OpenAI officially retired GPT-4o from ChatGPT on February 13, 2026, along with GPT-4.1, GPT-4.1 mini, and OpenAI o4-mini. The default model for all ChatGPT users is now GPT-5.2. Existing conversations that used GPT-4o will continue displaying previous responses, but all new messages route through the successor. The model that brought generative AI into mainstream consciousness lasted less than two years.
That timeline deserves attention. GPT-4o launched in May 2024 as OpenAI’s multimodal flagship, the model that could see, hear, and speak. It powered the voice mode that made ChatGPT feel like a conversation instead of a search bar. It was the model that crossed the chasm from early adopter curiosity to something your dentist asked about. And now it is gone, deprecated alongside three other models in a single support page update that reads like a firmware changelog.
Twenty Months From Flagship to Landfill
The compression of AI model lifecycles has no precedent in consumer technology. Microsoft supported Windows XP for thirteen years. Google maintained the original Gmail interface for nearly a decade. GPT-4o got twenty months. The replacement, GPT-5.2, had already captured the majority of ChatGPT usage before OpenAI pulled the trigger, according to the company’s deprecation notice. A fifth generation model made the fourth generation irrelevant not through a dramatic announcement but through quiet user migration. Most people switched without being told to.
This is the new rhythm. OpenAI also announced GPT-5.3 Codex this week, a model specifically designed for advanced coding tasks. The gap between model generations is no longer measured in years. It is measured in months. Each new release does not supplement the previous one. It replaces it entirely. The company is treating its own models the way fast fashion treats inventory: produce, ship, liquidate, repeat.
There is a version of this story that is straightforwardly good. GPT-5.2 is measurably better. Consumers get a superior product without lifting a finger. The companies that adapted fastest to GPT-4o’s capabilities will adapt fastest again. For the casual user, this deprecation is invisible progress. That reading is not wrong. It is just incomplete.
The pattern parallels a broader acceleration we are seeing across the industry. ByteDance, Alibaba, and DeepSeek all launched new models in early February, each one designed to leapfrog whatever existed the month before. The competitive dynamics are compressing timelines further. No company can afford to maintain an older model when a rival’s newer one is cheaper and more capable. The result is an industry where the product you built your business on can become obsolete before your annual contract renews.
The Developer Problem
For the hundreds of millions of casual ChatGPT users, this deprecation changes nothing. They were already on GPT-5.2 without knowing it. The real impact lands on the developers, enterprises, and startups that built workflows, fine tuned models, and integrated API calls around GPT-4o’s specific behavior. Every model has idiosyncrasies. Prompts that worked perfectly on GPT-4o may produce different outputs on GPT-5.2. Fine tuned models need retraining. Edge cases need retesting.
OpenAI’s API deprecation timeline is separate from the ChatGPT consumer deprecation, but the signal is the same: build on our platform and accept that the foundation shifts beneath you every few months. This is the trade off that every company using third party AI accepts, and it is one that Big Tech’s $650 billion AI capex bet is designed to lock in. The more infrastructure you build on someone else’s model, the harder it becomes to leave. The more frequently that model changes, the more dependent you become on the provider to keep things working.
The opposite argument has weight. Developers who treated GPT-4o as a permanent foundation were making a bet they should not have made. The documentation always warned that models would be deprecated. The companies that built abstraction layers, maintained model agnostic architectures, and tested across providers are fine today. The ones that hardcoded GPT-4o into production workflows chose convenience over resilience. That is a developer problem, not an OpenAI problem. Both readings contain truth. Which one you land on depends on whether you believe platform providers owe stability or whether users owe themselves adaptability.
The counterargument is that rapid deprecation is exactly what progress looks like. GPT-5.2 is measurably better than GPT-4o across every benchmark. Clinging to older models out of nostalgia or convenience slows the entire ecosystem. OpenAI’s willingness to kill its own darlings is precisely what makes it the market leader. The companies that survive are the ones that adapt to new models quickly, not the ones that demand backward compatibility forever. Every technology platform has upgrade cycles. AI’s are just faster.
The progress argument is valid, and it misses the point. Nobody is mourning GPT-4o’s capabilities. The concern is the business model underneath. OpenAI is training an entire economy to build on infrastructure it can unilaterally retire, and calling it innovation. Imagine if every commercial landlord could demolish your office building with thirty days notice and hand you a key to a different one across town. The new office might be better. You still lost everything on your walls. The companies that survive this era will not be the ones that build the best prompts. They will be the ones that build the thickest insulation between their products and the model provider’s deprecation schedule. The rest are renting intelligence on someone else’s terms and calling it a strategy.
What This Means for Everyday People
If you use ChatGPT, your experience just got better and you probably did not notice. GPT-5.2 is faster, more accurate, and handles complex reasoning more reliably than GPT-4o did. The transition was designed to be invisible.
The deeper implication is about control. When the tools you rely on can change overnight without your input, you are not a customer. You are a passenger. The AI companies are driving, and the destination changes whenever their engineering team ships a new model. For businesses, that means building contingency into every AI integration. For individuals, it means understanding that the AI assistant you are talking to today will not be the same one you are talking to in six months. The personality, the quirks, the way it phrases things, all of it is disposable. The only constant is the subscription fee.
This analysis assumes two things worth questioning. First, that the pace of deprecation will continue or accelerate. It is possible that model improvements plateau and lifecycles stabilize, the way smartphone upgrade cycles eventually slowed from annual breakthroughs to incremental refinements. Second, that dependency on a single provider is the default path. Open source models from Meta, Mistral, and others offer an alternative for companies willing to trade convenience for control. Whether that trade off is worth it depends on how much you trust OpenAI’s deprecation schedule to align with your business needs. That question does not have a universal answer.
For inquiries and analysis contact laterstack@proton.me
Infleqtion, the Boulder, Colorado quantum computing company, completed its merger with Churchill Capital Corp X on February 13, 2026, and will begin trading on the New York Stock Exchange under the ticker INFQ on Monday, February 17. The deal values the company at approximately $1.8 billion with over $550 million in gross proceeds. It is the newest addition to a small but increasingly crowded class of publicly traded quantum companies, arriving at a moment when the sector’s credibility is being tested in real time.
The Infleqtion listing is a neutral atom quantum computing play. Unlike IonQ, which uses trapped ion technology, or IBM, which builds superconducting circuits, Infleqtion’s platform manipulates individual atoms suspended in optical traps using precisely tuned lasers. The approach has theoretical advantages in scalability and parallelism, and recent breakthroughs in neutral atom error correction have strengthened the case for the modality. Infleqtion’s customer list includes NVIDIA, the U.S. Department of Defense, and NASA. The company also advanced to the final stage of the Wellcome Leap Q4Bio Challenge, securing $2 million to validate a quantum enabled biomarker discovery platform for oncology in partnership with the University of Chicago and MIT.
Those credentials read well on paper. The question is whether the public market is willing to pay for them.
The SPAC Problem
The vehicle matters. SPACs, special purpose acquisition companies, have a dismal post merger track record across every sector. According to data compiled by SPAC Research, the median SPAC that completed a merger between 2020 and 2023 was trading below $5 per share within eighteen months of its business combination. Quantum computing is not exempt. IonQ went public via SPAC in 2021. D-Wave went public via SPAC in 2022. Rigetti went public via SPAC in 2022. All three experienced significant post merger declines.
Infleqtion’s SPAC partner, Churchill Capital Corp X, is run by Michael Klein, a serial SPAC sponsor whose prior vehicles include the Churchill Capital Corp IV deal that merged with Lucid Motors in 2021 at a $24 billion valuation. Lucid traded below $2 per share by late 2024. The Klein name carries history, and not the kind that inspires confidence in long term value creation for public shareholders.
The timing compounds the concern. Infleqtion arrives on the NYSE the same week IonQ is trading near 52 week lows, down roughly 10% in five days following the Wolfpack Research short seller report published on February 4, 2026. Wolfpack alleged that 86% of IonQ’s revenue came from Pentagon earmarks that were subsequently canceled. IonQ’s stock has been in freefall ahead of its February 25 earnings call. Meanwhile, Quantinuum, the Honeywell backed quantum company that many consider the sector’s strongest player, filed its confidential S-1 in January 2026 for a traditional IPO targeting a $20 billion plus valuation. Quantinuum projects $2 billion in revenue. IonQ reported $24 million. Infleqtion’s revenue figures have not been disclosed at this scale.
The public quantum market is expanding in headcount while contracting in credibility. More tickers. More scrutiny. Less patience.
The charitable reading is that competition validates the sector. More public quantum companies means more capital, more talent, more engineering hours dedicated to solving the hardest problems in computing. IonQ’s troubles may be company specific, not sector defining. Infleqtion uses a fundamentally different technology. Painting all quantum SPACs with the same brush risks dismissing legitimate differences in approach, team, and execution. That reading deserves space alongside the skepticism.
Claim Versus Capability
Infleqtion has legitimate technology. Its Sqorpius portable quantum sensor has defense applications that go beyond computation into navigation, timing, and threat detection. The company’s neutral atom approach avoids some of the fabrication challenges that plague superconducting systems. And the Q4Bio oncology work, if validated, would represent one of the first genuine quantum advantages in life sciences.
But “if validated” is doing enormous work in that sentence. No neutral atom quantum computer has demonstrated a commercially meaningful quantum advantage over classical systems in production. Infleqtion’s technology is real. Its commercial viability at the stated valuation is unproven. The $1.8 billion price tag is a bet on future capability, not current revenue. That is the same bet the market made on IonQ. The same bet the market made on D-Wave. Both are still trying to prove it was justified.
The question investors need to answer is not whether neutral atom quantum computing works in the lab. It does. The question is whether a $1.8 billion public company is the right vehicle for technology that may need another five to ten years of development before it generates revenue at scale. The broader quantum sector’s answer to that question, as we have examined at length, remains deeply uncertain.
It is worth stating the assumptions embedded in that uncertainty. This analysis assumes the timeline to commercial quantum advantage is long, that SPAC track records are predictive of future SPAC outcomes, and that Infleqtion’s undisclosed revenue numbers are modest relative to the valuation. If any of those assumptions prove wrong, specifically if Infleqtion’s defense sensor contracts generate meaningful near term revenue, or if neutral atom architecture reaches error correction milestones faster than expected, the calculus changes entirely. The Wellcome Leap oncology work with UChicago and MIT, if it produces publishable results, would be among the first peer reviewed demonstrations of quantum advantage in life sciences. That alone could rewrite the valuation case. The smart move is to watch the science, not just the ticker.
The counterargument is that public markets are exactly where quantum companies should be. Private funding rounds create valuation distortions and lock up capital for years. Public listing forces transparency through quarterly reporting, subjects claims to market scrutiny, and gives institutional investors the ability to price risk in real time. Infleqtion going public via SPAC is not a red flag. It is capital efficiency. The company gets funded, investors get liquidity, and the market gets to decide what the technology is worth. Every quantum company that stays private is just delaying that reckoning.
The detail nobody is talking about is the vehicle itself. Quantinuum filed for a traditional IPO with Morgan Stanley and JPMorgan underwriting. Infleqtion went the SPAC route with a serial SPAC sponsor. That is not a neutral choice. Companies that can raise through traditional IPOs do. Companies that cannot, or whose numbers cannot survive the underwriting scrutiny, go SPAC. The technology might be legitimate. The choice of financial vehicle tells you the company’s own bankers had doubts about whether institutional investors would pay this price through the front door. Watch how INFQ trades in its first thirty days. If it follows the SPAC median, it will be below $5 by summer. If it holds, it will be one of the few exceptions that proves the rule. Either way, quantum computing’s public market chapter is being written by financial engineers as much as by physicists, and that should make everyone pay closer attention to the cap table than the qubit count.
What This Means for Everyday People
Quantum computing remains a technology that most people will never interact with directly. But the financial infrastructure being built around it affects everyone. When SPACs take speculative technology public, retail investors often buy the narrative before the revenue exists. The enthusiasm is real. The returns, historically, are not.
If you own index funds, you already have indirect quantum exposure through Google, Microsoft, and IBM. Those companies can absorb quantum R&D costs across diversified businesses. Pure play quantum stocks like INFQ, IONQ, and eventually Quantinuum carry concentrated risk in a sector that has not yet proven it can generate consistent commercial revenue. The gap between scientific capability and investable business remains the central tension in quantum computing. Infleqtion’s public debut does not resolve that tension. It monetizes it.
For inquiries and analysis contact laterstack@proton.me
Sapiom, a San Francisco startup building financial infrastructure for AI agents, raised a $15 million seed round on February 12, 2026. The round was led by Accel, with strategic participation from Okta Ventures, Gradient Ventures (Google’s AI fund), Array Ventures, Menlo Ventures, Anthropic, and Coinbase Ventures. The company was founded by Ilan Zerbib, a former engineering lead at Shopify, and is designed to solve a problem that sounds mundane until you think about it: AI agents cannot handle money.
That gap is about to matter. OpenAI launched its enterprise agent platform, Frontier, earlier this month with customers including HP, Oracle, State Farm, and Uber. Anthropic’s Claude agents are being deployed across customer service, research, and operations. Decagon raised $250 million at a multibillion dollar valuation to build AI agents for enterprise customer support. The agent economy is scaling fast. But every one of these systems hits a wall the moment a task requires a financial transaction. An AI agent can research a vendor, draft a purchase order, and get manager approval. It cannot pay the invoice.
Sapiom is building the layer that sits between the agent and the financial system. The infrastructure handles payment processing, account management, and transaction authorization for autonomous AI workflows. Think of it as the plumbing that allows an agent to hold a balance, execute a transfer, and maintain an auditable record of every dollar it touches.
The Investor List Tells the Story
The cap table is more revealing than the check size. Anthropic builds the AI models that power many of these agents. Gradient Ventures is Google’s AI investment arm, backing the ecosystem that connects Google’s models to real world tasks. Coinbase Ventures operates in the infrastructure layer where digital assets and programmable money intersect. Okta Ventures provides identity and authentication, the access control layer that determines what an agent is authorized to do.
These are not general purpose venture funds chasing the AI theme. These are the companies building the agent stack, and they are investing in Sapiom because they know their own products will need it. When the model provider, the identity layer, and the crypto infrastructure company all back the same seed stage fintech startup, they are pre wiring the plumbing for a system they expect to exist. The scale of capital flowing into AI infrastructure confirms this is not speculative. It is architectural.
The skeptical read of that same cap table: large platform companies invest in dozens of seed stage startups as option value, not conviction. Anthropic writing a seed check does not mean Anthropic believes Sapiom will become a pillar of the agent economy. It means Anthropic spent a small amount of money to maintain optionality on a category that might matter. Venture portfolios are built on the assumption that most bets fail. Reading strategic intent into a seed round requires distinguishing signal from spray, and at this stage, both explanations fit the evidence equally well.
The obvious risk is that this is a feature, not a company. Stripe already processes trillions in payments. Plaid connects applications to bank accounts. Both have the engineering resources and market position to add an AI agent layer to their existing infrastructure. If Stripe ships “Stripe for Agents” in six months, Sapiom’s entire product becomes redundant. The history of fintech is littered with startups that identified real infrastructure gaps only to watch the incumbents close those gaps with a single product launch. A $15 million seed buys time, not a moat.
The regulatory dimension is equally unresolved. Autonomous AI agents making financial transactions raises questions about liability, fraud prevention, and consumer protection that no regulator has answered. If an agent initiates a payment that turns out to be fraudulent, who is responsible? The agent’s owner? The model provider? The infrastructure company that processed the transaction? These questions will be answered by lawsuits, not whitepapers.
Stripe could absolutely eat this for lunch. That is the wrong reason to dismiss it. The signal here is not the product. It is the cap table. When Anthropic writes a check into a seed round for agent financial infrastructure, they are telling you they expect their own models to need this capability and they would rather fund a dedicated startup than build it themselves. That is the clearest market validation a seed stage company can get. The real play is not whether Sapiom survives. It is that the AI agent economy has reached the point where the biggest model providers are already planning for agents that spend money. That is a structural shift, not a startup story. Whether Sapiom or Stripe or some third player captures the infrastructure layer is a competitive question. The fact that the layer needs to exist at all is the headline.
What This Means for Everyday People
The near term impact is invisible. Sapiom is building infrastructure that other companies will use, not a product consumers will interact with directly. But the downstream effects are significant. When AI agents can handle money autonomously, the services built on top of them change fundamentally. Your insurance claim gets processed without a human touching it. Your subscription gets optimized by an agent that can cancel, renegotiate, and repurchase on your behalf. Your business expenses get categorized, approved, and paid without anyone opening an app.
The tradeoff is control. Every layer of automation between you and your money is a layer of abstraction you have to trust. The companies building that trust layer, Sapiom included, are betting that convenience will win. History suggests they are right. Whether that is good for consumers depends entirely on who writes the rules for what these agents are allowed to do with your money. Right now, nobody has.
This analysis rests on one core assumption that deserves scrutiny: that the AI agent economy will scale to the point where autonomous financial transactions become routine. If agents remain primarily informational, answering questions and drafting documents rather than executing real world transactions, the entire financial infrastructure layer becomes unnecessary. The bet is that agents will graduate from assistants to operators. That graduation is not guaranteed. It depends on trust, regulation, technical reliability, and consumer willingness to let software spend their money without asking first. Every one of those dependencies is unresolved.
For inquiries and analysis contact laterstack@proton.me
Four companies are about to spend more on artificial intelligence infrastructure in a single year than most nations spend on their entire economies. Amazon, Alphabet, Meta, and Microsoft have collectively committed between $635 billion and $665 billion in capital expenditure for 2026, a figure that represents a 67 to 74 percent increase over the $381 billion they spent in 2025. The Big Tech AI spending spree has no precedent in corporate history, and the question it raises is deceptively simple: is this the greatest infrastructure investment since the railroads, or the greatest misallocation of capital since the dot com era?
Bloomberg called the combined figure “essentially unprecedented in modern economic history, even accounting for the telecom boom of the 1990s and the construction of railroads.” CNBC reported that total spending approaches $700 billion when factoring in additional operational expenses. Fortune noted that the sum rivals the entire gross domestic product of Sweden. These are not comparisons made for dramatic effect. They are the most accurate analogies available, and even they may understate the scale.
The Numbers Behind the Conviction
Amazon leads the pack with a staggering $200 billion commitment, a figure that sent its stock plunging more than 11 percent in extended trading before closing down 5.55 percent on the day of the announcement. CEO Andy Jassy justified the spending by citing “very high demand” for AI compute capacity. The market’s response was telling: AWS revenue grew 24 percent to $35.6 billion, the fastest growth rate in 13 quarters, and investors still punished the stock. Strong demand was not the concern. The concern was whether any level of demand could justify spending at this velocity.
Alphabet committed between $175 billion and $185 billion, a figure that stunned Wall Street analysts who had expected approximately $119.5 billion, nearly double its 2025 capital expenditure. The stock whipsawed after earnings before ultimately recovering, a pattern that suggests the market is uncertain rather than hostile. Meta announced a range of $115 billion to $135 billion, again nearly doubling its prior year. Microsoft, based on its quarterly figures, is running at approximately $145 billion annualized.
Analysts at Barclays acknowledged that infrastructure costs of this magnitude will weigh on near term profitability but argued that cloud growth combined with progress from divisions like DeepMind “start to justify” the expenditure. The operative word is “start.” That is not a ringing endorsement. It is a hedge dressed as optimism.
The Railroad Analogy and Its Limits
The comparison to railroads is instructive, though not in the way its proponents intend. The railroad boom of the 19th century did transform the American economy. It also produced spectacular failures, fraudulent financing, and decades of overcapacity. Many of the companies that built the railroads went bankrupt. The infrastructure they created was valuable. The equity they issued was often worthless. The people who profited most were those who used the railroads, not those who financed them.
The dot com parallel cuts deeper. In the late 1990s, telecommunications companies spent hundreds of billions laying fiber optic cable based on demand projections that proved roughly correct but arrived on a timeline that bankrupted the builders. WorldCom and Global Crossing constructed the future and then collapsed into it.
The question is not whether AI will be valuable. It almost certainly will be. The question is whether the companies spending $650 billion will capture sufficient returns, or whether they are building infrastructure whose value accrues primarily to the customers who use it rather than the shareholders who financed it.
This Is an Arms Race
This looks like an arms race because it is one. None of these companies can afford to stop spending. Amazon cannot let Azure capture enterprise AI workloads while AWS builds capacity. Microsoft cannot let Google’s DeepMind models run on competitors’ hardware. Meta cannot afford to rent someone else’s AI infrastructure while building its own models. The logic is not “this will definitely pay off.” The logic is “the cost of being wrong about spending is lower than the cost of being wrong about not spending.” When every competitor reaches the same conclusion, the result is a $650 billion pile of chips on the table.
The dot com comparison is useful but lands somewhere in the middle. The telecom companies that laid fiber in the 1990s were right about demand and wrong about timing. The infrastructure they built was valuable. The companies that built it went bankrupt. The same pattern could repeat. The critical difference is that Amazon, Alphabet, Meta, and Microsoft have balance sheets that can absorb years of spending before profitability breaks. WorldCom did not. These companies will likely lose money on AI infrastructure in the short term and earn it back over the long term, because the use cases are not hypothetical. Enterprise AI adoption is accelerating. The compute has to live somewhere.
The deeper strategic play that most coverage misses is how this AI infrastructure connects to the next wave. Quantum computing requires classical compute for error correction, simulation, and hybrid algorithms. The companies building the largest AI compute footprints today are also positioning themselves as the infrastructure layer for quantum computing tomorrow. Amazon’s Braket, Google’s quantum lab, Microsoft’s Azure Quantum. The $650 billion is not just an AI bet. It is a multi decade infrastructure play that extends into quantum, robotics, and whatever comes after. That does not make it wise. It makes it unavoidable.
The Labor Market Disconnect
The spending announcements arrived against a labor market backdrop that makes the scale of investment more unsettling. ADP reported that the U.S. economy added only 22,000 jobs in January, a figure far below expectations. Corporate layoffs sit at levels not seen since 2009. The pattern emerging across American industry is one where companies are simultaneously increasing capital investment in AI systems and reducing investment in human workers, a dynamic that previous analysis on this site has examined in detail.
This is not a coincidence. It is a strategy. The capital flowing into AI infrastructure is explicitly intended to automate tasks currently performed by humans. The companies spending $650 billion are not adding this capacity alongside their existing workforce. They are adding it, in significant part, as a replacement for portions of that workforce. The physical constraints gating AI progress make the buildout even more capital intensive, as data centers, power generation, and chip fabrication all require massive upfront investment before returns materialize.
The macroeconomic implications are significant. When corporations redirect hundreds of billions from labor to capital, wages deflate while asset prices inflate. Workers lose bargaining power. Shareholders capture the gains. The distribution of economic benefit shifts further toward capital owners, a trend accelerating for decades and now receiving a $650 billion accelerant.
The Sovereignty Dimension
There is a geopolitical layer to this spending that receives insufficient attention. When four American companies control the majority of the world’s AI compute infrastructure, they become instruments of American strategic power whether they intend to or not. France’s decision to ban American technology platforms from government use reflected precisely this concern: that dependence on American digital infrastructure creates vulnerabilities that sovereign nations cannot accept.
A world in which AI compute is concentrated in data centers owned by Amazon, Alphabet, Meta, and Microsoft is a world in which access to artificial intelligence is mediated by American corporate decisions and American legal frameworks. The CLOUD Act, sanctions enforcement, and export controls all apply. The $650 billion is not just a corporate investment. It is the construction of a strategic asset with implications that extend far beyond quarterly earnings.
What This Means for Everyday People
For workers, the message embedded in these capital expenditure figures is blunt. The largest and most powerful companies on earth have decided that their future involves substantially more machines and substantially fewer people. That decision has already been made. The $650 billion is the execution budget.
For consumers, the near term effect will be improved AI products and services. More compute capacity means faster models, better applications, and lower prices for AI tools.
For citizens, the longer term questions are harder. Who governs an economy in which four companies control the infrastructure that powers artificial intelligence? What happens when these systems displace workers faster than the economy can absorb them? What oversight exists for investment decisions that will reshape labor markets, energy consumption, and the distribution of economic power for decades?
These are not questions that earnings calls answer. They are questions that societies must answer, and the $650 billion clock is already ticking.
For inquiries and analysis contact laterstack@proton.me
Bedrock Robotics, the San Francisco startup built by former Waymo engineers, closed a $270 million Series B on February 4, 2026, to bring autonomous construction equipment from supervised prototype to fully unmanned deployment. The round, jointly led by CapitalG (Google’s growth fund) and Valor Atreides AI Fund, pushes Bedrock’s total funding past $350 million and represents one of the largest bets yet that the same AI systems that learned to navigate city streets can learn to move earth.
The investor list reads like a coordinated thesis on physical AI: NVentures (NVIDIA’s venture arm), 8VC, Eclipse, Emergence Capital, Perry Creek Capital, Tishman Speyer, MIT, Georgian, and others. When a real estate developer, a semiconductor giant, and a university endowment all write the same check, they are not chasing software margins. They are pricing in a conviction that autonomy’s next frontier is measured in cubic yards, not click rates.
The Team That Built the Playbook
The founding team matters here more than usual. Boris Sofman, Bedrock’s CEO and cofounder, previously cofounded Anki, the consumer robotics company that shipped over 3.5 million devices and raised more than $200 million before shutting down in 2019. He then spent roughly five years at Waymo as Director of Engineering and Head of Trucking, leading the autonomous freight program through its expansion into major U.S. cities. Cofounder and CTO Kevin Peterson also came from Waymo. The remaining cofounders, Ajay Gummalla and Tom Eliaz, both VPs of Engineering, round out a leadership bench that has collectively shipped autonomous systems at commercial scale.
Sofman’s trajectory tells a particular story. He built consumer robots, watched that company die from funding starvation, then spent five years inside Google’s most ambitious physical AI project. Now he is applying everything he learned to construction, an industry where the labor economics practically beg for automation. “The construction industry is being asked to build more than it can deliver,” Sofman said in the company’s Series B announcement. “Contractors are pulled across competing priorities with the same limited workforce and equipment.”
An Industry Bleeding Workers
The numbers confirm what Sofman describes. According to the Associated General Contractors of America, 92% of construction firms report difficulty finding workers, with 45% citing labor shortages as the primary cause of project delays. The ITIF reported in January 2026 that the industry faces a shortage of roughly 439,000 workers, driven in large part by the explosion of data center construction. Deloitte’s 2026 Engineering and Construction Outlook projects the gap will widen to 499,000 unfilled positions this year.
Construction also remains the deadliest sector in American industry by total fatalities, with 1,075 worker deaths recorded in 2024 according to OSHA data, accounting for 19% of all U.S. workplace fatalities. Over 60% of construction accidents occur within a worker’s first year on the job. This is not an industry resisting technology out of preference. It is an industry running out of people.
The pattern connects directly to a broader constraint we examined in our analysis of how physical limitations are gating AI’s real progress. Software intelligence is abundant. The bottleneck is getting that intelligence into the physical world, into machines that dig, pour, and grade. Bedrock is positioning itself exactly at that bottleneck.
What They Have Actually Built
Bedrock’s approach is retrofit, not replacement. The company installs a hardware rack on top of existing excavator cabs, equipped with LiDAR, GPS, inertial measurement units, eight high definition cameras, and an onboard computer. The system works across multiple excavator models from 20 ton to 80 ton machines. In November 2025, Bedrock partnered with Sundt Construction on what the company calls the construction industry’s largest known supervised autonomy deployment: mass excavation on a 130 acre manufacturing facility site in Phoenix, Arizona. The autonomous systems moved over 65,000 cubic yards of material by loading human operated articulating dump trucks using the same workflow as manual operations.
The company is now targeting its first fully unmanned excavator deployments with customers in 2026. If that timeline holds, Bedrock will have gone from stealth to unmanned commercial operation in under two years.
The Legal and Political Minefield
The Waymo pedigree looks strong on a pitch deck. It looks less convincing when you consider that Waymo itself is navigating wrongful death litigation, regulatory scrutiny in multiple states, and an operating model that still requires significant human oversight in edge cases. The self driving playbook transfers, but so do the liabilities. An 80 ton excavator operating without a human in the cab presents a legal and political exposure that no amount of venture capital can engineer away. OSHA has no framework for unmanned heavy equipment on active construction sites. Who is responsible when one of these machines kills someone? That question does not have an answer yet.
The political dimension cuts in two directions. The current administration’s infrastructure spending, data center buildout, and manufacturing reshoring create enormous tailwinds for autonomous construction. More projects, more demand, more urgency. On the other hand, these are the same political forces pushing for American jobs and union labor. Autonomous excavators do not pay union dues. They do not vote. The companies deploying them will face the same political friction that has dogged autonomous vehicles, except the workers being displaced wear hardhats and carry union cards, not rideshare apps.
The pattern is one we have tracked before in detail. Venture capital funds the technology that displaces the workforce, then frames the displacement as solving a shortage. The shortage is real today. It will not be real forever. When autonomous machines become cheaper per cubic yard than human operators, the narrative will shift from “we cannot find workers” to “we do not need workers.” The gains flow to the capital owners who funded the machines. The losses flow to the operators who used to drive them. The financial logic follows a similar arc to what private equity has done in other industries, as we documented with Snout’s $110 million raise to address a veterinary crisis that consolidation itself created. First you create the conditions for a shortage. Then you fund the solution to the shortage you created. Then you capture the margins on both sides.
What happens to the displaced operators is the question nobody on the cap table has an incentive to answer. The construction worker who used to run a $500,000 excavator does not become a software engineer. He becomes a gig worker or an unemployed statistic. The gains are real. The PR for everyday Americans is terrible.
What This Means for Everyday People
For general contractors hemorrhaging money to delays and labor gaps, autonomous excavation could compress timelines and reduce the single largest variable cost on a job site. For consumers, faster construction means lower costs on housing, infrastructure, and commercial development. Those are real benefits that affect real people.
For construction workers, the picture is darker than the investor presentations suggest. Bedrock frames autonomy as solving a shortage, not replacing a workforce. That framing holds as long as the shortage persists. The moment it doesn’t, every financial incentive in the system points toward fewer humans and more machines. The rich get richer. The workers who built the country’s infrastructure get a pink slip and a LinkedIn notification suggesting they learn to code.
The construction industry accounts for roughly $1.4 trillion in annual U.S. spending. It is one of the least digitized sectors in the global economy. The talent pipeline is shrinking, the safety record is grim, and the demand curve, driven by data centers, manufacturing reshoring, and infrastructure legislation, points in only one direction. Bedrock is not the first company to promise autonomous construction. Built Robotics started autonomous excavator trials in 2017 and eventually narrowed its focus to solar farm installation. Caterpillar has been running semi autonomous field trials for years without reaching full autonomy in construction applications.
What separates Bedrock is the pedigree of the team, the speed of execution, and the size of the capital behind them. Whether that is enough to solve a problem that has humbled every prior attempt is the $350 million question. The dirt will tell.
For inquiries and analysis contact laterstack@proton.me
The Quantinuum IPO represents something the quantum computing industry has not yet produced: a company confident enough in its fundamentals to pursue a traditional initial public offering. On January 14, 2026, Quantinuum, the trapped ion quantum computing subsidiary majority owned by Honeywell International, filed a confidential S-1 registration statement with the Securities and Exchange Commission. The company is expected to seek a valuation north of $20 billion and raise approximately $1 billion in proceeds, numbers that would make this the largest quantum computing capital markets event in the industry’s history.
Every other quantum company that has gone public chose a different path. IonQ, D-Wave, and Rigetti all reached public markets through SPAC mergers between 2021 and 2022, a mechanism that allowed them to make forward looking revenue projections that traditional IPOs prohibit. Quantinuum’s decision to file a conventional S-1 is a deliberate statement: the company believes its financials can withstand the scrutiny of a full SEC registration process without the narrative scaffolding that SPACs provide.
That distinction matters more than it might appear.
The Hardware Advantage
Quantinuum was formed in 2021 through the merger of Honeywell Quantum Solutions and Cambridge Quantum, combining Honeywell’s precision manufacturing expertise with Cambridge Quantum’s software and algorithms capabilities. The resulting entity operates a full stack trapped ion platform built on what the company calls QCCD architecture, or quantum charge coupled device, which physically shuttles individual ions within the processor to perform gate operations.
The company’s Helios quantum computer currently operates 98 qubits with single qubit gate fidelity of 99.9975% and two qubit gate fidelity of 99.921%. Those fidelity numbers are among the highest published by any quantum computing company. More significantly, Quantinuum has demonstrated 48 fully error corrected logical qubits, a milestone that moves the conversation from theoretical error correction to operational error correction.
Here is what the claim actually means in practice. Forty eight logical qubits is a genuine engineering milestone. It is also not close to commercially useful scale. Applications in drug discovery, cryptographic analysis, and materials simulation require error corrected systems orders of magnitude larger. The capability is real. The distance between here and commercial utility is also real. Investors pricing $20 billion are not buying today’s 48 qubits. They are buying a bet that Quantinuum closes that gap before competitors do and before patience runs out.
The Valuation Question
Quantinuum’s last private fundraising round valued the company at approximately $10 billion on a pre money basis. That $600 million round drew participation from NVIDIA NVentures, Amgen, and JPMorgan, a mix of strategic and financial investors that signals broad confidence across technology, pharmaceutical, and financial sectors. The IPO is expected to roughly double that valuation.
Whether $20 billion is justified depends entirely on your time horizon and your assumptions about quantum computing’s commercial trajectory. The company employs several hundred people across the United States, United Kingdom, Germany, and Japan. Revenue figures remain undisclosed pending the S-1 becoming public, which makes the valuation a bet on capability and positioning rather than current earnings.
Compare this to the public quantum companies. IonQ, currently the largest publicly traded pure play quantum company, projected approximately $109 million in 2025 revenue while trading at a market capitalization that fluctuates around $8 billion. IonQ also faces pressure from a Wolfpack Research short report alleging that roughly 86% of its revenue derived from Pentagon contracts that were subsequently canceled. Whether those allegations prove accurate, they illustrate the fragility of quantum company revenue at this stage.
D-Wave and Rigetti, the other public quantum companies, trade at substantially lower valuations and continue to generate modest revenue relative to their R&D expenditures. None of the SPAC quantum companies has delivered the post listing performance their sponsors projected.
The S-1 filing arrives at a moment that deserves scrutiny. Quantinuum’s closest public competitor is under short seller attack. The broader SPAC quantum cohort has underperformed. Filing now, through a traditional IPO, positions Quantinuum as the credible alternative at exactly the moment investors are looking for one. That is good strategy. It is also not the same thing as having the strongest technology. The filing is as much about market positioning as it is about technical readiness.
Why Honeywell’s Backing Changes the Calculus
The Honeywell relationship is not cosmetic. Honeywell’s precision manufacturing capabilities, honed over decades of producing aerospace and defense components to exacting tolerances, provide Quantinuum with fabrication advantages that no pure play startup can replicate. Trapped ion quantum computers require extraordinary precision in ion trap manufacturing. Honeywell’s infrastructure provides that precision at a level that competitors building their own manufacturing from scratch cannot easily match.
Honeywell’s majority ownership also provides Quantinuum with something its public competitors lack: a parent company with $36 billion in annual revenue, a stable balance sheet, and no existential dependence on quantum computing succeeding on any particular timeline. If quantum commercialization takes longer than optimists expect, Quantinuum has a backstop. IonQ, D-Wave, and Rigetti do not.
This structural advantage may matter more than any technical benchmark. Quantum computing remains a capital intensive field where the timeline to profitability is measured in years, not quarters. The companies most likely to survive are those with the longest financial runway, and Honeywell’s backing gives Quantinuum arguably the longest runway in the industry.
The Broader Signal
Quantinuum’s IPO filing arrives at a moment when quantum computing investment is accelerating across multiple geographies. Europe has committed tens of millions through initiatives like the SUPREME consortium to build domestic quantum manufacturing capability. Australia is backing companies like Diraq with sovereign investment conditions to prevent quantum IP from migrating offshore. Academic breakthroughs at institutions like Stanford continue pushing the theoretical ceiling for qubit counts and architectures.
A successful Quantinuum IPO would validate the thesis that quantum computing has matured enough to attract mainstream public market capital, not just venture speculation and government grants. It would also establish a valuation benchmark against which every other quantum company, public or private, will be measured.
A $20 billion valuation against undisclosed revenue in an industry that generated roughly $1.5 billion total in 2025. The sentiment has moved ahead of the feasibility timeline. That does not make the investment wrong. It makes the risk asymmetric. If quantum computing delivers on a 5 to 7 year horizon, this valuation looks prescient. If the timeline stretches to 10 or 15 years, the capital patience required will outlast most investors’ willingness to wait. The market is pricing conviction. Whether that conviction is justified is a question the S-1 numbers will begin to answer, but only begin.
What This Means for Everyday People
For most people, quantum computing remains an abstraction. The Quantinuum IPO will not change daily life. But the capital flows it represents will shape which companies survive long enough to deliver quantum applications in drug discovery, materials science, financial modeling, and cryptography.
If public market investors embrace this offering, it accelerates the entire field. More capital means more hardware development, more software investment, more hiring. If they reject it, the message is that quantum remains a venture stage technology unready for mainstream investment. That outcome would slow commercial development and consolidate the industry around fewer, better capitalized players.
The stakes for the quantum industry are straightforward. Quantinuum’s IPO is not just a liquidity event for Honeywell and early investors. It is a referendum on whether quantum computing has crossed the threshold from laboratory science to investable industry.
For inquiries and analysis contact laterstack@proton.me
For the first time, major corporations are saying the quiet part loud.
Amazon announced plans to reduce its corporate workforce by 16,000 employees. Pinterest cut up to 15% of its workforce. Meta’s Mark Zuckerberg declared that 2026 will be when “AI starts to dramatically change the way that we work.” Across the technology industry and beyond, companies explicitly cited artificial intelligence as a driver of layoffs that affected more than 50,000 workers in 2025 alone.
This is new. Previous waves of technology-driven displacement were discussed euphemistically. Automation. Efficiency gains. Workforce optimization. The machines were taking jobs, but companies rarely said so directly. The social contract held that corporations would not publicly attribute job losses to technology, even when technology was clearly responsible.
That contract has broken.
But before accepting the narrative that AI is now actively replacing human workers at scale, we must ask a harder question: Is it true? Are these layoffs genuinely driven by AI capabilities that now exceed human performance? Or is something else happening, something that tells us as much about corporate communications and investor relations as it does about artificial intelligence?
The Case That AI Is Real
Let us begin with the bull case, because it is not without merit.
Generative AI has advanced remarkably over the past three years. Large language models can now write code, draft marketing copy, summarize documents, conduct research, and perform dozens of tasks that previously required human labor. Tools like GitHub Copilot demonstrably increase developer productivity. Customer service chatbots handle inquiries that once required human agents. Content generation systems produce text at scales impossible for human writers.
Companies deploying these tools are reporting productivity gains. Klarna, the payments company, has stated that its AI assistant performs the work equivalent of 700 full-time customer service agents. If AI can do the work of hundreds of employees, why would companies continue paying those employees?
The logic is straightforward. Technology substitutes for labor. This has been true since the spinning jenny. AI is the latest instance of a pattern that has repeated throughout the history of capitalism.
From this perspective, the 50,000 layoffs explicitly attributed to AI represent an honest acknowledgment of technological progress. Companies are admitting what has always been true but was previously impolite to say: technology replaces workers.
The Case for Skepticism
The bear case is equally compelling.
Forrester, the technology research firm, published a report in January arguing that many companies announcing AI-related layoffs “do not have mature, vetted AI applications ready to fill those roles.” The report identifies a trend of “AI-washing,” companies attributing financially motivated cuts to future AI implementation that may or may not materialize.
Molly Kinder, a senior research fellow at the Brookings Institution, noted that saying layoffs were caused by AI is a “very investor-friendly message.” The alternative, admitting that a business is struggling or that management made poor decisions, is far less palatable to shareholders.
Consider the incentives. Wall Street loves AI narratives. Companies that credibly claim to be AI-forward receive valuation premiums. Stock prices respond positively to announcements about AI adoption. In this environment, attributing layoffs to AI efficiency gains rather than business weakness is not merely acceptable. It is strategically optimal.
The skeptical view holds that AI is being used as cover. Some portion of these layoffs would have happened regardless of AI capabilities. Economic conditions, competitive pressures, overcapacity from pandemic hiring, and strategic missteps all contribute to workforce reductions. AI provides a socially acceptable explanation that simultaneously signals technological sophistication and deflects blame from management.
The Truth Is Probably Both
Reality rarely conforms to clean narratives. The honest assessment is that both factors are at work simultaneously.
Some jobs are genuinely being replaced by AI systems that now perform tasks better, faster, or cheaper than humans. The customer service representative whose role consisted primarily of answering routine questions from a knowledge base is vulnerable. The junior copywriter producing formulaic content at volume is vulnerable. The data entry clerk transcribing information from one system to another is vulnerable.
Simultaneously, some layoffs attributed to AI are opportunistic. Companies facing pressure to reduce headcount can now cite AI as the reason rather than acknowledging overcapacity or strategic failure. The narrative is convenient. Investors reward it. The workers lose their jobs either way, but the framing matters for corporate reputation and stock price.
Distinguishing genuine AI displacement from AI-washing in any specific case is difficult. Companies do not provide the detailed productivity data that would allow external observers to verify their claims. We see the announcements. We do not see the internal analysis, if any, that preceded them.
The Policy Question
For policymakers, the uncertainty creates a dilemma. If AI is genuinely displacing workers at scale, policy responses are urgent. Workforce retraining programs, social safety net expansions, and potentially more radical interventions may be necessary to manage a transition whose speed exceeds historical precedents.
If AI displacement is overstated, different interventions are appropriate. The focus should be on counter-cyclical employment policy, supporting workers affected by economic downturns rather than technological transformation, and resisting the temptation to create programs addressing a problem that exists more in corporate communications than in labor market reality.
The policy response depends on accurate diagnosis. That diagnosis is currently impossible because companies have incentives to exaggerate AI’s role and no obligation to provide verifiable data.
A pragmatic approach acknowledges uncertainty while preparing for multiple scenarios. Invest in workforce adaptation programs that help workers transition regardless of whether the cause is AI or traditional economic displacement. Monitor labor market data for patterns that would distinguish AI-driven structural change from cyclical fluctuations. Require more transparency from companies making AI-related layoff claims, perhaps through enhanced disclosure requirements in securities filings.
The worst outcome would be either complacency (assuming AI displacement is overstated and being caught unprepared) or panic (overreacting to a narrative that serves corporate interests more than it reflects reality).
The Moral Question
Beyond policy, there is a moral question that both left and right should consider.
From the left, the concern is worker welfare. Fifty thousand people lost their jobs. Whether AI or management failure or economic conditions caused those job losses, the human impact is real. Families face disruption. Communities lose economic activity. The psychological toll of unemployment falls on individuals regardless of the cause.
From the right, the concern is honesty. If companies are citing AI as cover for ordinary business failures, they are misleading investors and the public. Market efficiency depends on accurate information. AI-washing, if widespread, distorts capital allocation and undermines the integrity of corporate communications.
Both perspectives converge on a demand for truth. Workers deserve to know why they lost their jobs. Investors deserve accurate explanations for corporate decisions. Policymakers deserve data that reflects reality rather than narratives constructed for strategic advantage.
The current situation satisfies none of these demands. Companies make claims. Workers lose jobs. The truth remains obscured by incentives that favor narrative over accuracy.
What the Data Actually Shows
The available evidence is mixed.
Challenger, Gray & Christmas, the employment consulting firm, documented that AI was cited in connection with approximately 50,000 layoffs in 2025. This represents a significant increase from previous years when AI was rarely mentioned explicitly.
However, total layoffs in the technology sector have remained elevated since the post-pandemic correction began in 2022. The pattern predates the current AI narrative. Companies hired aggressively during the pandemic, discovered they had overcapacity as growth normalized, and have been reducing headcount through multiple rounds of layoffs.
AI may be accelerating this correction. It may be providing cover for a correction that would have happened anyway. The aggregate data cannot distinguish between these possibilities.
What is notable is the change in rhetoric. Companies previously avoided explicitly blaming technology for job losses. Now they embrace it. This shift reflects changed incentives in capital markets, where AI capability is rewarded, rather than necessarily changed conditions in labor markets.
What This Means for Everyday People
For workers, the practical implications are the same regardless of whether AI displacement is real or exaggerated. The jobs are gone. The skills that secured previous employment may not secure future employment. Adaptation is required.
The strategic response for individuals is to develop capabilities that complement AI rather than compete with it. Tasks requiring judgment, creativity, interpersonal connection, and physical presence are less vulnerable than tasks that are routine, digital, and scalable. This advice would be valid even if AI were not a factor. It becomes more urgent if AI displacement proves real.
For citizens evaluating public policy, the appropriate stance is skepticism toward corporate narratives combined with preparation for genuine disruption. Companies have incentives to exaggerate AI’s impact. But the technology is real, and its capabilities are expanding. Prudent policy prepares for scenarios that may not materialize rather than assuming benign outcomes.
The 50,000 layoffs attributed to AI may be the beginning of a transformation that reshapes labor markets over the next decade. They may also be a convenient narrative that serves corporate interests while obscuring ordinary business dynamics.
The honest answer is that we do not know. What we know is that companies are now willing to say what they previously would not: that technology is taking jobs. Whether that statement is accurate remains to be determined.
For inquiries and analysis contact laterstack@proton.me
Frequently Asked Questions
Did AI really cause 50,000 layoffs?
Companies explicitly cited AI as a factor in over 50,000 layoffs in 2025, including major cuts at Amazon, Pinterest, and others. However, analysts debate whether AI actually replaced these workers or whether companies are “AI-washing,” using AI as a convenient explanation for cuts driven by overcapacity, economic conditions, or strategic failures. The truth likely involves both genuine displacement and opportunistic framing.
What is AI-washing in the context of layoffs?
AI-washing refers to companies attributing layoffs to AI capabilities and efficiency gains when the actual drivers may be unrelated to AI. According to Forrester research, many companies announcing AI-related layoffs “do not have mature, vetted AI applications ready to fill those roles.” Citing AI for layoffs is an investor-friendly message that can make ordinary business problems appear like technological progress.
What jobs are most at risk from AI?
Tasks that are routine, digital, and scalable are most vulnerable to AI displacement: customer service handling standard inquiries, formulaic content generation, data entry and transcription, and basic research and summarization. Jobs requiring judgment, creativity, interpersonal skills, and physical presence are less vulnerable. However, AI capabilities are expanding, and the boundary between vulnerable and protected work continues to shift.
Consider the audacity. A deepfake candidate applied for a security researcher position at Evoke, an AI security company whose entire business is threat modeling for artificial intelligence systems. The CEO, Rebholz, has spent years researching deepfakes. He has used them in presentations. He has built a career on understanding synthetic media.
And he almost hired the deepfake.
“Everything in me, everything I know about deepfakes was screaming at me: this is a deepfake,” Rebholz told The Register last week. “But there was something blocking me, the one percent chance that I’m wrong, this is actually a good candidate, and he’s going to think poorly of me if I confront him.”
This is the moment when an emerging threat becomes a systemic crisis. When a deepfake expert, interviewing for his own security company, experiences “inner turmoil” about whether to trust his expertise, the verification systems that underpin remote work have failed. The question is no longer whether deepfakes will compromise hiring processes. The question is what comes next.
The Anatomy of the Attack
The incident began ordinarily. Rebholz posted job openings on LinkedIn. Within hours, someone he did not know messaged him, recommending a candidate for the security researcher role. The referral itself was not suspicious. People refer candidates. Networks operate through introductions.
The first red flag was the candidate’s profile picture: not a photograph but something resembling an anime character. In the security community, this is not automatically disqualifying. Privacy concerns lead many professionals to avoid displaying their real faces online. Aliases and stylized avatars are common.
Rebholz gave the candidate the benefit of the doubt.
When the video interview began, the candidate sat in front of a virtual background. His face appeared “a bit blurry and plastic.” There was a greenscreen reflection visible in his glasses. At one point, dimples appeared on his face and then disappeared as he moved. The “softness of his face” came and went.
Rebholz noticed behavioral indicators as well. The candidate repeated interview questions back before answering them, a technique that buys processing time for systems generating responses. Many of his answers were nearly word-for-word quotes of things Rebholz himself had said or written publicly.
“It was almost an out-of-body experience where I felt like I was talking to myself,” Rebholz said.
Despite all of this, the deepfake expert experienced doubt. The candidate might be real. The visual artifacts might be compression issues. The familiar answers might be coincidence or good research. The social pressure not to accuse someone of being fake competed with the technical evidence that something was wrong.
After the interview, Rebholz sent video clips to a colleague at Moveris, which develops deepfake detection technology. The analysis confirmed what his expertise already told him: the candidate was synthetic.
The Implications for Remote Work
The remote work revolution that accelerated during the pandemic created enormous value. Talent could be hired from anywhere. Geographic constraints on labor markets loosened. Companies accessed skills that were previously unavailable in their local markets. Workers gained flexibility that many describe as transformational for their quality of life.
The same structural changes that enabled remote work also created vulnerabilities that are now being exploited.
Remote hiring relies on video interviews conducted over platforms that were not designed with identity verification in mind. Zoom, Teams, and Google Meet assume that the person on the video feed is who they claim to be. There is no cryptographic proof of identity. There is no biometric verification. There is trust, backed by the assumption that creating a convincing synthetic identity is difficult.
That assumption is no longer valid.
Experian’s fraud forecast for 2026 identified deepfake job candidates as a top threat. Nearly every major technology company, from Amazon to small startups, has encountered fake IT workers applying for positions. Some have been hired. Some have passed background checks. Some have gained access to internal systems before being detected.
The underlying threat is not merely fraudulent employment. It is infiltration. An attacker who passes an interview, receives credentials, and gains access to internal systems can exfiltrate data, compromise code repositories, establish persistent access, or conduct espionage. Remote work and global hiring widen talent pools, but they also weaken the signals of identity verification that previously protected organizations.
The Scale of the Problem
Industry data suggests the problem is larger than isolated incidents.
A 2024 survey found that 15% of high school students had encountered explicit deepfake imagery of their peers. The same technology that creates fake intimate images creates fake job candidates. The underlying models are general-purpose.
Challenger, Gray & Christmas, the employment consulting firm, has documented cases of synthetic candidates at scale. North Korean operatives have been identified using fake identities to obtain remote IT positions at American companies, earning salaries that fund the regime while potentially conducting espionage.
The FBI has documented cases where deepfake extortion, using synthetic intimate imagery created from social media photographs, has led to self-harm among victims, including minors. The same technology ecosystem enables job fraud, identity theft, and targeted harassment.
What makes the Evoke incident notable is not that a deepfake attempted to infiltrate a company. It is that the target was a company specifically focused on AI security threats, the interviewer was a deepfake expert, and the attack nearly succeeded anyway.
If experts cannot reliably detect deepfakes in real-time video interviews, what hope do ordinary hiring managers have?
The Emerging Response
The market is responding with verification technologies, but adoption lags threat development.
Companies like Moveris and others offer deepfake detection as a service. The technology analyzes video feeds for artifacts: inconsistent lighting, unnatural facial movements, audio-visual synchronization errors, and other indicators of synthetic generation. These tools can be integrated into hiring workflows.
Biometric verification services offer alternatives to video interviews for identity confirmation. A candidate can be required to verify their identity through a trusted third party before an interview begins. This does not prevent the interview itself from using deepfakes, but it establishes that the person claiming the identity is who they claim to be.
Some organizations are returning to in-person interviews for sensitive positions. The geographic flexibility of remote hiring is sacrificed for the security of physical presence. A deepfake cannot shake your hand.
Others are implementing multi-stage verification: initial video screening, followed by live coding exercises observed in real-time, followed by reference checks conducted through established professional networks rather than provided contacts. The friction increases. The security improves. The talent pool shrinks.
There is no costless solution. Remote work created value by reducing friction. The friction was also security. Restoring security means restoring friction.
The Policy Dimension
This is not merely a corporate security problem. It is a labor market problem with policy implications.
Employment verification systems, background check infrastructure, and credential validation processes were designed for a world where identity documents were difficult to forge and video communication did not exist. The entire apparatus assumes that physical presence or documented history establishes identity.
Deepfakes undermine these assumptions systematically. A synthetic candidate can provide fake credentials, appear convincingly in video interviews, and pass initial screening processes. The verification happens after hiring, when access has already been granted.
Policymakers face difficult tradeoffs. Mandating biometric verification raises privacy concerns. Requiring in-person verification restricts labor market flexibility. Imposing liability on employers for deepfake infiltration may be unfair when detection is genuinely difficult.
The honest answer is that no policy solution currently exists that preserves the benefits of remote hiring while eliminating the risks of synthetic candidates. Technology created this problem. Technology may eventually solve it. In the interim, organizations must make risk decisions with imperfect information and imperfect tools.
What This Means for Everyday People
For job seekers, the deepfake threat creates a new burden: proving that you are real. Legitimate candidates may face increased scrutiny, additional verification steps, and suspicion that would have been absent in earlier eras. The friction imposed by security measures falls on everyone, not just attackers.
For workers in remote positions, the question is whether their employers can distinguish them from synthetic impostors. If not, what prevents an attacker from impersonating a current employee, attending meetings in their place, or redirecting their communications?
For society broadly, the erosion of identity verification extends beyond employment. If you cannot trust that the person on a video call is who they claim to be, the implications ripple through every domain that relies on remote communication: telemedicine, legal proceedings, financial services, family relationships.
The Evoke incident is a warning. The CEO of an AI security company, an expert in exactly this threat, experienced doubt about whether to trust his own expertise. He almost hired a synthetic candidate because the social pressure not to accuse someone of fraud competed with the technical evidence that something was wrong.
This is the future of identity in the age of generative AI. The signals we relied upon to verify that people are who they claim to be are failing. What replaces them remains uncertain.
For inquiries and analysis contact laterstack@proton.me
Frequently Asked Questions
What happened with the deepfake job applicant at Evoke?
A synthetic identity using deepfake video technology applied for a security researcher position at Evoke, an AI security company. The CEO, who has years of experience researching deepfakes, conducted the video interview and noticed multiple red flags including visual artifacts and answers that quoted his own public statements. Despite his expertise, he experienced doubt about confronting the candidate and later confirmed through third-party analysis that the applicant was a deepfake.
How widespread is the deepfake job applicant problem?
Nearly every major technology company has encountered fake IT workers applying for positions. North Korean operatives have been identified using synthetic identities to obtain remote positions at American companies. Experian’s 2026 fraud forecast identifies deepfake job candidates as a top threat. The problem extends beyond tech: any organization using remote video interviews is potentially vulnerable.
How can companies protect against deepfake job candidates?
Emerging solutions include deepfake detection services that analyze video feeds for artifacts, biometric identity verification through trusted third parties, multi-stage verification processes combining video screening with live exercises and reference checks through established networks, and returning to in-person interviews for sensitive positions. No solution is costless; all involve tradeoffs between security and hiring flexibility.
There is a moment in the life of any dependency when the dependent party recognizes, suddenly and with clarity, the nature of the relationship. For European governments, that moment arrived when Microsoft canceled Karim Khan’s email.
Khan is the chief prosecutor of the International Criminal Court. In late 2025, the Trump administration imposed sanctions on him personally in response to ICC investigations. Microsoft, complying with U.S. sanctions law, terminated Khan’s access to his Microsoft email account. The prosecutor of an international tribunal established by treaty, investigating alleged war crimes, lost his email because an American company decided he should.
The implications rippled through foreign ministries, intelligence agencies, and government IT departments across Europe. If Microsoft could cancel the ICC prosecutor’s email at the command of the American executive branch, what else could American technology companies do? What communications could they intercept? What services could they terminate? What “kill switches” existed in the infrastructure that governments had come to depend upon?
Last week, France announced that it would ban Zoom and Microsoft Teams for public officials, replacing them with a domestically developed platform called Visio. The announcement is the most visible manifestation of a deeper shift: democratic nations are reconsidering their dependence on American technology platforms in ways that would have seemed paranoid five years ago and now seem prudent.
The Timeline of a Reckoning
The French announcement follows years of escalating concern.
In 2020, the European Court of Justice struck down Privacy Shield, the framework governing transatlantic data transfers, ruling that American surveillance law provided insufficient protection for European citizens’ data. The decision, known as Schrems II, created legal uncertainty for every European organization using American cloud services.
In 2022, the European Commission proposed the Data Act, establishing rules for data sharing and limiting international data transfers. The regulation reflected growing concern that European data processed by American companies was subject to American jurisdiction in ways that European governments could not control.
In 2025, the sanctions on the ICC prosecutor demonstrated that these concerns were not theoretical. American companies would comply with American government demands, regardless of the impact on foreign customers. The “kill switch” was real.
French Minister Delegate David Amiel stated the rationale plainly: maintaining classified exchanges and strategic information on external infrastructure poses an unacceptable national security risk. The question was not whether American platforms were convenient or cost-effective. The question was whether French government communications should be subject to the potential control of a foreign power.
The answer was no.
The Visio Alternative
Visio is not new. The platform has been in testing for approximately a year and already supports around 40,000 users within French government networks. It is hosted on Outscale’s sovereign cloud, a subsidiary of Dassault Systèmes, ensuring that all user data remains within French jurisdiction. The encryption keys are French. The servers are French. The legal framework governing access is French.
France announced that it will roll out Visio across all government departments by 2027. Leading adopters include the CNRS (French National Centre for Scientific Research), which will replace 34,000 Zoom licenses by March 2026, along with the French National Health Insurance Fund, the Directorate General of Public Finances, and the Ministry of Armed Forces.
The economic case is straightforward. France projects €1 million in annual savings for every 100,000 users migrating from licensed American solutions. For a government employing millions, the savings are substantial. But the primary motivation is not financial. It is strategic.
The Broader European Context
France is not acting alone.
The European Union has been constructing a regulatory framework to assert control over digital infrastructure for years. The Digital Markets Act imposes obligations on large platforms designated as “gatekeepers.” The Digital Services Act establishes content moderation requirements. The GDPR created the world’s most stringent data protection regime.
These regulations represent an attempt to exercise sovereignty over digital space without building alternative infrastructure. They regulate American platforms operating in Europe rather than replacing them. The French decision represents a different approach: building domestic alternatives and migrating away from American platforms entirely, at least for government functions.
Other European nations are watching. Germany has expressed similar concerns about digital sovereignty. The Netherlands has conducted assessments of risks associated with American cloud providers. The UK, despite its closer alignment with the United States, has begun examining the security implications of platform dependence.
Beyond Europe, the backlash is global. India has endorsed “Made in India” alternatives to American platforms. Australia has examined social media platform risks. International organizations have begun diversifying their technology providers to avoid the vulnerability that the ICC prosecutor experienced.
The American Perspective
From an American perspective, this development is strategically problematic but legally predictable.
American technology companies operate under American law. When the American government imposes sanctions, American companies must comply or face criminal penalties. Microsoft did not choose to cancel Khan’s email because it wanted to. It complied with legal requirements imposed by the executive branch.
The same logic applies to surveillance. American intelligence agencies have broad authorities to compel American companies to provide data, particularly data involving foreign nationals. The CLOUD Act of 2018 clarified that American companies must provide data to American law enforcement regardless of where the data is stored. European data on American platforms is American data for legal purposes.
This is not a bug in American law. It is a feature. American policymakers deliberately constructed a legal framework that extends American jurisdiction to data processed by American companies worldwide. The framework serves American intelligence and law enforcement interests.
The consequence is that foreign governments now recognize what American law always implied: using American platforms means accepting American jurisdiction. For routine commercial purposes, this may be acceptable. For sensitive government communications, it increasingly is not.
The Technology Nationalism Question
Critics will characterize the French decision as technology nationalism, a retreat from the globalized digital economy that has generated enormous value over the past decades.
The critique has merit but misses the strategic context. Technology is not neutral. The platforms through which governments communicate, store data, and conduct operations are infrastructure as critical as roads, ports, and power plants. No serious nation would allow a foreign power to control its physical infrastructure. The question is why digital infrastructure should be different.
The honest answer is that it should not be different, and nations are belatedly recognizing this reality.
The American government understood this long before European governments did. American defense and intelligence agencies do not use foreign platforms for sensitive communications. American sanctions law reflects an understanding that technology control is a form of power. American policy has consistently sought to maintain American dominance over global digital infrastructure.
European governments are now applying the same logic in reverse. If technology control is power, then dependence on American technology is subordination. Strategic autonomy requires technological autonomy, at least for critical government functions.
The Market Implications
For American technology companies, the French decision represents the beginning of a trend that will accelerate. Government contracts in democratic nations outside the United States will increasingly require sovereign infrastructure. The addressable market for American platforms in sensitive applications is shrinking.
This creates opportunities for European technology companies that can provide sovereign alternatives. It creates opportunities for American companies willing to establish genuinely independent subsidiaries operating under local legal frameworks. It creates complexity for multinational organizations that must now navigate diverging technology ecosystems.
The fragmentation of the global digital market along national and regional lines was always a possibility. American platforms benefited from a period of relatively uncontested global expansion. That period is ending. The kill switch that canceled Khan’s email demonstrated the risks of dependence, and nations are responding rationally by reducing that dependence.
What This Means for Everyday People
For ordinary citizens in France and elsewhere, the shift to sovereign platforms will be largely invisible. Visio will replace Zoom in government offices. Most citizens never attended those meetings anyway.
The deeper implications are systemic. The fragmentation of the global digital economy into sovereign blocs will reduce efficiency and increase costs. Interoperability between national systems will be imperfect. The seamless global communication that characterized the internet’s first decades will become more constrained.
Whether this is a price worth paying depends on values that reasonable people weigh differently. Privacy and sovereignty have costs. Convenience and efficiency have costs. The Khan incident forced a reckoning with tradeoffs that had been deferred, and nations are making choices about which costs they are willing to bear.
France has chosen sovereignty over convenience. Other nations will make their own calculations. The era of unquestioned American dominance over global digital infrastructure is ending. What replaces it remains to be determined.
For inquiries and analysis contact laterstack@proton.me
Frequently Asked Questions
Why did France ban Zoom and Microsoft Teams for government use?
France announced it will phase out American video conferencing platforms for public officials due to security concerns about data sovereignty and potential foreign government access to communications. The decision follows the 2025 incident where Microsoft canceled the ICC prosecutor’s email account in compliance with U.S. sanctions, demonstrating that American companies will comply with American government demands regardless of impact on foreign customers.
What is Visio and how does it differ from American platforms?
Visio is a French-developed video conferencing platform hosted on Outscale’s sovereign cloud, a subsidiary of Dassault Systèmes. Unlike American platforms, Visio keeps all data, encryption keys, and traffic within French jurisdiction and under French legal frameworks. The platform has been in testing for a year and already serves 40,000 government users.
Will other countries follow France’s example?
Multiple countries and international organizations are reconsidering their dependence on American technology platforms. Germany, the Netherlands, India, and Australia have all examined digital sovereignty concerns. The International Criminal Court itself dropped Microsoft as a service provider following the sanctions incident. The trend toward sovereign digital infrastructure is accelerating across democratic nations.
Europe has excelled at quantum research. Its universities and national laboratories have contributed foundational discoveries to the field. European physicists have won Nobel Prizes for quantum work. The theoretical groundwork for quantum computing was laid in part by European scientists.
What Europe has not excelled at is turning research into industry. The pattern repeats across technology sectors. European institutions invent, and American or Asian companies commercialize. The internet was created with significant European contributions. The dominant internet companies are American. Machine learning draws on decades of European academic research. The leading AI companies are American and increasingly Chinese.
The European Union is determined to break this pattern with quantum computing.
This week, the EU launched SUPREME, a consortium with a €50 million (approximately $59 million USD) budget to industrialize superconducting quantum technology. The initiative brings together leading European research institutions, quantum startups, and industrial partners with the explicit goal of establishing manufacturing capability for quantum processors on European soil.
This is not research funding. This is industrialization funding. The EU has concluded that quantum research excellence without manufacturing capability is strategically insufficient, and it is committing public resources to ensure that European research translates into European industry.
The Superconducting Focus
SUPREME focuses specifically on superconducting qubits, the same qubit technology employed by IBM, Google, and several leading quantum computing companies. Superconducting qubits encode quantum information in electrical circuits cooled to temperatures near absolute zero. They offer fast gate operations and have demonstrated some of the largest quantum processors to date.
The technology has disadvantages. It requires extreme cooling infrastructure, making the systems expensive and power-intensive. The qubits must be individually calibrated, creating manufacturing challenges at scale. But superconducting qubits represent one of the most mature approaches to quantum computing, with a clearer path to near-term applications than some competing technologies.
By focusing SUPREME on superconducting technology, the EU is making a strategic choice. It is not pursuing every possible qubit modality. It is concentrating resources on an approach where European research has strong foundations and where industrial scaling is the primary remaining challenge.
This concentration reflects a broader shift in European technology policy: from spreading resources thinly across many possibilities to focusing intensively on a smaller number of strategic priorities.
The Consortium Structure
SUPREME brings together partners across the quantum value chain. Research institutions provide scientific expertise and access to advanced laboratory facilities. Quantum startups contribute entrepreneurial energy and commercialization focus. Industrial partners offer manufacturing know-how, supply chain connections, and eventual customer demand.
The structure is designed to address Europe’s historical weakness: the gap between research and commercialization. By integrating researchers, startups, and industrial partners into a single consortium, SUPREME aims to ensure that advances in the laboratory translate rapidly into improvements in manufacturing and ultimately into products.
The €50 million budget, while substantial for a single initiative, is modest compared to total European quantum investments. The EU Quantum Flagship, launched in 2018, committed €1 billion over ten years. National programs in Germany, France, and the Netherlands have added additional billions. SUPREME represents a focused intervention within this broader funding landscape, targeting specifically the industrialization bottleneck.
The Strategic Autonomy Imperative
The phrase “strategic autonomy” has become central to European technology policy. It captures the recognition that dependence on foreign suppliers for critical technologies creates vulnerability, both to supply disruption and to geopolitical leverage.
Europe’s experience with semiconductor supply chains during 2020 and 2021 crystallized this concern. European automakers, among the world’s largest, discovered that they could not build cars because they could not obtain chips manufactured primarily in Asia. The economic cost was billions of euros in lost production. The strategic lesson was that dependence on foreign manufacturing in critical technologies is untenable.
Quantum computing represents an opportunity to avoid creating new dependencies. The industry is nascent. No region dominates manufacturing. The choices made now will determine whether Europe is a participant in the quantum industry or merely a customer.
SUPREME is part of a broader European effort to establish quantum manufacturing capability before dependencies form. The initiative complements national programs, private investments, and European research funding to create an ecosystem capable of producing quantum processors competitively with American and Asian suppliers.
The Competitive Landscape
Europe’s quantum ambitions face formidable competition.
The United States dominates the quantum startup landscape. Google, IBM, and Amazon are investing billions in quantum hardware and cloud services. American venture capital flows more freely into quantum companies than European capital. The National Quantum Initiative provides sustained federal support.
China is investing heavily in quantum technologies, though with less transparency about specific programs and funding levels. Chinese research groups have demonstrated world-leading capabilities in quantum communication and certain quantum computing modalities.
European companies including IQM (Finland), Alice & Bob (France), OQC (UK), and Kiutra (Germany) have emerged as credible contenders. But they face capital constraints, market access challenges, and competition from better-funded American rivals.
SUPREME and broader European quantum funding aim to tilt the competitive landscape. By providing capital that might not be available from private markets, by building manufacturing infrastructure that individual companies could not fund alone, and by creating connections between research and industry, public investment can accelerate the development of a European quantum industry.
Whether this strategy succeeds depends on execution. Government funding can enable or distort. It can fill gaps in private capital formation or it can create dependence on continued public support. It can accelerate commercialization or it can insulate companies from market discipline. The outcomes depend on how programs like SUPREME are implemented, how success is measured, and how funding evolves as the industry matures.
The Subsidy Question
Critics of industrial policy will note that SUPREME, like other government quantum investments, represents a subsidy to technology development that might occur naturally through private markets. The critique has merit but requires contextualization.
Quantum computing is a long-horizon technology. Meaningful commercial applications remain years away. Private capital, which typically seeks returns within five to seven years, may underinvest in technologies with longer development timelines. Government capital, which can take a generational view, fills this temporal gap.
Moreover, quantum computing has characteristics of a strategic technology where early capability creates lasting advantage. The physics of learning curves means that early entrants can reduce costs and improve quality faster than later entrants. Nations that develop quantum manufacturing capability now may maintain advantages for decades.
From this perspective, government investment in quantum industrialization is not a distortion of markets but a correction for market failures: the short time horizons of private capital and the public goods characteristics of strategic technology capability.
What This Means for Everyday People
For ordinary Europeans, the SUPREME consortium is distant from daily concerns. Quantum computers will not affect most people’s lives for years or decades, if ever.
The relevance is indirect but real. Technology industries create employment, economic growth, and tax revenue. They attract talent and catalyze innovation in adjacent sectors. Nations that participate in building strategic technologies benefit economically and strategically from that participation.
The alternative is dependence. Europeans already depend on American companies for cloud computing, social media, search, and mobile operating systems. They depend on Asian companies for semiconductors, batteries, and electronics manufacturing. Each dependency represents a constraint on European autonomy and a transfer of economic value outside the continent.
SUPREME represents one small effort to ensure that quantum computing does not become another such dependency. The €50 million investment will not by itself create a European quantum industry. But combined with national programs, private investment, and the deep research capabilities of European institutions, it contributes to an ecosystem that could make Europe a participant in the quantum future rather than merely a consumer of it.
For inquiries and analysis contact laterstack@proton.me
Frequently Asked Questions
What is the SUPREME consortium?
SUPREME is a European Union initiative with a €50 million budget to industrialize superconducting quantum technology. The consortium brings together research institutions, quantum startups, and industrial partners to bridge the gap between laboratory research and commercial manufacturing of quantum processors.
Why is Europe investing in quantum manufacturing?
Europe has historically excelled at technology research while struggling to commercialize innovations into industrial capability. The EU is investing in quantum manufacturing specifically to avoid creating dependencies on American or Asian suppliers as the quantum computing industry develops, pursuing what European policymakers call “strategic autonomy.”
How does SUPREME fit into broader European quantum funding?
SUPREME is one component of a larger European quantum investment landscape that includes the €1 billion EU Quantum Flagship program, national programs in Germany, France, Netherlands, and other member states, and private venture funding. SUPREME focuses specifically on the industrialization bottleneck rather than fundamental research.