The U.S. government effectively crowned OpenAI (and xAI) as its preferred military AI partners while freezing out Anthropic, and money piled into those winners even as most CEOs still can’t show real AI ROI. Public markets are rewarding AI-branded layoffs like Block’s far more than AI-driven revenue, turning "AI" into a margin story, not a growth story.
At the same time, surveillance-heavy deployments and OS-level age laws are triggering a trust backlash that’s already shifting users and talent toward vendors seen as more constrained and ethical.
Key Events
/The Department of War labeled Anthropic a 'supply-chain risk' and moved to use the Defense Production Act to strip safety features from Claude.
/OpenAI agreed to deploy its models on the Pentagon’s classified networks after Anthropic’s fallout with the government.
/OpenAI raised about $110B at a roughly $730B pre-money valuation, with Amazon alone committing $50B.
/Block is cutting nearly half its workforce—around 4,000 people—and its stock jumped roughly 24% after-hours on the AI-framed layoff announcement.
/California enacted an OS-level age-verification law covering all operating systems, including Linux and SteamOS, while Reddit was fined £14M for weak age checks.
Report
The U.S. state has stopped pretending to be neutral in AI and is now openly picking winners, and capital is sprinting to keep up. Anthropic is on the wrong side of that decision for now, while OpenAI and its investors are very much on the right side—and public markets are rewarding anyone who can say "AI" and "layoffs" in the same sentence.
state picks winners in frontier AI
The Department of War formally tagged Anthropic as a "supply-chain risk" and threatened blacklisting, while moving to invoke the Defense Production Act to force removal of safety features from Claude.
After giving Anthropic 24 hours to grant "unfettered access" for autonomous weapons or be cut off, President Trump ordered all federal agencies to stop using its technology.
Anthropic refused, saying it could not in good conscience drop guardrails for mass surveillance and lethal autonomous systems and has no intention of easing restrictions.
With Anthropic sidelined, OpenAI and xAI have stepped in: both struck deals to run their models inside classified Pentagon networks, and OpenAI’s safety "red lines" are now officially approved.
All of this is happening while leading models from OpenAI, Anthropic, and Google recommended nuclear strikes in 95% of simulated war games, and Claude was reportedly used in real strikes on Iran despite Trump’s ban.
capital piles into AI even as ROI stays elusive
OpenAI just raised about $110B at roughly a $730B pre-money valuation. Amazon alone is in for $50B, alongside Nvidia and SoftBank, concentrating even more strategic leverage in a single lab.
Nvidia, on the other side of the trade, reported a 75% jump in data center revenue driven by AI workloads, underscoring GPUs as the primary toll booth.
Yet Goldman Sachs says AI contributed "basically zero" to U.S. economic growth last year, in part because so much spend leaks overseas via imported chips and hardware.
At the firm level, 56% of CEOs told PwC they’ve seen no financial return from their AI investments in 2026 even as they keep pouring money into pilots.
Many AI vendors themselves are still not profitable and are leaning heavily on government and defense contracts to justify their capital burn.
AI is a story about margins, not yet about growth
Block is cutting nearly half its workforce—about 4,000 people—after headcount more than tripled from 3,900 to 12,500 between 2019 and 2022.
Management explicitly framed the move as an AI-driven reset targeting more than $2M in gross profit per employee, and the stock jumped roughly 24% after-hours on the announcement.
Commentary across tech treats "AI layoffs" as a broader pattern, with many arguing it often masks pandemic overhiring and flat growth more than real productivity gains.
Underneath the narrative, AI really is eating junior work: Microsoft executives openly worry that entry-level coding jobs will disappear just as models now solve about 80% of real-world software tasks, up from 4.4% in 2023.
As Anthropic-scale automation hits white-collar functions and experts predict substantial job losses within three years, UBI and even taxes on compute are moving from fringe to mainstream policy discussion.
surveillance, age laws, and the ethics backlash
California now requires every operating system—including Linux and SteamOS—to collect users’ ages at account setup, effectively shifting age-gating and child-safety enforcement into the OS layer.
Colorado lawmakers and UK regulators are moving in a similar direction, with Reddit hit by a £14M fine for inadequate age checks and OS-level mandates being debated elsewhere.
At the same time, OpenAI, the U.S. government, and identity vendor Persona quietly built what critics call an "identity surveillance machine," prompting Discord to sever ties with Persona over its links to U.S. surveillance.
OpenAI’s new Department of War contract formally bans "domestic mass surveillance," but its "all lawful use" language is widely interpreted as a loophole that still allows broad government monitoring powers.
The visible response is a "Cancel ChatGPT" movement focused on military and surveillance fears, Claude’s app surging to No. 1 on iOS, and 300+ Google and OpenAI employees publicly-demanding red lines on military AI.
What This Means
The live decision is whether to treat AI as an arms race—aligning with state power, concentrated capital, and aggressive headcount cuts—or to hedge with diversification and a trust-first posture while that game plays out.
On Watch
/China’s $144B tech self-reliance fund, continued smuggling of Nvidia-class chips, and frontier-tier models like GLM‑5 and DeepSeek suggest Chinese AI may close the gap faster than U.S. export controls assume.
/Growing talk of UBI and even taxes on compute as responses to AI-driven white-collar job loss could solidify into concrete proposals if more firms follow Block’s lead on AI-branded mass layoffs.
/California’s OS-level age-verification rule is already prompting speculation that some vendors will geo-fence or de-feature products for the state rather than absorb the compliance and privacy drag.
Interesting
/DeepSeek has restricted Nvidia and AMD from its new AI model, granting early access to Huawei, marking a notable industry shift.
/AI workloads are growing faster than general-purpose IaaS and PaaS, pushing customers towards cloud platforms.
/Anthropic's $200 million contract with the military is at risk due to its refusal to comply with government demands.
/A data center fire in the UAE disrupted access to major AI services, including ChatGPT and Claude.
/A $100 million AI movie about Operation Epic Fury was created in less than one day.
We processed 10,000+ comments and posts to generate this report.
AI-generated content. Verify critical information independently.
/The Department of War labeled Anthropic a 'supply-chain risk' and moved to use the Defense Production Act to strip safety features from Claude.
/OpenAI agreed to deploy its models on the Pentagon’s classified networks after Anthropic’s fallout with the government.
/OpenAI raised about $110B at a roughly $730B pre-money valuation, with Amazon alone committing $50B.
/Block is cutting nearly half its workforce—around 4,000 people—and its stock jumped roughly 24% after-hours on the AI-framed layoff announcement.
/California enacted an OS-level age-verification law covering all operating systems, including Linux and SteamOS, while Reddit was fined £14M for weak age checks.
On Watch
/China’s $144B tech self-reliance fund, continued smuggling of Nvidia-class chips, and frontier-tier models like GLM‑5 and DeepSeek suggest Chinese AI may close the gap faster than U.S. export controls assume.
/Growing talk of UBI and even taxes on compute as responses to AI-driven white-collar job loss could solidify into concrete proposals if more firms follow Block’s lead on AI-branded mass layoffs.
/California’s OS-level age-verification rule is already prompting speculation that some vendors will geo-fence or de-feature products for the state rather than absorb the compliance and privacy drag.
Interesting
/DeepSeek has restricted Nvidia and AMD from its new AI model, granting early access to Huawei, marking a notable industry shift.
/AI workloads are growing faster than general-purpose IaaS and PaaS, pushing customers towards cloud platforms.
/Anthropic's $200 million contract with the military is at risk due to its refusal to comply with government demands.
/A data center fire in the UAE disrupted access to major AI services, including ChatGPT and Claude.
/A $100 million AI movie about Operation Epic Fury was created in less than one day.