TL;DR
Washington just picked sides in AI: Anthropic is being punished for refusing surveillance and weapons work while OpenAI and xAI move into classified systems backed by a $110B mega‑round and hyperscaler capex. Markets are rewarding AI‑branded layoffs and GPU vendors, even though Goldman says AI added basically nothing to GDP and most CEOs still see no financial return.
The real trade is how much you’re willing to tie your AI roadmap to a defense‑aligned, NVIDIA‑centric stack before the productivity story actually shows up in the numbers.
Key Events
Report
Defense and capital just picked their early winners in frontier AI: OpenAI and xAI are moving into classified systems while Anthropic is being pushed out of U.S. government entirely.
At the same time, $100B‑class infra bets and AI‑justified layoffs are landing before the technology has delivered any measurable macro productivity.
The Pentagon labeled Anthropic a supply‑chain risk, issued a 24‑hour ultimatum to drop guardrails on autonomous weapons, and Trump ordered all federal agencies to stop using Claude.
Anthropic refused a major Pentagon contract on ethical grounds, saying it could not in good conscience support mass surveillance or lethal autonomous weapons.
OpenAI stepped into that vacuum, agreeing to deploy its models on the Department of War’s classified networks under an 'all lawful purposes' framework while publicly claiming to share Anthropic’s red lines.
xAI likewise signed a deal to run Grok in classified systems, lining up alongside OpenAI as a defense‑aligned lab. In parallel, more than 300 Google and OpenAI employees signed a public letter backing Anthropic’s stance, and consumer backlash has pushed Claude to the top of Apple’s free apps chart while 'Cancel ChatGPT' trends.
OpenAI closed a $110B funding round, the largest single raise in tech history. The round valued the company around $730B pre‑money, with Amazon pledging tens of billions alongside NVIDIA and SoftBank and an eight‑year $100B AWS spend commitment.
Hyperscaler capex is already growing about 70% per year since GPT‑4, with AI‑driven spend projected to reach $770B in 2026 across Amazon, Microsoft, Google, Meta and others.
OpenAI itself projects burning $157B of cash through 2028 even as its COO says AI has barely penetrated enterprise workflows so far. Meanwhile Goldman Sachs estimates AI contributed 'basically zero' to U.S. GDP last year and 56% of CEOs in a PwC survey report no financial return from AI investments by 2026.
Anthropic’s launch of a COBOL‑focused Claude Code tool knocked around 10–13% off IBM’s share price in days, as investors questioned a mainframe and COBOL services franchise that still underpins roughly 95% of U.S. ATM transactions.
In cybersecurity, the debut of Claude Code Security coincided with about a $15B sell‑off across listed cyber names as markets reconsidered whether human‑heavy detection and services models can withstand AI‑native tools.
Anthropic’s own automation is already biting into revenues in legal tech, cybersecurity, IT consulting and finance, feeding claims that February 2026 marked 'the death of white‑collar work.' Jack Dorsey’s Block laid off nearly half its workforce in an AI pivot and the stock jumped 24% after‑hours, while a FAANG company is reportedly planning a 25% workforce reduction linked to AI investments.
These cuts are hitting not only programmers but engineers, marketing and HR, even as many CEOs still see zero direct AI financial return and Goldman’s macro data show no visible productivity bump.
Meta signed a multi‑year deal with AMD to deploy Instinct GPUs across roughly 6GW of data‑center capacity in an arrangement valued up to $100B including a 10% stock component.
Nvidia reported $68.13B in quarterly revenue, up 73% year‑on‑year, with data‑center sales up 75%, and is already sampling its Vera Rubin AI GPU with 288GB of HBM4 per card.
At the same time, a $100B Micron memory megafab in New York has been delayed by a lawsuit from six local citizens, and data centers are hoarding SSDs and RAM to the point that a 390TB game archive is shutting down over storage costs.
Intel server‑class CPUs are in short supply with rising prices and long lead times, pushing some users toward AMD for reliability in home and lab servers.
China’s DeepSeek is training and optimizing frontier models on Chinese chips, giving early access to Huawei while excluding Nvidia and AMD, on top of an energy system that already generates 40% more electricity than the US and EU combined and is adding record solar capacity.
What This Means
State power and hyperscaler capital are fusing into a single AI stack around OpenAI, NVIDIA and defense work, while the same technology is eroding white‑collar moats and stressing compute supply before it has proved out real productivity gains. The live decision is whether to anchor your AI roadmap on that state‑aligned stack or keep optionality with neutral vendors, knowing that choice now defines your exposure to politics, talent and supply chains as much as to model quality.
On Watch
Interesting
We processed 10,000+ comments and posts to generate this report.
AI-generated content. Verify critical information independently.
Sources
Key Events
On Watch
Interesting