A small circle of US labs and infra vendors is locking up GPUs, talent, and sensitive data just as regulators and the public start pushing back hard on surveillance, government cloud dominance, and AI-driven layoffs.
The real action this quarter is around chokepoints—compute, health and driver records, and specialist staff—becoming both the core profit centers and the core political and legal flashpoints.
Key Events
/Anthropic locked up more than 220,000 NVIDIA GPUs by signing a multi-year lease for SpaceXAI's Colossus 1 supercomputer.
/Meta is cutting 8,000 jobs while committing about $145B to AI, including a $10B Louisiana data center with major tax breaks.
/EU regulators are moving to block Microsoft, Amazon, and Google from handling sensitive government data.
/Palantir won 'unlimited access' to identifiable NHS patient data, triggering political backlash and regional opt-outs.
/The FCC approved SpaceX's purchase of 65 MHz of spectrum from EchoStar to power Starlink Mobile.
Report
Anthropic, NVIDIA, and a handful of labs are cornering frontier compute at the same time regulators and publics are putting hard constraints around data and infrastructure.
The result is a tightening triangle around GPUs, sensitive datasets, and permits for AI build-outs, exactly where large tech and industrial capex is now concentrated.
the new GPU cartel
Anthropic has leased the entire Colossus 1 facility from SpaceXAI, locking in access to more than 220,000 NVIDIA GPUs and 300MW of power. xAI itself has deployed a similar-scale stack at Colossus 1—about 150,000 H100s, 50,000 H200s, and 20,000 GB200s—concentrating a vast amount of frontier compute in one site.
NVIDIA has committed roughly $40B to equity AI deals and now has a valuation larger than the GDP of every country except the US and China, giving it outsized leverage over the GPU supply chain.
Anthropic reported 80-fold growth in Q1 and has had to chase additional compute, while OpenAI is reportedly processing 57 billion tokens per day and has raised a new $4B corporate unit to fund expansion.
Frontier labs are also securing multi-year cloud and infrastructure deals, from Anthropic’s $1.8B AI cloud agreement with Akamai to its $200M partnership with the Gates Foundation.
hyperscalers, sovereignty, and the state
European authorities are moving to bar Microsoft, Amazon, and Google from handling sensitive government health, financial, and legal data, directly restricting US hyperscalers’ access to core public-sector workloads.
At the same time, the EU Commission is in talks with OpenAI and Anthropic about their models and is backing Italy’s push to force Meta to pay news publishers.
In North America, Apple is warning that Canada’s Bill C-22 could effectively mandate encryption backdoors even as it rolls out end-to-end encrypted RCS messaging between iPhone and Android devices.
The FCC has approved SpaceX’s purchase of about 65 MHz of spectrum from EchoStar for Starlink Mobile and is simultaneously proposing stricter Know Your Customer rules that would require providers to collect more personal data.
The FCC also reversed its earlier ban on software updates for certain foreign-made drones and routers, extending update permissions through 2029 as a time-bounded compromise on hardware security.
surveillance-grade data platforms and the backlash
Palantir staff have been granted 'unlimited access' to identifiable NHS patient data under a new platform deal, while ICE agents in the US reportedly carry a list of 20 million people on their devices via Palantir software.
The same company is facing political pushback, with Greater Manchester rejecting an NHS data platform centered on Palantir and German intelligence services declining to adopt its tools after evaluation.
Separately, GM agreed to a $12.75M settlement under the California Consumer Privacy Act after selling driver data without consent through its OnStar program, which has generated substantial revenue from data sales.
Texas is suing Netflix for allegedly spying on children and misleading users about data collection in what the state calls 'bait and switch' advertising practices.
Content and training-data practices are also under legal fire as a judge refused to dismiss a copyright suit over 197,000 pirated books used with NVIDIA’s NeMo framework, and a family sued OpenAI claiming ChatGPT’s advice led to an overdose.
white-collar jobs, cash-outs, and the AI pivot
Meta is laying off about 8,000 people as part of a roughly $145B AI initiative, while Cloudflare cut 1,100 staff—around 20% of its workforce—despite record revenue and a 600% jump in AI usage over three months.
Oracle has reportedly laid off around 30,000 employees tied to cost-cutting and changes in restricted stock units, and LinkedIn has cut 5% of its staff, or about 875 roles.
GM is eliminating hundreds of salaried IT jobs while hiring workers with stronger AI skills, including foreign employees, prompting internal concerns about operational and security gaps.
Microsoft says 30% of its new code is now generated by AI and it already employs over 100 AI agents to hunt for Windows vulnerabilities.
Over 600 OpenAI employees have sold $6.6B of stock in a secondary sale averaging $11M per person, even as the company raises a new $4B corporate-focused unit for AI initiatives.
What This Means
Capital, compute, and data are consolidating into a narrow US-centric stack just as governments, courts, and local communities start to treat that stack as something to tax, constrain, and litigate. The near-term story is not just AI capability growth but a visible build-up of chokepoints—GPUs, health records, driver data, specialized talent—sitting under political and legal spotlights at the same time.
On Watch
/An 18-day strike threat at a key Samsung chip fab, with workers rejecting a $340,000 bonus and management considering scaling down production, could tighten semiconductor and SSD supply after prices in Japan have already jumped 300%.
/Google and SpaceX are exploring orbital data centers tied to Starlink and Gemini, a concept that could create an entirely new compute and distribution layer if it clears the substantial engineering and economic skepticism.
/The FCC’s push for stronger Know Your Customer rules, potentially forcing providers to collect extensive personal data, could raise the compliance and onboarding burden across connectivity and communications businesses.
Interesting
/AI referrals are outperforming traditional Google traffic in engagement, indicating a shift in user interaction with digital content.
/Anthropic is reportedly destroying rare books as a legal safeguard while acquiring millions for scanning.
/Google DeepMind workers voted to unionize due to military AI contracts, reflecting growing concerns over ethical implications of AI development.
/SoftBank has reduced its planned margin loan backed by OpenAI shares from $10 billion to approximately $6 billion due to creditor concerns.
/Amazon's internal culture pressures employees to adopt AI technologies, leading to inflated usage metrics that may compromise data integrity.
We processed 10,000+ comments and posts to generate this report.
AI-generated content. Verify critical information independently.
/Anthropic locked up more than 220,000 NVIDIA GPUs by signing a multi-year lease for SpaceXAI's Colossus 1 supercomputer.
/Meta is cutting 8,000 jobs while committing about $145B to AI, including a $10B Louisiana data center with major tax breaks.
/EU regulators are moving to block Microsoft, Amazon, and Google from handling sensitive government data.
/Palantir won 'unlimited access' to identifiable NHS patient data, triggering political backlash and regional opt-outs.
/The FCC approved SpaceX's purchase of 65 MHz of spectrum from EchoStar to power Starlink Mobile.
On Watch
/An 18-day strike threat at a key Samsung chip fab, with workers rejecting a $340,000 bonus and management considering scaling down production, could tighten semiconductor and SSD supply after prices in Japan have already jumped 300%.
/Google and SpaceX are exploring orbital data centers tied to Starlink and Gemini, a concept that could create an entirely new compute and distribution layer if it clears the substantial engineering and economic skepticism.
/The FCC’s push for stronger Know Your Customer rules, potentially forcing providers to collect extensive personal data, could raise the compliance and onboarding burden across connectivity and communications businesses.
Interesting
/AI referrals are outperforming traditional Google traffic in engagement, indicating a shift in user interaction with digital content.
/Anthropic is reportedly destroying rare books as a legal safeguard while acquiring millions for scanning.
/Google DeepMind workers voted to unionize due to military AI contracts, reflecting growing concerns over ethical implications of AI development.
/SoftBank has reduced its planned margin loan backed by OpenAI shares from $10 billion to approximately $6 billion due to creditor concerns.
/Amazon's internal culture pressures employees to adopt AI technologies, leading to inflated usage metrics that may compromise data integrity.