Friday's headlines turned on one question. Who inside this industry believes their own warnings? Jacob Coxon answered by resigning from Anthropic and naming both his former employers: "I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives"1. Three colleagues answered by staying at their desks and agreeing with him in public. Capital moved at the speed of conscience. Qualcomm committed to sell Amazon up to $60 billion in AI data-center chips through 2036, and Mistral closed Europe's largest technology raise on record at a €21 billion valuation23.
Read together, the day argues one thing. The people closest to the technology see the stakes most clearly, and keep building anyway.
A Resignation Draws Agreement
Jacob Coxon spent three years on pretraining research at OpenAI and Anthropic before he resigned from the second of them on Sept. 91. Samuel Marks, who leads scalable oversight at the company, answered in public: "AI developers believe their technology could cause human extinction (or similarly bad outcomes)". He added that "Many AI developer staff desperately want to slow down to figure out how to build AI more safely". Anna Wang, a member of the technical staff, was blunter about the gap: "There is not yet a viable scientific plan to solve risks from recursively self-improving AI". Ethan Perez, who leads the alignment team, kept it to one line: "100% agree with him that AI poses serious risks to society, and I'm glad he's speaking out!". Three researchers stayed on the payroll and agreed, in public and by name, with the man who walked out. Staying inside to steer the outcome is its own bet on Anthropic's chances.
Qualcomm Wagers on the Data Center
Cristiano Amon's Qualcomm signed a deal on Sept. 82. It lets Amazon buy up to $60 billion in chips, systems, and optical gear through 2036. Amazon also received warrants for 25 million Qualcomm shares at $161.26 apiece. That $4 billion stake ties the retailer's upside to the chipmaker's bet. Qualcomm shares climbed 3 percent. The company targets $15 billion in annual data-center revenue by fiscal 2029. It wants 5 percent of a market it sizes at $1 trillion, even as its phone business braces for a steep Apple-related drop this quarter. A second growth engine arrives exactly when the first one wobbles.
Samsung Leads Europe's Biggest Bet
A €3 billion round led by Samsung Electronics landed in Mistral on Sept. 83. It values the Paris lab at €21 billion, nearly double its mark a year earlier, with EQT's Scaleup Europe Fund and existing backer PSG Equity as co-leads. Nvidia, Salesforce Ventures, Advent, BlackRock, and the Grand Duchy of Luxembourg returned as investors. Mistral called it the largest equity raise ever completed by a European technology company. Proceeds go toward a full gigawatt of European compute by 2030. Chief executive Arthur Mensch frames the mission as building "a third way in AI". That pitch aims at governments wary of routing sovereign data through Silicon Valley or Beijing.
Lehane Asks Congress to Legislate
Chris Lehane, OpenAI's chief global affairs officer, published an appeal on Sept. 94. He asked Congress to act before it adjourns in December. His argument ran plainly: "The prospect of AI-accelerated AI development demands more than voluntary commitments. The United States needs mandatory, capability-based national regulation that can evolve as the technology does". OpenAI's ask covers testing standards, independent assessments, incident reporting, and mandatory alignment-evaluation gates before deployment, plus support for California bills SB 813 and AB 1405, both already signed into law. The company drew one line explicitly: "Fully autonomous recursive self-improvement...is not happening today and we should not pursue it unless and until it can be done safely". Petitioning lawmakers for oversight that could slow its own roadmap is a reversal. This is a company built on shipping fast.
Positron Stacks Nine Figures on Memory
Mitesh Agrawal's Positron AI closed $875 million on Sept. 105. The money split two ways. A $375 million Series C came first, then a Series C-1 anchored by Jim Clark and NEA. That pushed the inference-chip startup to a $5 billion valuation. Dylan Patel's SemiAnalysis Capital joined. Patel praised the memory-first approach for hitting what he called the real constraint in AI hardware. Proceeds fund the tapeout of Positron's Asimov chip on TSMC's N3P process by year's end, plus a Titan system that pairs up to eight Asimov chips to serve models topping 16 trillion parameters. Positron bet nine figures on cheap memory, ahead of exotic packaging. That makes it the loudest challenger yet to Nvidia's grip on inference costs.
Anthropic Catalogs Its Own Abuse
A threat-intelligence report landed on Sept. 10, the most detailed Anthropic has published6. It documents cases spanning cyber operations, influence campaigns, and biological misuse between December 2025 and August 2026. One case tied to Chinese-speaking operators produced more than a dozen zero-day findings in a single month. A 13-agent collection fleet did the work. Another campaign took more than 300,000 national identity records and 500,000 company registry entries from a North African target before Anthropic's team found it and shut the operation down. Publishing the specifics, down to terabyte counts and agent designs, doubles as a pitch. The governments Lehane courts that same week are the audience.
Microsoft Plans a Tripled Footprint
Bloomberg reported the plan on Sept. 107. Microsoft would more than triple its data-center footprint, from roughly 12 gigawatts today to over 38 gigawatts by 2032. People familiar with the matter gave the figures. Roughly a third of that would run AI-specific silicon, up from about 2 gigawatts today. The buildout follows real capacity limits. Microsoft turned away cloud and AI customers and capped subscriptions in recent months. Scaling at this pace signals that Satya Nadella's operation treats surging demand as a permanent condition.
Microsoft plans to more than triple its data-center footprint, from roughly 12 gigawatts to over 38 by 2032
Microsoft's data-center capacity today and planned for 2032, in gigawatts, as Bloomberg reported it
Each row is one measure of capacity: the blue dot is today, the magenta dot is the 2032 plan, and the rod between them is the planned growth.
Source: Bloomberg [7]. Chart by Ryan Elliott Dennis.
The numbers behind this chart
| Item | Today | Planned by 2032 | Note |
|---|---|---|---|
| All data centers | 12 GW | 38 GW | Roughly 12 gigawatts today, over 38 by 2032 |
| AI-specific silicon | 2 GW | 12.7 GW | Roughly a third of the 38 gigawatts, up from about 2 today |
Meta Ships an Agent That Can Pay
Muse, Meta's personal AI agent, arrived Sept. 8, built to book travel, fill forms, and run multistep tasks across a user's email and calendar8. Supervision stays light. Mark Zuckerberg described the free tier as usable up to 100 million tokens a week. Paid Power and Maximum tiers cost $20 and $100 a month. Stripe's Link technology backs every purchase Muse makes, issuing one-time-use cards and extending Link's buyer protections to an AI agent for the first time. Meta handed an algorithm a payment method built for repeat purchases. That puts it ahead of Google and Microsoft in the race to build agents that spend on command.
NASA Opens the Moon's Data
Scientists at NASA and IBM open-sourced the Lunar Foundation Model on Sept. 10, a system trained on more than 30 spatially aligned data layers9. Nine instruments across four lunar missions supplied them. Weights landed on Hugging Face under an Apache 2.0 license. Fine-tuning code runs through TerraTorch and a NASA-IMPACT repository. It finds craters, ice deposits, and volcanic terrain up to 23 percent more accurately than current methods. Planetary scientists gain a public tool for picking Artemis landing sites. A research release becomes direct input on where humans next set foot.
Apple Signs Every Pixel
Reference Image debuted on Sept. 910. The iPhone 18 Pro feature has the camera sensor sign every pixel it captures. Photos compares that unalterable file against the final image. Private Cloud Compute handles the processing, and Apple plans to add support for the SynthID standard later this year to flag AI-generated or edited images. Google and others back the C2PA-based Content Credentials standard. Apple chose a proprietary scheme of its own, which sets it apart. Photojournalists gain a tool built for their credibility problem. It arrives four days before the iPhone 18 Pro goes on sale in more than 65 countries.
What to watch
Congress adjourns in December4. Lehane's deadline gives lawmakers a fixed window before campaign season swallows the calendar. Watch Qualcomm's next earnings call2. It will show whether the Amazon deal moves that $15 billion target closer. Anthropic's next disclosure will show whether Coxon's resignation shifts any internal timeline1.
Sources
- HuffPost, "More Anthropic Employees Sound The Alarm About AI Safety Risks," HuffPost, Sept. 9, 2026, https://www.huffpost.com/entry/anthropic-ai-risks_n_6aa1d186e4b086ecc55fc61d
- The Motley Fool, "Forget Smartphones: Qualcomm Just Landed a Massive AI Deal With Amazon," The Motley Fool, Sept. 9, 2026, https://www.fool.com/investing/2026/09/09/forget-smartphones-qualcomm-just-landed-a-massive-ai-deal-with-amazon/
- TechCrunch, "Mistral Raises €3B as Sovereign AI Becomes Big Business," TechCrunch, Sept. 8, 2026, https://techcrunch.com/2026/09/08/mistral-raises-e3b-as-sovereign-ai-becomes-big-business/
- KFGO, "OpenAI Pushes for Mandatory National AI Safety Requirements," KFGO, Sept. 9, 2026, https://kfgo.com/2026/09/09/openai-pushes-for-mandatory-national-ai-safety-requirements/
- Positron AI, "Positron AI Raises $875 Million at a $5 Billion Valuation to Bring Its Next-Generation Inference Silicon to Market," PR Newswire, Sept. 10, 2026, https://www.prnewswire.com/news-releases/positron-ai-raises-875-million-at-a-5-billion-valuation-to-bring-its-next-generation-inference-silicon-to-market-302874601.html
- Anthropic, "Detecting and Countering Misuse of AI: September 2026," Anthropic, Sept. 10, 2026, https://www.anthropic.com/threat-intelligence-report-september-2026
- Bloomberg, "Microsoft AI-Focused Data Center Plan to Add 26 Gigawatts of Compute," Bloomberg, Sept. 10, 2026, https://www.bloomberg.com/news/features/2026-09-10/microsoft-ai-focused-data-center-plan-to-add-26-gigawatts-of-compute
- Meta, "Introducing Muse: The World's First Personal AI Agent Built for Everyone," Meta Newsroom, Sept. 8, 2026, https://about.fb.com/news/2026/09/introducing-muse-personal-ai-agent/
- NASA, "NASA, IBM Launch AI Foundation Model for Lunar Science," NASA Science, Sept. 10, 2026, https://science.nasa.gov/science-research/artificial-intelligence-lunar-foundation-model/
- TechCrunch, "Apple Has a New Way to Prove Your iPhone Photos Are Genuine," TechCrunch, Sept. 9, 2026, https://techcrunch.com/2026/09/09/apple-has-a-new-way-prove-your-iphone-photos-arent-ai-slop/