Welcome to AI Decoded, Fast Company‘s weekly newsletter that breaks down the most important news in the world of AI. I’m Mark Sullivan, a senior writer at Fast Company, covering emerging tech, AI, and tech policy.
Sign up to receive this newsletter every week via email here. And if you have comments on this issue and/or ideas for future ones, drop me a line at sullivan@fastcompany.com, and follow me on X @thesullivan.
Another OpenAI exec leaves the company
OpenAI hired Chris Malone to manage its data center build-out in March 2025, just after the company announced its massive Stargate data center project with SoftBank and Oracle, The Wall Street Journal reported. Stargate was unveiled with fanfare in a chummy Oval Office announcement featuring OpenAI’s CEO, Sam Altman; SoftBank’s CEO, Masayoshi Son; Oracle’s chairman and CTO, Larry Ellison; and President Donald Trump.
Malone, who left OpenAI last week, had already been moved down in the organization. After a reorganization earlier this year, he no longer reported to the president, Greg Brockman, and instead became cohead of a technical engineering team under VP Sachin Katti. In July, OpenAI promoted Uday Ruddarraju to chief technology officer of computing capacity, reporting to Brockman, while Brent Mayo, hired from xAI this year, was put in charge of keeping projects on schedule.
A shuffle in the data center arm of OpenAI’s operation is conspicuous. The exorbitant cost of data centers is a major reason the big AI labs have taken on so much investment money. Labs like OpenAI and Anthropic believe they need massive computing power both to serve customer demand for artificial intelligence apps and services and to support their own R&D, which they hope will eventually produce AI far more intelligent than humans. Meanwhile, OpenAI remains far from profitable and has its sights set on an IPO next year.
And there’s no end in sight. In July, OpenAI raised its projected compute budget to roughly $750 billion through 2030, up from about $600 billion. For much of its history, OpenAI relied on leasing cloud-computing capacity from third parties like Microsoft and Amazon Web Services, but later shifted toward owning its own compute infrastructure, the central idea behind Stargate. Now the company is moving back toward a middle ground by leasing entire data center facilities. This month, it signed a 10-gigawatt data center lease in Ohio with SoftBank’s SB Energy that is partly backed by an Nvidia financial guarantee.
OpenAI is committing close to its own valuation in future infrastructure payments, increasingly underwritten by the companies selling it the chips, and it has now cycled through the leadership overseeing that build-out twice in 18 months.
Malone’s exit is one of 13 senior departures from OpenAI in 2026, by Business Insider‘s count. Among them are the chief revenue officer, Denise Dresser; the longtime COO, Brad Lightcap; and the product and business chief, Fidji Simo, Altman’s second-in-command, who stepped down in July.
Bill Gates says nobody is planning for what comes next
Bill Gates published an essay of nearly 6,000 words on Wednesday outlining his growing anxiety over how slowly society is responding to the changes AI will bring. He believes AI will permanently eliminate a large share of jobs, put cyberattack and bioweapon capabilities in the hands of people who never had them, and harm children’s development by replacing human relationships. Gates argues these aren’t distant threats, but ones that are already beginning to arrive. “There is no plan to ease the entry into the AI era,” he writes. He has some ideas for how society should respond. These three stand out.
- Jobs go first, so protect some by law. Gates expects AI to replace workers in law, customer service, medicine, software, and manufacturing within a decade rather than over generations, leaving far fewer jobs than exist today. He proposes creating “human reserved” jobs: work that machines could perform but that society decides should remain in human hands. He compares the idea to nature reserves, “places where we could put buildings and roads, but we choose not to.”
- Levy a tax on AI tokens and robots. Right now, the tax code pays companies to fire people. An employer who hires a worker pays payroll taxes, but an employer who buys a robot writes it off as a business expense. “The tax system nudges you toward replacing people with machines,” Gates says. Meanwhile, income tax revenue falls exactly when displaced workers need the safety net most. Gates wants that tax on AI tokens and robots in order to close the gap and fund retraining.
- Create overarching regulatory bodies. As things stand, no agency owns problems created by artificial intelligence. AI cuts across employment, energy, elections, public health, law enforcement, and national security, and Gates argues that existing institutions each see only their own piece of it. That creates the risk that an AI-enabled attack could succeed simply because no one thought stopping it was their responsibility. “None of our current institutions were designed to handle a technology that spreads so fast,” Gates writes. He proposes national bodies that can set priorities across agencies, along with an international organization drawing on models from nuclear inspections, aviation regulation, and treaties protecting the ozone layer.
The SEC asks banks about the AI hedge fund that nearly blew up
The Securities and Exchange Commission has subpoenaed Goldman Sachs, JPMorgan, Citigroup, and Bank of America for information about the hedge fund Situational Awareness’s trades, use of leverage, and communications with the banks, The New York Times reports. The AI-focused fund, run by a former OpenAI researcher, Leopold Aschenbrenner, reportedly plunged from about $45 billion to roughly $10 billion in late July. That’s when a tech selloff triggered margin calls and forced the fund to unwind concentrated positions in companies including the memory-chip maker SK Hynix and the AI cloud-computing company CoreWeave. Both stocks have since rallied.
Microsoft won’t tell investors what Azure earns
Microsoft’s latest annual report said revenue from Azure and other cloud services grew 41% last fiscal year without giving a dollar figure, a prior-year comparison, or any expense or profit data, The Wall Street Journal reports. Microsoft reports Azure earnings in the same segment with its high-margin legacy software business, obscuring the cost of its expensive AI datacenter buildout. That may not sit well with investors who are—rightly—nervous about the “AI bubble.”
The model that quietly took over OpenRouter is Chinese and open
Beijing-based Z.ai (also known as Zhipu) announced a new open-weights AI model called GLM-5.3-Flash on Wednesday. It can work with text and images has a huge 1-million-token context window, meaning it can process unusually large amounts of information at once.
Z.ai says it outperforms its previous GLM-5.2 model at roughly a tenth of the price, while running entirely on Chinese AI chips. The company also confirmed that GLM-5.3-Flash is the model that spent the past week at or near the top of OpenRouter and OpenCode leaderboards under the anonymous name Ox Alpha.
More AI coverage from Fast Company:
- Meet Eloise, the Florida donkey taking on the data center boom
- What’s up with all the tech titan manifestos?
- Why Spirit Airlines’ internal data has become a hot commodity for AI companies
- Corporate America is embracing AI more slowly than the hype suggests—but the pace is increasing
Want exclusive reporting and trend analysis on technology, business innovation, future of work, and design? Sign up for Fast Company Premium.