OpenAI Just Became a $500B Company
4IR - Daily AI News
Welcome back to 4IR. Here’s today’s lineup:
OpenAI just became a $500B company—and Microsoft owns 27% of it - ChatGPT maker completes restructuring into public benefit corporation, giving Microsoft massive stake and clearing path to IPO while Sam Altman gets zero equity
NVIDIA is building robot factories in America—starting with its own chips - Jensen Huang announces Blackwell GPUs now manufactured in Arizona, 100,000 robotaxis with Uber by 2027, and a $1B Nokia partnership, all while major manufacturers build AI-powered factories with Omniverse
Anthropic claims Claude 4 beats everyone at coding—including OpenAI - New Opus 4 model leads SWE-bench at 72.5% and can work continuously for hours, launching with extended thinking that uses tools like web search mid-reasoning
OpenAI just became a $500B company—and Microsoft owns 27% of it
The story: OpenAI finally finished the restructuring everyone’s been waiting for. The company went from its messy nonprofit-plus-profit setup to a clean public benefit corporation worth $500 billion. Microsoft got 27% ownership—that’s a $135 billion stake—plus guaranteed access to all OpenAI models through 2032, including any future super-smart AGI systems. The deal frees OpenAI from the restrictions that started when Microsoft first invested in 2019, when Microsoft basically controlled everything in exchange for providing computing power. Now OpenAI can raise money from anyone and make deals with other cloud providers. And Sam Altman? He gets zero equity. His salary is $76,000 a year and that’s it.
What we know:
OpenAI restructured into public benefit corporation valued at $500 billion
Microsoft receives 27% stake worth $135 billion with model access through 2032
OpenAI committed to purchase $250 billion in Azure cloud services over time
Sam Altman gets no equity in restructured company (makes $76,000 salary)
ChatGPT now has 800 million weekly users, IPO is “most likely path”
Why it matters: This is OpenAI admitting it can’t stay nonprofit and compete with everyone else. The $500B valuation puts OpenAI up there with companies like Visa and forces every other AI company to justify their own sky-high numbers. Microsoft’s 27% stake is smart: they get huge upside without owning everything, keep access to the models even if OpenAI builds AGI, and locked in $250 billion in cloud computing sales. For OpenAI, losing Microsoft exclusivity means they can finally shop around for computing power and raise money without asking Microsoft’s permission first. The IPO hint is the big tell—they’re getting ready to go public because even $500B isn’t enough money for what’s coming next.
The Sam Altman no-equity thing is interesting theater. It looks humble—the CEO who just wants to build AGI, not get rich. But he already made his money from Y Combinator and other investments. The real winner is Microsoft, who turned a risky bet on a nonprofit into a $135B stake without the headaches of actually owning the company. OpenAI gets freedom, Microsoft gets guaranteed profits, and both can tell regulators they’re not a monopoly. That $250B Azure deal is insurance—even if things go bad later, Microsoft already won. Meanwhile, 800 million weekly users proves ChatGPT isn’t hype anymore—it’s essential.
NVIDIA is building robot factories in America—starting with its own chips
The story: Jensen Huang turned NVIDIA’s Washington DC keynote into a massive manufacturing announcement. The big news: Blackwell AI chips are now being made in Arizona—the first time NVIDIA’s best GPUs have been built in America. But that was just the start. NVIDIA announced deals with Uber to put 100,000 self-driving robotaxis on the road by 2027, a $1 billion investment in Nokia to build AI-powered 5G and 6G networks, and a partnership with Oracle to build the Department of Energy’s biggest AI supercomputer using 100,000 Blackwell GPUs. Meanwhile, companies like Foxconn, Toyota, TSMC, Caterpillar, and Lucid Motors are all using NVIDIA’s Omniverse software to build virtual copies of their factories. The message was clear: America’s manufacturing comeback runs on NVIDIA.
What we know:
Blackwell GPUs now being made in Arizona (first US manufacturing of top chips)
Partnership with Uber for 100,000 robotaxis by 2027 with Stellantis, Lucid, Mercedes-Benz
$1 billion Nokia investment for AI-enhanced 5G and 6G networks
DOE supercomputer with Oracle featuring 100,000 Blackwell GPUs at Argonne National Lab
Major manufacturers building virtual factory copies: Foxconn, Toyota, TSMC, Caterpillar, Lucid
Why it matters: NVIDIA is positioning itself as essential to American manufacturing, not just a chip seller. Making chips in Arizona answers Trump’s “bring jobs back” demand and reduces dependence on Taiwan. The Nokia deal takes on Chinese telecom companies. The DOE supercomputer shows NVIDIA is embedded in government work. And the factory software deals mean every major manufacturer now depends on NVIDIA’s platform to design their AI factories. This isn’t just selling chips—it’s becoming the foundation for America’s entire AI-powered industrial future.
The robotaxi deal with Uber is the hidden story here. NVIDIA already owns AI training and data centers. Now they’re grabbing self-driving cars before Tesla locks down that market. The “physical AI” branding is brilliant—it makes factory robots sound like the next big thing instead of just better software for old equipment. Watch China’s response—this whole announcement basically said America is betting everything on NVIDIA for the next industrial revolution. Jensen just turned a chip company into national infrastructure.
Anthropic claims Claude 4 beats everyone at coding—including OpenAI
The story: Anthropic launched Claude 4 with two versions—Opus 4 and Sonnet 4—and came out swinging on performance. Opus 4 scored 72.5% on SWE-bench, the toughest coding test out there, claiming the title of “world’s best coding model.” The real improvement is staying power: Opus 4 can work for hours on complicated tasks needing thousands of steps, something older models couldn’t do without losing focus or screwing up. Both models got “extended thinking with tool use,” which means Claude can now switch between deep thinking and grabbing stuff like web search results while solving problems. Claude Code, their developer tool, left beta with plugins for VS Code and JetBrains. Prices stayed the same: Opus 4 at $15/$75 per million tokens, Sonnet 4 at $3/$15.
What we know:
Claude Opus 4 leads SWE-bench at 72.5%, Sonnet 4 at 72.7%
Models can work for hours on complex tasks with many steps
Extended thinking now uses tools (web search, etc.) while reasoning
Claude Code now officially available with VS Code and JetBrains plugins
Pricing unchanged: Opus 4 at $15/$75 per million tokens, Sonnet 4 at $3/$15
Why it matters: The coding competition just got fiercer. Every AI company says their model writes the best code because that’s where the money is—developer tools, code helpers, automatic bug fixing. Anthropic’s 72.5% on SWE-bench beats OpenAI’s recent scores and forces Google and Meta to respond. The “works for hours” claim matters more than test scores—if Claude can actually stay focused on big refactoring jobs without supervision, that’s a real AI agent, not just fancy autocomplete. The extended thinking with tools is smart: instead of making users pick between quick answers or deep thinking, Claude decides when to think harder and when to look stuff up.
The real fight isn’t about test scores—it’s about locking customers in. OpenAI restructured to grow bigger. NVIDIA is building physical infrastructure. Anthropic is betting developers will want their model for serious coding work. Different strategies, same goal: become essential before everyone’s models are basically the same. Keeping prices flat while getting better is Anthropic’s quiet power move—we’re improving without squeezing you for more money, unlike some “open-source” companies that just tripled their prices overnight. If Claude Code becomes what serious developers use by default, Anthropic wins even if their model isn’t always the fastest or cheapest.
Note: Commentary sections are editorial interpretation, not factual claims
The Microsoft stake is huge. What if that much control limit future public benefit?