Anthropic filed confidentially with the SEC on Monday to go public. If it prices anywhere near its current valuation, it would be the largest IPO in history.

The company was valued at $965 billion in its most recent funding round last week, making it the world’s most valuable startup. That’s ahead of OpenAI, which clocked in at $852 billion in its last round. For context, the previous record IPO was Saudi Aramco at $29.4 billion in 2019. Even accounting for the difference between pre-IPO valuation and actual IPO pricing, we’re talking about a completely different scale.

This has been coming. Both Anthropic and OpenAI have been building toward public markets for months, but Anthropic moved first. The confidential filing means we won’t see the full S-1 with financials and risk factors for a while, but the company is now in the IPO pipeline.

The timing makes sense if you look at the business. Anthropic’s Claude has real traction with enterprises. The company has major distribution deals with AWS and Google Cloud. Revenue is reportedly in the billions, though we’ll need to wait for the S-1 to see actual numbers and, more importantly, burn rate.

What this tells us about the AI market

A $965 billion valuation prices in a lot of future growth. Investors are betting that foundation model companies can become platforms, not just API providers. They’re betting that whoever controls the best models will capture a massive chunk of enterprise software spending.

That’s a big assumption. The counter-argument is that models are commoditizing fast, margins will compress, and the real value will accrue to application companies or infrastructure providers like Nvidia. But Anthropic’s ability to raise at this valuation, and now file for an IPO, suggests that plenty of institutional investors believe the platform thesis.

It also puts pressure on OpenAI. Being second to market in an IPO isn’t necessarily bad, but OpenAI has positioned itself as the leader in AI. Getting beaten to the public markets by a competitor that was founded by ex-OpenAI employees adds a narrative wrinkle. OpenAI is expected to file soon, but “soon” now means playing catch-up.

The big test

The question is whether public market investors will buy in at anything close to these valuations. Private market AI valuations have been in their own universe for the past two years. Public markets tend to be more skeptical, especially when it comes to unprofitable high-growth companies.

Anthropic will need to show a path to profitability, or at least a credible story about how the unit economics work at scale. The S-1 will tell us a lot. What’s the gross margin on Claude API calls? How much is the company spending on compute? What does customer acquisition look like? How sticky are enterprise customers?

If Anthropic prices successfully and the stock holds up, it opens the door for a wave of AI IPOs. If it struggles, it could freeze the market for a year or more.

Nvidia makes its PC play

In other big industry news, Nvidia officially announced it’s going after the PC CPU market with RTX Spark, its new chip for AI agent laptops. Microsoft, Dell, and HP are all launching devices.

This is Nvidia’s bet that AI agents will drive a PC refresh cycle, and that it can grab a piece of the $200 billion CPU market that Intel and AMD have controlled for decades. The pitch is that RTX Spark can run sophisticated AI agents locally, which matters for privacy, latency, and cost.

Apple proved that Arm-based chips can deliver great performance and battery life on consumer devices. Qualcomm has been trying to replicate that in the Windows world with mixed results. Nvidia thinks it can do better, especially on the graphics and AI side.

The risk is price. These machines are expected to be expensive, potentially $2,000 or more for the high-end models. That’s a tough sell unless the AI agent features are genuinely useful and not just demos. The Verge’s hands-on with Google’s Gemini Spark agent, which would run on these machines, was positive but noted concerns about cost and privacy.

If Nvidia pulls this off, it’s another massive revenue stream for a company that’s already printing money from data center AI chips. If the devices flop, it’s an expensive distraction. But Nvidia has earned the benefit of the doubt.