business 6 min read

Anthropic IPO Proves the Market Still Loves AI, Just Not the Stocks

Anthropic staged the largest IPO in history this week, yet the Nasdaq posted its worst five-day loss since 2020. That contradiction tells you everything about where AI valuations stand—and who's really winning.

  • AI Investing
  • Venture Capital
  • Wall Street
  • Anthropic IPO
  • Tech Selloff

The Party and the Panic

Ben Narasin is throwing a party. There will be Krug champagne, vodka, and two jars of caviar per guest at his Silicon Valley home later this month. He hasn’t even invested in Anthropic.

Meanwhile, the Nasdaq has lost 10 percent over five trading days — the worst week for tech since March 2020. Nvidia dropped 23 percent in a single week, its sharpest pullback in 18 years. The C-suite crisis at the chipmaker has added fresh uncertainty to an industry already groaning under rate fears and anti-AI sentiment.

Two things happened at once this week. Neither cancels out the other.

The Number Nobody Is Debating

Anthropic raised $93.5 billion in its IPO and sold an additional $26 billion to Big Tech buyers at the offering price. Its stock gained 17 percent on debut, lifting the company’s market capitalization to roughly $1.8 trillion before settling near $1.6 trillion on a fully diluted basis. Only six American companies sit above it: Apple, Microsoft, Alphabet, Nvidia, Amazon, and SpaceX.

This was the single largest IPO in history, and it priced above its expected range after upsizing. In January, investors feared Anthropic might delay entirely. Brad Gerstner predicted the company would land at the low end of its pricing band. Instead, it walked away with more money than anyone expected.

For the venture firms that backed it, the returns rewrite career trajectories. Menlo Ventures and Lightspeed each hold roughly $25 billion worth of Anthropic stock at the IPO price. Bessemer holds about $10 billion. Matt Murphy told CNBC that more than five of Menlo’s limited partners received over $1 billion in returns from the company alone.

Somesh Dash at IVP called it a punctuation mark in the history of both humanity and computation. That may sound like founder hype. The numbers back it up.

The Contradiction That Matters

Here is what most reporting misses: Anthropic is valued as a software business. It generates revenue by selling access to Claude, its language model, not by manufacturing chips or building data centers. Yet the broader AI ecosystem it depends on — Nvidia GPUs, Amazon and Google cloud compute, memory supplies, energy contracts — is having a brutal week.

The market is effectively saying: we love your outcome, we doubt your infrastructure.

That distinction matters because it reveals a structural anxiety running through every institutional portfolio right now. Investors can tolerate a single spectacular winner. They cannot reconcile that winner with the accelerating costs of doing AI at scale. Compute prices have not come down. Data center permits face regulatory headwinds after November ballot measures passed alongside the midterms. The House and Senate flip to Democratic control in January adds another variable the market cannot yet price.

And Nvidia — the company Anthropic relies on and co-invests with — is suddenly leaderless after Jensen Huang’s departure, even if temporarily. Every AI startup that ties its roadmap to accelerated computing just watched its primary supply chain send shockwaves through its own valuation model.

Who Actually Wins Here

The winners are clear. The venture firms that held early stakes walk away with returns most will never replicate across an entire fund cycle. Eleven hundred and twenty-six technology companies went public in 2021; only 34 did so in 2025, according to Jay Ritter’s data. The exit drought is over.

Big Tech is a secondary winner, though the calculus is subtler. The $26 billion they deployed at the offering price locks in anchor positions in the most capable open-weight competitor to OpenAI at a known entry point. If Claude continues closing the gap on GPT-class models, those holdings appreciate. If they don’t, the companies still own strategic positioning in a category everyone else is forced to play in regardless.

The loser this week is any investor who tried to assign a clean sector multiple to AI publicly traded names. The market rewarded one isolated company while punishing the entire hardware and cloud stack. That kind of bifurcation does not signal confidence in the thesis. It signals confusion about where value actually sits inside the thesis.

Second-Order Effects You’re Not Seeing Yet

The fallout from this split-screen dynamic is already rippling through parts of the market that have nothing to do with large language models. Cloud providers that had been betting heavily on AI-driven demand growth are recalibrating their capital expenditure timelines. Amazon Web Services and Google Cloud both faced pressure from analysts questioning whether AI revenue would materialize fast enough to justify the billions being spent on power infrastructure and server deployment. Their shares moved lower even before Nvidia’s stumble amplified the concern.

Energy companies that had positioned themselves as essential partners to data center expansion are also feeling the shift. AI’s insatiable electricity demand was supposed to create a secular tailwind for utilities and independent power producers. But when the market questions whether the entire AI buildout is proceeding too rapidly relative to regulatory and grid constraints, that story loses its polish. Power purchase agreements that looked airtight twelve months ago are now being renegotiated or delayed.

There is a quieter consequence as well. The venture community has spent three years telling founders that going public was dead. The narrow window for tech IPOs, combined with punishing volatility, drove countless companies toward SPACs or extended private lifespans. This week proved that windows can reopen overnight when the right company arrives. Expect a rush of late-stage startups to reassess timing rather than accept discount Secondary rounds from distressed buyers.

What Comes Next

SpaceX hit the market five months ago and made 2026 a monster year for venture IPO proceeds before Anthropic arrived. OpenAI is widely expected to file for 2027. IVP’s Dash said half a dozen AI companies with north-of-$10 billion valuations are already in prep mode. The window is open.

But the macro picture makes each subsequent filing riskier than the last. Rising rates compress growth multiples. Anti-AI sentiment has moved from blog posts into budget meetings at major corporations. Regulatory uncertainty around data centers and model licensing is accumulating faster than policy can address it.

Narasin called this the bedtime story for future VCs. He may be right about the story. The question investors need to answer is whether the lesson — that extraordinary returns are possible in AI — outweighs the warning embedded in the week’s market action: the ecosystem that enables those returns is far less stable than the valuations suggest.

The paradox of this moment is that Anthropic’s success could make the next round of AI debuts harder, not easier. Every company that follows will be measured against a benchmark that assumed perfect conditions — strong demand, supportive sovereign interest, and a hardware partner that wasn’t simultaneously collapsing. None of those assumptions held for a full week. The investors who understand that distinction will position accordingly. Those who simply read the headline number and assume the gold rush is unstoppable will repeat the same mistake Nvidia’s sellers just made.