Headline valuations flatten AI worth into simple dollar cards, but the actual conversation is about scarcity, timing, and trust. Anthropic's Series H made it the most valuable private AI lab by headline price, secondary markets then pushed its implied value past $1 trillion, and Chinese model competition is forcing everyone to rethink what those multiples actually buy.
The 30-day read
Anthropic has the momentum in both primary and secondary markets, but the premium is driven as much by a seller drought as by fundamentals. OpenAI is still larger in users and funding raised, yet its shares are trading at a discount to the last primary round. Meanwhile, cheap Chinese models are eroding the usage share of US frontier labs. The valuation story is no longer just about who has the best model — it is about who can raise, burn, and convert users before public-market discipline arrives.
1 · What people are sayingAnthropic passes OpenAI, and the market argues about what the number means
The dominant thread in the last 30 days is Anthropic's valuation flip. In late May it raised a $65 billion Series H at a $965 billion post-money valuation, putting it above OpenAI's March round at $852 billion. By early July, secondary platforms were pricing Anthropic shares at an implied $1.2 trillion — a level Caplight's CEO called the most sought-after price in venture-secondary history.
Not everyone trusts the figure. Menlo Ventures partner Matt Murphy, an early Anthropic backer, described secondary-market prices as a "noisy signal" driven by near-zero supply and IPO speculation. The scarcity is real: Rainmaker Securities' CEO said that even at $1.2 trillion, virtually no one is willing to sell.
"Anthropic is the most sought-after company in the history of the venture secondary market."Javier Avalos, CEO of Caplight — July 2026
On X, traders treated the numbers as a Rorschach test. One account highlighted three high-conviction AI stocks Wall Street sees doubling; another noted that bulls counting on xAI winning the race must believe the company is really an AI play, not just a data-center landlord. Semiconductor names also appeared as valuation caution tales: Arm and Marvell both pulled back as investors rotated out of richly valued AI and semiconductor stocks.
2 · The numbers behind the flipRevenue, multiples, and the IPO clock
The valuation crossover rests on diverging revenue trajectories. Anthropic disclosed an annualized run-rate above $47 billion, up from $30 billion earlier in the year. OpenAI's run-rate is widely cited around $24 billion. At their primary valuations, that implies Anthropic trades at roughly 20 times run-rate revenue while OpenAI sits near 35 times.
The IPO timing adds pressure. Anthropic filed confidentially in June and is widely expected to target a Q4 2026 listing; OpenAI has also filed but faces reported internal disagreement between Sam Altman's 2026 ambition and CFO Sarah Friar's preference for 2027. The first public S-1 will force the question of whether private-market scarcity premiums survive contact with liquid markets.
3 · Two framingsDevelopers vs. investors
The same set of facts produces two very different reads.
Developers & practitioners
- Chinese models took 80% of OpenRouter token share as US providers fell to 20%
- Open-source and low-cost APIs are making frontier models interchangeable
- Anthropic is giving YC startups up to $500K in free credits to lock them in early
- Frame: model quality is converging; price and distribution are the new moats
Investors & finance
- OpenAI's 35x revenue multiple vs. Anthropic's 20x shapes relative demand
- Secondary scarcity, SPV structures, and rights of first refusal distort price discovery
- AI mega-rounds absorbed so much capital that seed funding fell 27% in H1 2026
- Frame: private markets are pricing scarcity, not just fundamentals
Where the two frames agree is on the risk: both note that current prices assume continued hypergrowth and a friendly IPO window. If either assumption breaks, the cascade could hit every name in the ranking.
Headline rankings treat these companies as static cards with fixed 2026 price tags. The last 30 days show the opposite: the numbers are moving, contested, and heavily dependent on whether sellers exist, IPOs happen on schedule, and cheap Chinese models keep eating usage share. The ranking is a snapshot; the market is a fight.