#Anthropic与OpenAI竞争升级 OpenAI and Anthropic Race: Who Will Be the Top US Large-Model Stock?

The most watched competition in Silicon Valley’s AI track in 2026 will undoubtedly be OpenAI and Anthropic’s fight to become the top US large-model stock. Both leading companies plan to complete an IPO by the end of 2026, and a contest over listing speed, financial strength, and industry voice has officially begun.
Behind this race are multiple challenges. Internal conflicts at OpenAI have become prominent: CEO Sam Altman is pushing for a listing within the year, but its CFO has opposed it, citing that organizational preparation is insufficient and the risk of compute investment is too high. The company expects that cumulative losses before listing will exceed 2000 billion US dollars, that compute expenditures in 2028 will reach 121 billion US dollars, and that the outlook for profitability is unclear.
By contrast, Anthropic’s revenue growth momentum is strong. Its annualized revenue has already surpassed 300 billion US dollars, the number of enterprise customers has doubled, and its financial model suggests it could achieve profitability earlier than OpenAI. However, the company also carries massive compute costs, has signed large compute agreements with Google and Broadcom, and its revenue calculation method has sparked controversy over whether it is “presenting financials more favorably.”
In addition, the large-model industry generally faces a pricing dilemma. Low-cost Tokens and calls to third-party tools can easily lead to huge losses, and the business model has not yet been fully proven. Although both companies are sprinting toward an IPO, they are both mired in the predicament of “burning money for growth.” Compared with the hollow prestige of being the “top stock,” how to establish a sustainable profitability model is the core proposition that AI large-model companies urgently need to solve.
OpenAI and Anthropic’s listing race is the IPO competition most closely watched in Silicon Valley.
Both companies do not want to be left behind; both want to complete their IPOs by the end of 2026. Yet behind the title of “top large-model stock,” their financial conditions and internal timing differ markedly.
OpenAI’s CEO Sam Altman hopes to list as soon as possible, but his CFO believes the company is not ready yet. Anthropic’s revenue growth is fast, but it is also under enormous pressure from computation costs. Both companies rely on large-scale compute investment to maintain competitiveness, and the payback cycle for such investment is uncertain. 

01
Internal Disagreements at OpenAI
Altman wants OpenAI to list at the earliest in this year’s fourth quarter, but according to confidential financial documents that OpenAI presented to investors ahead of the latest funding round, the company expects cumulative losses will exceed 200 billion US dollars before it begins generating positive operating cash flow.
A financial document shows that OpenAI expects its compute spending to reach 121 billion US dollars in 2028. Even if sales nearly double compared with the prior year, the company still expects to lose 85 billion US dollars. This scale of losses is extremely rare among publicly listed companies.

But CFO Sarah Friar’s view differs from Altman’s; she does not believe the company can be ready to go public in 2026.
Friar’s reasoning is that process-oriented and organizational work has not yet been put in place, and the risk brought by spending commitments is too high. She is also not sure whether OpenAI will need to invest that much money in AI servers in the coming years, and whether revenue growth—which has slowed—can support those commitments.
In addition, Amazon and Nvidia currently hold a substantial amount of equity in OpenAI. As tightly bound, “strategic shareholders” that come with strong commitment and high-stakes bets, they may also affect the timing of the company’s listing.
As for the disagreement between the CEO and CFO, in public, Friar has shown a willingness to play it down—she merely emphasizes that an IPO is “currently not under consideration,” because OpenAI is still working to “bring the company to a sustained upgrade state that matches our current scale.”
From the stance toward the IPO, it becomes clear that subtle changes have emerged between Altman and Friar.
In August 2025, Friar stopped reporting directly to Altman and instead began reporting to Fidji Simo, who had joined at that time to lead OpenAI’s application business. Such an arrangement is not common in large companies; the CFO typically reports directly to the CEO.
Multiple people who have worked with Friar told The Information that Altman excluded her from certain conversations related to the company’s financial plans. For example, in recent months, when Altman discussed server spending with one of OpenAI’s largest investors, Friar was not present. By contrast, in earlier conversations on the same topic, she had been involved.
Another person who attended an OpenAI senior leadership meeting earlier this year said that the meeting involved major financial decisions, and Friar was not invited—also unusual.
It is worth noting that the concerns Friar has privately expressed are quite similar to Anthropic CEO Dario Amodei’s recent public comments.
In a podcast this February, Amodei said, “Even if the technology really develops at the fastest speed I forecast, it’s still not clear whether revenue can keep up. But the problem is, you buy data centers according to that (expected revenue) pace. If your judgment is off by one or two years, it could be catastrophic.”
Amodei believes that even if you’re only off by one year—or if the growth rate is not tenfold but fivefold—the result is bankruptcy. He then added, “I have a feeling some companies haven’t really calculated this seriously. They don’t even know how much risk they are carrying.”
What companies does he mean?

02
Is Anthropic “Beautifying the Financial Statements”?
Based on financial data obtained from The Wall Street Journal, Anthropic’s revenue growth momentum is stronger than OpenAI’s.
Its annualized revenue has already surpassed 30 billion US dollars, while the figure was about 9 billion US dollars at the end of 2025. When announcing its Series G financing in February, Anthropic stated that more than 500 enterprise clients had annual spending exceeding 1 million US dollars. Now, that number is over 1000.
It doubled in less than two months.
Profit Comparison Between OpenAI and Anthropic
According to figures compiled by The Wall Street Journal, even including training costs (bar chart), Anthropic is expected to achieve profitability in 2028, while OpenAI would need to reach 2030; and if training costs are excluded (line chart), Anthropic would basically break even in 2024 and 2025.
Mizuho Financial Group analysts estimate that Broadcom’s AI revenue from Anthropic will reach 21 billion US dollars in 2026 and 42 billion US dollars in 2027.
Annualized Revenue by Business Segment for OpenAI and Anthropic
It should be pointed out that the two companies differ in how they calculate revenue, which makes OpenAI’s revenue growth rate look less rapid than Anthropic’s.
One key difference is that—Anthropic includes its technology sales conducted through cloud partners in revenue, while OpenAI does not. This makes Anthropic’s reported revenue look more attractive on paper; Anthropic responds that this aligns with standard accounting practices because the company is the principal in the transaction.
In addition, although it says it is afraid revenue won’t keep up, Anthropic has never paused compute investment.
According to Anthropic’s official disclosures, it has already signed new agreements with Google and Broadcom to obtain next-generation TPU compute capacity in the range of several gigawatts, which is expected to go live starting in 2027. The vast majority of new compute facilities will be located in the United States. Anthropic CFO Krishna Rao said this is “the most important compute investment commitment to date.”
Inference cost is another heavy burden.
Free Cash Flow Comparison Between OpenAI and Anthropic
Although ChatGPT’s consumer user revenue is relatively large, paying users make up only a small portion—meaning that most inference costs are not converted into revenue. Anthropic’s situation is slightly better: most of its revenue comes from enterprise customers.
An OpenAI spokesperson said the company supports free users to promote technology adoption, and that it can profit in ways such as advertising or converting free users into subscribers. The spokesperson emphasized that the company prioritizes growth over profit.

 03
The Predicament of Pricing Models
How large-model companies should price to avoid losses is an unsolved problem.
Luo Fuli, the head of Xiaomi’s large-model business, recently analyzed this in a post. She believes Claude Code’s subscription system is ingeniously designed, but it may not be profitable—perhaps even operating at a loss—unless Anthropic’s API profit margin can reach 10 to 20 times, which she doubts.

“In a single user query, some wrapped tools will initiate multiple rounds of low-value tool calls, and each round is a separate API request. Each request carries an extremely long context window, often exceeding 100,000 tokens. Even when caching hits, it’s still wasteful,” Luo Fuli said.
According to Luo Fuli’s calculations, the actual number of requests per query is several times that of Claude Code’s own framework. Converted into API pricing, the real cost could be dozens of times the subscription price, which is a “huge pit.”
Luo Fuli said that “before large language model companies find a way to set pricing that is both reasonable and does not result in losses, they should not blindly launch price wars.”
She believes that selling Tokens at extremely low prices, while opening the door widely for third-party encapsulation tools, seems beneficial to users but is actually a trap. “Selling tokens at extremely cheap prices while opening the door to third-party harnesses looks wonderful to users, but it’s a trap. If users waste their attention on low-quality agent harnesses, unstable and slow inference services, and models downgraded to reduce costs, then in the end they still won’t accomplish anything—this is not a healthy cycle for user experience and retention.” 

04
Conclusion
Both OpenAI and Anthropic are competing for the US “top large-model stock,” and both are tied to continuous fundraising and high-stakes wagers. Both face the need to keep burning money, but their commercial returns have not yet been fully validated.
However, their situations are also clearly different.
OpenAI has internal disagreements about the timing of the listing, while Anthropic needs to control compute costs as revenue grows quickly. And judging by industry buzz, Anthropic’s reputation has begun to overtake OpenAI.
It could be said that on the path of large-model exploration, no one will be the first forever. Once the technical roadmap is misjudged, it’s possible to be surpassed by competitors. Although OpenAI is the first company to unlock ChatBot AI assistants, it may not always be able to stay ahead in every business area.
In fact, from the perspective of industry health, amid steadily rising compute costs and still-mature pricing models, how to build a sustainable business model may be more important than the title of “top stock.”
But that judgment must exclude the people who tell the story.