OpenAI Misses, Anthropic Takes All: The Power Balance of the AGI Game is Tilting

When the AGI clause is revoked and revenue targets are missed, who is the real frontrunner?

The financial figures from the past week are more worth reading carefully than any model benchmark score. For the first time, OpenAI has been reported by its own CFO’s internal memo to have missed revenue and user targets; in the same week, Anthropic secured a $40 billion commitment from Google. This is no coincidence; it is the power structure of the AGI game being re-marked.

Background Context

I wrote in last Sunday’s weekly report: On 4/27, OpenAI and Microsoft re-signed their contract, removing the “AGI Clause” that had been entangled for years. At the time, the market was focused on who would lose what once this “AGI achievement” trigger disappeared. But I tend to think that was just a surface-level action. The real story is: when AGI is no longer a contractual trigger, the winner can only be measured by the most basic metrics—cash flow, user count, and profit margins. And over the past seven days, these metrics have pulled OpenAI and Anthropic onto different tracks.

More importantly, once the AGI clause was revoked, Microsoft lost the constraint of “losing exclusivity upon achieving AGI,” but OpenAI also lost the buffer of “being able to freely use other clouds upon achieving AGI.” Now that both sides have become a standard commercial relationship, OpenAI must support its global AI capital expenditure (capex) plan—reportedly costing $660 billion—through revenue, without the halo of the AGI myth. This pressure exploded on 4/28.

Multiple Perspectives

Official Version

On 4/28, OpenAI adopted an “official denial” stance regarding the Wall Street Journal’s report, stating that the company is “focused on expanding compute capacity,” avoiding a direct response to whether the internal revenue memo was true. On the same day, Anthropic had no specific official statement because they didn’t need to—on 4/22, Bloomberg had already reported that Google would invest up to $40 billion; on 4/24, a series of media outlets confirmed that the first $10 billion injection was finalized at a $350 billion valuation, with the remaining $30 billion to be disbursed based on milestones.

Industry Perspective

The Wall Street Journal, citing internal memos and multiple people familiar with the matter, pointed out that OpenAI failed to meet its own target of “1 billion weekly active users for ChatGPT by the end of the year”; the annual revenue target for ChatGPT has also slipped out of range. The report cited two reasons: Google Gemini accelerated its market grab at the end of the year, while Anthropic continued to win orders in code and enterprise clients, causing OpenAI’s monthly revenue to fall below targets several times in the past few months.

Even more noteworthy is that Fortune subsequently reported: OpenAI CFO Sarah Friar has privately disagreed with Sam Altman on this issue. Friar warned colleagues, “If revenue growth does not accelerate, financing for future compute contracts will become difficult.” This is the first concrete evidence I have seen of a C-level split within OpenAI—not a safety researcher leaving, but the CFO directly questioning the financeability of the capex plan.

The market reaction was even more direct. During trading on 4/28, Oracle, which is most deeply tied to OpenAI’s compute, fell about 4%, Broadcom fell 4%, and AMD fell 3%; in Asia, SoftBank, one of OpenAI’s largest investors, plunged nearly 10% during the session. This is the first time in over a year that “OpenAI bad news” has single-handedly dragged down a basket of tech stocks.

Objective Data

• Anthropic Annualized Revenue (ARR): End of 2024 ~ $1 billion → End of 2025 ~ $9 billion → 2026/3 exceeded $30 billion (tripled in about 4 months)
• Anthropic Investors Combined: Amazon (up to $30 billion), Google (up to $40 billion), Microsoft (up to $5 billion), Nvidia (up to $10 billion). The four players in Silicon Valley who can decide GPU and cloud allocation have all bet on Anthropic.
• VC Valuation Offers: Over $800 billion (Anthropic is not accepting for now)
• OpenAI Annualized Revenue: Over $25 billion (self-claimed), but growth rate questioned by internal memo
• After the AGI clause was removed, OpenAI’s revenue sharing with Microsoft was changed to “ends in 2030 + total cap,” unlinked from technical milestones

Xiao Jian’s Judgment

This is the structural signal I believe is most worth paying attention to: When the AGI clause is revoked, and when OpenAI can no longer use “AGI achievement” as an implicit amulet for future valuation, the market is left with only the most classical method to test it—can your revenue growth support your compute commitments? Currently, the answer is “questionable.”

I tend to think that the “Big Three” (OpenAI / Anthropic / Google DeepMind) tripartite narrative of the past 18 months has now converged into a new pattern: “Anthropic + Google axis vs. isolated OpenAI.” Microsoft’s role is particularly interesting: It retains a 27% stake in OpenAI, while having switched its Copilot agent backend to Claude, and has also invested $5 billion in Anthropic. This is not a double bet; this is a hedging behavior of a company putting eggs into three baskets. When the “biggest backer” starts hedging, can you say it still has confidence in OpenAI?

But I also need to hit the brakes. OpenAI’s $25 billion annualized revenue is still the highest in the industry, and ChatGPT remains the product with the most brand power on the consumer side; Anthropic’s revenue structure is more concentrated in the enterprise (Claude Code is the main driver), which means it is more vulnerable to a single track. Anthropic looks like it has won everything, but it is actually betting all its chips on the “autonomous code agent” table. If one day OpenAI releases the next-generation Codex that matches Claude’s performance and starts an API price war, Anthropic’s growth curve risks bending again.

Finally, I want to return to an old question: Do these financial figures tell us how far we are from AGI? No, zero point zero. METR’s Time Horizon 1.1 shows that the duration of tasks AI can complete independently is still in the hour range; on ARC-AGI-3, humans are less than 1% against the strongest models. But what financial indicators tell us is another equally important thing: who will decide the training of the next generation of models, who will control who can rent compute, and who will define what is “good enough.” On this long road to AGI, this struggle for allocation power is what readers should really be tracking every week.

Worth Continuing to Observe

1. Whether Sarah Friar will speak publicly or resign—The divergence in strategy between CFO and CEO usually appears in internal memos first and only surfaces 3-6 months later. If Friar leaves OpenAI before the end of Q2, this is the strongest signal of the collapse of the internal financing route.
2. The official stance at Microsoft Build 2026 (mid-May)—Whether Microsoft will increase Claude’s integration to a level close to OpenAI’s. If so, “Microsoft after the AGI clause” confirms it has chosen the “diversify risk” route.
3. Whether OpenAI launches the next-generation Codex early—This is the only lifeline of Anthropic that can be directly attacked. Once a price war starts, the story of Claude’s $30B ARR will be immediately repriced.
4. Whether Anthropic accepts the $800 billion VC offer—Refusing means they are more optimistic about their own valuation ceiling; accepting means they believe the bubble is peaking and it’s time to cash out.

---

References

• OpenAI Misses Its Own User and Sales Goals, WSJ Reports — Bloomberg
• OpenAI reportedly missed revenue targets. Shares of Oracle and these chip stocks are falling — CNBC
• OpenAI CFO reportedly at odds with Sam Altman over missed revenue target — Fortune
• Google to invest up to $40B in Anthropic in cash and compute — TechCrunch
• Google’s $40B Anthropic move is Big Tech’s latest huge AI bet — Axios
• Microsoft and OpenAI gut their exclusive deal — VentureBeat
• Anthropic shrugs off VC funding offers valuing it at $800B+ — TechCrunch
• Anthropic’s rise is giving some OpenAI investors second thoughts — TechCrunch
• Microsoft Loses OpenAI Exclusivity and AGI Clause in Amended Deal — Unite.AI
• Time Horizon 1.1 — METR

Further Reading

• AGI Observation This Week | The Week the “AGI Clause” Was Buried — Issue No. 1 Weekly Report

—Xiao Jian, April 29, 2026
This article tracks the real progress of AI / AGI / ASI daily. All data is from public sources.