OpenAI, Google, and Anthropic Are in a Race Nobody Can Win — or Afford to Lose

The AI industry is currently experiencing a unique tension that is almost like panic. All three of the businesses that the majority of people believe will define the future of artificial intelligence are burning capital at rates that would sink practically any other industry. Massive sums of money are moving, and leadership teams are changing. The odd thing is that everyone seems to be aware of this. Nevertheless, spending continues to rise.

Through 2028, OpenAI anticipates cumulative operating losses of $74 billion. The $200 monthly ChatGPT Pro subscription loses money on its most frequent users, as Sam Altman has freely admitted. This means that the product at the core of OpenAI’s consumer strategy is a liability at scale. With Sequoia Capital joining despite already owning shares in OpenAI and xAI, Anthropic is on its way to a $350 billion valuation following a $10 billion funding round. This is a subtle indication that even Silicon Valley’s most disciplined venture capital is hedging its bets rather than picking a winner. Even with all of its advantages in terms of infrastructure and search power, Google is still struggling to be relevant in a field that it arguably ought to have dominated from the beginning.

Even for those who appear most uncomfortable with the competitive logic, it is nearly impossible to avoid it, which is what makes this race so peculiar. The CEO of Anthropic, Dario Amodei, has used language that sounds more like a public health alert than a product roadmap when discussing AI development. It’s clear from watching him in interviews that he genuinely thinks the thing he’s building could be dangerous, but he’s still going ahead with it, in part because it makes sense that if someone is going to build something, it should be someone who is concerned about the danger. It doesn’t sound like a comfortable position to be in.

However, the financial picture is clearer than the philosophical one. After overtaking OpenAI’s prior dominance in just eighteen months, Anthropic now holds about thirty-two percent of the enterprise LLM API market. Business clients account for 80% of Anthropic’s revenue, which is a structural advantage when attempting to create a business that can withstand the associated computing costs. The math is beginning to show that consumer AI is a more difficult place to make money when inference costs stay the same, and OpenAI’s revenue split still heavily favors consumers. Claude is available to Azure users thanks to a $500 million investment from Microsoft. It’s not a minor detail that OpenAI’s main distribution partner is hedging with its main rival at the same time.

It is more difficult to read Google’s stance. DeepMind’s research depth, Google Cloud’s infrastructure scale, and integration routes through Search and Android that no startup could match gave the company real early advantages. However, Google was forced to take a defensive stance due to the AI moment because they were concerned about how a chatbot might affect search revenue, which took time. There is a sense that Google has been catching up to a story it should have written, even though Gemini is now competitive and the Pixel and Android integrations are genuine. That might change. Google has previously survived threats from competitors. However, the time when it felt safe to be the obvious default in AI has obviously passed.

What the competitive game demands of all three is the deeper issue. Concentration is attracted by the physics of frontier AI training: larger data centers, more specialized chips, and exclusive energy agreements. That can only be funded by a small number of organizations. The political and regulatory landscape pulls in the exact opposite direction, requiring localization, value alignment, and constraints that are difficult to reconcile with the fast-paced logic of catching rivals. OpenAI is attempting to influence regulations in its favor while competing on capability. Anthropic is competing commercially while maintaining a safety-first stance, which is a very challenging needle to thread. Google is attempting to avoid cannibalizing the biggest user relationship in tech history.

It’s difficult to ignore the fact that all three businesses are acting in this manner out of fear of what might happen if they slow down while someone else doesn’t, rather than because they are certain it’s the best course of action. That is not a business plan, but the structure of a prisoner’s dilemma. Everyone benefits from the logical cooperative result, which is slower, more coordinated development with shared safety research. However, since each player gains from moving more quickly than the others, everyone moves quickly, which worsens the overall result. Engineers can model this pattern, and economists can identify it. That doesn’t make stopping any simpler.

It’s still genuinely unclear what will happen in the end. One of these businesses might emerge with a long-lasting technical advantage. In an AI economy where the true value is realized downstream, all three might wind up serving as infrastructure providers. It’s possible that in hindsight, the entire contest will resemble the internet browser wars, which were costly, intense, and ultimately less decisive than they appeared to be at the time. No one is aware. And that’s what makes the entire situation so difficult to ignore, more than the billions being spent.