OpenAI has told its investors that its initial public offering could be delayed until next year. The decision, communicated in San Francisco, coincided with the entry into force of the first phase of the EU’s Artificial Intelligence Act in Brussels and opened a crack in the hierarchy of generative AI. This was not a mere technical postponement: for the first time, capital—that relentless judge that measures everything in returns—treated the creator of ChatGPT as an asset in relative decline. The irony could not have been more cutting. OpenAI, which had planned to beat rival Anthropic to the public markets, now watches as its competitor accelerates its own public offering for the fall, with a higher valuation and revenue growth that has outpaced its own. The king has fallen, and capital has already chosen its successor.
OpenAI delays its IPO as cash burn sets off alarm bells
Concern over the speed of cash burn relative to the company’s actual growth has been voiced privately by major OpenAI investors, according to information published by Chinese outlet 36Kr. The problem is not just that OpenAI spends heavily; it is that it spends heavily and no longer grows at the pace that once justified that burn rate. The equation that for years sustained its astronomical valuation—exponential growth in exchange for multibillion-dollar losses—has stopped adding up in the eyes of the most demanding funds.
The consequence has been immediate and predictable: some of those same investors have started hedging their bets by directly funding Anthropic, the company founded by former OpenAI executives that has become its nemesis. This is not a knee-jerk reaction or an emotional decision; it is a cold reading of the market. Anthropic has surpassed OpenAI in both revenue growth and valuation, according to data compiled by 36Kr. And while OpenAI postpones its plans, Anthropic is already meeting with potential investors for its fall IPO, highlighting in those presentations its competitive edge over the maker of ChatGPT.
Distrust has even spread to former OpenAI employees. Shortly after leaving the company, one of them issued a public warning reported by Chinese outlet QbitAI: he advised those holding stakes to cash out as soon as possible and not to wait for the IPO. The message, shared across social networks and industry forums, reflects growing pessimism about the valuation of frontier labs. This is not an isolated opinion; it is a symptom of a regime change in risk perception.
The price war that changed the rules
To understand this reversal of the narrative, one must look east. Competitive pressure from Chinese labs has completely reshaped the landscape. According to the South China Morning Post, OpenAI has sharply cut prices on some of its models to fend off its Asian rivals. The most striking case is GPT-5.6 Luna, whose entry-level price has fallen to a level that would have been unthinkable just two years ago.
OpenAI no longer sets the rules of the market; it suffers them. The company that for years set the pace of innovation and pricing now reacts to the moves of Qwen, DeepSeek, Kimi, GLM, Hunyuan, and Ernie—the Chinese labs that have proven it is possible to compete with efficient, low-cost models. Efficiency has become the new battleground, and there OpenAI starts at a disadvantage: its architecture, historically more expensive to run, leaves it vulnerable in an environment where profit margins per token shrink every week.
The irony is that OpenAI’s own strategy—scaling ever-larger models—has led it into a position of structural weakness. While Chinese labs optimize their architectures to cut costs, OpenAI keeps burning cash in a race for scale whose diminishing returns are becoming evident. Investors have read that dynamic and adjusted their portfolios accordingly.
The European regulation as a backdrop
The entry into force of the first phase of the EU’s AI Act adds another layer of regulatory pressure on OpenAI in one of its key markets. The European rule, the most ambitious of its kind in the world, imposes transparency, documentation, and risk-assessment obligations that raise compliance costs and slow deployment cycles.
For a company that needs to demonstrate accelerated growth to justify its valuation, European regulation is an added drag. It is not that Anthropic is exempt—the same obligations apply to it—but its more contained cost structure and stronger growth allow it to absorb that impact better. The paradox is that the regulation, designed to protect European citizens, is helping reshape the balance of power between U.S. labs: it accelerates the leader’s fall and favors the challenger.
The new geopolitics of capital
What we are witnessing is not just a corporate battle; it is a reordering of the geopolitics of artificial intelligence. OpenAI’s dependence on private capital and its inability to control cash burn make it vulnerable against rivals with more efficient models, with consequences that extend beyond Silicon Valley’s borders.
Investors’ decision to diversify toward Anthropic is a sign that the market has understood something fundamental: in AI, power is no longer measured by the ability to scale models, but by the ability to scale revenue sustainably. OpenAI built its empire on the promise of artificial general intelligence; Anthropic builds its own on the reality of return on investment. That difference—subtle but decisive—explains why capital has switched sides.
Anthropic’s accelerated IPO for the fall is not an opportunistic move; it is a statement of intent. The company wants to capture the moment when the sector’s narrative has shifted, before OpenAI can regroup. If Anthropic secures a solid public valuation while OpenAI continues to wait, the message to the market will be unmistakable: the future of generative AI no longer has a single owner.
An irreversible regime change
The question that hangs in the air is whether OpenAI can reverse this dynamic. The company retains real advantages: its brand remains the most recognized in the sector, its technology maintains a competitive level, and its ability to attract talent has not disappeared. But markets do not reward potential; they reward execution. And on that front, Anthropic has taken the lead.
The delay of OpenAI’s IPO until next year, if confirmed, is not a simple scheduling adjustment. It is the implicit admission that the company needs time to shore up its finances, restructure its pricing strategy, and prove to investors that it can grow without burning cash at an unsustainable pace. It is, in short, an acceptance that the cycle of unlimited growth financed by private capital has come to an end.
Generative AI has entered its maturity phase, and maturity has its own rules. The era of unquestioned leaders has given way to a competitive market where efficiency, profitability, and regulatory adaptability matter as much as technological innovation. OpenAI, the fallen king, will have to learn to govern in a world where it no longer dictates the laws. All of us, spectators of this transition, are witnessing the construction of a new order whose outcome is far from written. The only certainty is that capital has already delivered its verdict.