OpenAI's Astra Is Taking Anthropic's Enterprise Lead. Anthropic May Answer With a New Model

Anthropic, the company that has spent the past year positioning itself as the responsible leader in enterprise artificial intelligence, is now considering a new model launch to defend that position, Reuters reported on September 18. Three people familiar with the plans said the launch is being weighed as a response to the momentum OpenAI has built since it released GPT-6 Astra on September 3, and that Anthropic is evaluating the safety of its next model as part of the deliberations.

Related reading: OpenAI Revenue Below Expectations Sends AI-Linked Stocks Lower · OpenAI Safety Researcher David Robinson Resigns, Calling the Culture Broken.

Anthropic declined to comment. The company has not announced a model or a date, and the plans could change.

The enterprise spending numbers have moved

The shift that prompted the discussions is visible in third-party spending data. Astra now accounts for about 13% of enterprise AI spending tracked by the corporate expense platform Ramp, compared with roughly 8% for Anthropic's Claude Fable, according to the Reuters report. The gap matters because enterprise customers, not consumers, have been Anthropic's strongest franchise. For months the company has been described as the leader in business AI, a position built on coding tools, long-context document work and cautious deployment practices that appeal to regulated industries.

There is a second signal. OpenRouter, a widely used routing platform that channels developer traffic across models, said its users spent more on OpenAI models than on Anthropic models last week. According to Reuters, it was the first time OpenAI had led on that measure in more than two and a half years. Routing data reflects enthusiast and developer behavior more than procurement, but it moves earlier than enterprise contracts do, which is why investors watch it.

Anthropic's counter-move has to work on capability as much as messaging. Its business customers pay for coding agents, long-document analysis and the kind of controlled deployment that legal, financial and healthcare buyers require. OpenAI's Astra arrived with unusually broad ambitions: the company said at launch that the model could autonomously build websites, run scientific analysis and handle cybersecurity and coding tasks, and it demonstrated cutting an apartment hunt from six hours to under ten minutes. If Astra narrows the quality gap in the workloads enterprises actually run, Anthropic has to answer with a model of its own, which is the calculation the Reuters sources describe.

The safety-versus-speed bind

The timing is awkward in a specific way. On September 12, Anthropic chief executive Dario Amodei published a roughly 3,800-word essay arguing that the industry should slow the pace at which it improves model capabilities, warning about swarms of AI agents outpacing human control. The essay drew public support from OpenAI chief executive Sam Altman and SpaceX chief executive Elon Musk, a rare alignment across rivals. It also came after a summer in which AI agents from OpenAI, Anthropic and Meta were reported to have behaved unexpectedly during security testing.

A company that has just asked the industry to slow down is now weighing a faster release to avoid losing share. That tension is not new in the AI business, but it is unusually visible here. Reuters reported that the discussions include how to balance investment in releasing new models against the push to strengthen profitability, a calculus that Amodei's own essay acknowledged is difficult when rivals keep moving.

Anthropic is not the only lab whose public safety posture sits alongside a competitive imperative. OpenAI delayed parts of the Astra rollout over cybersecurity concerns before releasing it. What is different is that Anthropic made pacing-the-frontier its public identity, and the enterprise market is now testing whether that identity costs it revenue.

The IPO makes the pressure sharper

The model debate is unfolding inside an expected initial public offering. Reuters previously reported that Anthropic is in talks with Nvidia to become an anchor investor, considering as much as $10 billion in the offering. Anthropic is seeking to raise up to $100 billion at a valuation of around $2 trillion, which would make it the largest listing in history. The company raised $65 billion in May at a $965 billion post-money valuation and said its annualized revenue run rate passed $65 billion by the end of July, up from about $9 billion at the end of 2025.

The earlier plan called for marketing to begin in mid-October at the earliest, before the November US midterm elections. The new reporting says the timing may slip past the midterms, and other coverage has pointed to a November window that would let Anthropic report a strong third quarter first. Public-market investors have grown more focused on the timing of profits as interest rates have stayed higher, which makes every dollar of model development a line item that has to be justified against a near-term revenue curve.

A $2 trillion valuation implies a revenue trajectory that is still mostly ahead of the company. Reuters has reported that Anthropic's projections point to roughly $190 billion to $200 billion in revenue in 2028, up from a run rate above $65 billion now. That is a steep curve even by AI standards, and it depends on enterprises continuing to expand their spending rather than consolidating on a single vendor. It also raises the familiar circularity question: Nvidia backing a $100 billion IPO for a company that is one of its largest customers blurs the line between demand that exists and demand that is financed.

Why enterprise share is the moat

Anthropic's case to public investors rests on the idea that businesses will pay a premium for models that are reliable, auditable and safe to put in front of customers. That market has been slower to consolidate than consumer AI, and switching costs are real: enterprises build workflows, evaluations and internal tooling around a specific model.

Astra attacks that position directly. If OpenAI's latest model is good enough for enterprise work and cheaper to run, the safety premium becomes harder to charge. The Reuters report captures the investor anxiety precisely: some investors told the outlet they are re-evaluating whether Anthropic still deserves its status as the leader in enterprise AI.

Amazon and Google, both backers and compute suppliers, have their own incentive to keep Anthropic strong. Amazon has committed billions to Anthropic and expects the company to spend more than $100 billion on AWS over a decade. Google and Broadcom are adding TPU capacity. None of that guarantees model leadership, because the enterprise choice is made by buyers, not investors.

The enterprise market is still growing fast enough to support more than one winner, which is why the story is about share rather than survival. Businesses that were experimenting with a single model a year ago now route different tasks to different systems, and procurement teams add second and third vendors to avoid lock-in. That fragmentation cuts both ways for Anthropic. It makes the market harder to dominate, but it also means no single quarter of share loss is fatal. What Reuters describes is a change in trajectory, not a collapse.

The China factor and the cost of staying still

A second competitive pressure comes from open-weight models, especially Chinese ones, which are improving quickly and can be run far more cheaply than closed frontier systems. If enterprises can get adequate quality from a model they host themselves, the premium closed labs charge comes under pressure from below at the same time that OpenAI pushes from above.

That squeeze is the real reason a safety-first lab would consider moving faster. Slowing release schedules only works if customers wait. If they do not, the caution that built Anthropic's reputation becomes a market-share leak, and the company has to decide which it values more: the pace it argues for publicly, or the enterprise lead it built on that argument.

What to watch

The next several weeks will answer three questions. Whether Anthropic actually ships a new model before it lists, and at what price and capability. Whether Astra's enterprise share keeps rising or plateaus once the launch bump fades. And whether the listing lands before the midterms, because an IPO priced on enterprise leadership is a very different story if that leadership is being contested in the data.

For now, the company that told the industry to slow down is weighing whether it can afford to.