Anthropic is deliberating whether to accelerate the release of a new frontier AI model ahead of its planned initial public offering, as OpenAI's GPT-6 Astra opens its first material lead over Claude Fable in enterprise spending in more than two and a half years, according to a Reuters report published September 18.
The internal debate lands at a delicate moment: the company is preparing a stock market debut that could define the next phase of the AI race, and investors are pressing for a demonstrable answer to Astra before the road show begins. It is the sharpest competitive test yet for the safety-first lab, and the latest sign that the AI industry's business battle is intensifying — you can follow the latest AI developments as the IPO window approaches.
The Enterprise Spending Shift
The numbers behind the deliberation come from corporate expense data. OpenAI's GPT-6 Astra now accounts for about 13% of enterprise AI spending tracked by the corporate expense platform Ramp, versus roughly 8% for Anthropic's Claude Fable, Reuters reported.
It is also the first time in over 30 months that OpenAI has led Anthropic on developer spend via OpenRouter, a routing service widely used as a proxy for which models developers actually pay to use. The shift matters because enterprise AI economics have become the primary story investors are underwriting for both companies heading into Anthropic's listing.
Adding to the pressure, Meta is separately reported to be actively looking to reduce its reliance on Anthropic-hosted models inside its own products — a quiet but potentially costly erosion of Anthropic's distribution footprint among the hyperscalers.
Revenue Still Favors Anthropic — for Now
The paradox is that Anthropic's financial trajectory remains the stronger of the two. The company ended July at more than a $65 billion annualized revenue run rate, versus roughly $40 billion for OpenAI, according to figures previously reported by both companies and cited in the Reuters coverage.
Anthropic's full-year 2025 revenue came in around $9 billion, and internal projections cited by Reuters point to $190 billion to $200 billion by 2028. But that trajectory depends on holding enterprise mindshare — hence the debate over whether to ship a new model now rather than later.
Investors are not uniformly convinced the growth can hold. The Financial Times reported that investors warn Anthropic could struggle to sustain revenues after its IPO, a concern that raises the stakes of the product roadmap: a strong new model launch would set the messaging for the road show, while a misstep could give the skeptics their opening.
Safety Pacing Meets Ship-It Pressure
The deliberation cuts against the public positioning of Anthropic's own leadership. CEO Dario Amodei has argued that "we must slow the pace at which we improve the capabilities of AI models," and Reuters reports the company is running additional safety evaluations on the candidate model before any release decision.
That caution now collides directly with shareholders pushing for a competitive answer to Astra ahead of the offering. Anthropic is also balancing incremental profitability against continued spending on frontier training runs — the same tension every frontier lab faces, but with a public listing on the calendar, it gets resolved under investor scrutiny rather than in private.
The episode is notable for what it says about the industry's center of gravity. Safety evaluations, once a differentiator announced on the lab's own schedule, are becoming a variable negotiated against market timelines.
The listing itself would be a landmark for the sector. OpenAI has explicitly ruled out going public in 2026, with Sam Altman calling an IPO on that timeline ill-advised, which would make Anthropic the first frontier lab to test public-market appetite for frontier-model economics. How investors value its compute-heavy growth, its safety overhead and its concentration of enterprise customers will set reference points for every private AI company waiting behind it.
What to Watch Into the IPO Window
According to Reuters, the next 60 days are pivotal. An IPO road show is expected to begin in mid-October at the earliest, but could be deferred until after the November U.S. midterm elections.
A slip past November would let OpenAI extend the Astra window uncontested — but it would also give Anthropic time to demonstrate that revenue continues to climb even without a new top-line model. Either way, the decision on the candidate model will likely land before investors do, and it will frame how the market prices the first pure-play frontier lab IPO.
For enterprise buyers, the near-term effect may be simple: sharper pricing, faster release cadences, and more aggressive enterprise commitments from both labs as each fights for the expense-line share that now defines the race.
Stay Ahead of AI
Following the business of AI — IPOs, enterprise shifts and the model race? Read more AI news on AI Buzz Wire — independent AI coverage, every day.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
