OpenAI, the creator of ChatGPT and a global leader in AI applications, is in early-stage discussions with prospective investors about a new funding round that would place its valuation above $1.2 trillion, according to reports citing people familiar with the matter.

These discussions were initiated by institutional investors, with the decision to proceed hinging on the company's timeline for going public. OpenAI Chief Executive Sam Altman has indicated that an IPO is unlikely before 2027, ahead of full public market pricing. Investors see significant upside in acquiring more OpenAI equity now, betting that advanced models will expand into broader work scopes and boost subscription, enterprise, and API revenue potential. The sustainability of this valuation will ultimately depend on how quickly that demand translates into revenue and cash flow.

The potential OpenAI round highlights a broader scramble for stakes in leading model platforms, alongside moves from rivals Anthropic and DeepSeek. Anthropic completed a $65 billion equity raise in May, reaching a post-money valuation of $965 billion. Reports from September 11 suggest it may prepare a potential IPO at roughly a $2 trillion valuation, aiming to raise up to $100 billion—a sum that would surpass the record $86.3 billion SpaceX raised in June. That reported figure, however, represents a potential valuation target rather than a realized post-funding market cap. Meanwhile, DeepSeek, China's leading AI model developer, is reportedly negotiating a funding round at a pre-money valuation of around $71 billion, up from approximately $52 billion in its first round. Across different markets, capital flows continue to target AI application growth, though the final terms of any deal remain subject to execution.

Where the money is going

Reports indicate the talks involve a possible funding round for OpenAI that would surpass the $1.2 trillion mark ahead of its eventual market debut. One person familiar with the situation, who asked not to be identified, said the decision to move forward will depend on when OpenAI decides to list. The talks were initiated by investors. Such a round would pave the way for a long-anticipated IPO. Altman has told Fortune that while an IPO remains in the pipeline, it is not happening this year.

If completed at this valuation, OpenAI would leapfrog its main competitor, Anthropic, which reached a $965 billion valuation including new investments in May. The Financial Times first reported the discussions on Tuesday, noting the round would let long-standing OpenAI backers expand their exposure before the company goes public. Anthropic is preparing its own IPO and has selected Nasdaq as its listing venue. According to Bloomberg, the Claude chatbot maker aims to raise as much as or more than SpaceX did in June, which set a record with $86.3 billion raised. Altman recently told Fortune that an IPO for OpenAI could come as late as 2027.

Investor enthusiasm for OpenAI and Anthropic hinges on turning model capability into repeatable professional work. Pretraining builds foundational ability, while reinforcement learning and task evaluation improve reasoning and execution; tool use, context management, and result verification enable longer workflows. OpenAI's new Astra model, which has stoked AGI debates, offers a concrete example. In delayed-simulation testing on OSWorld 2.0, Astra scored 72.6% across tasks, averaging about 40 minutes per task, compared with GPT-5.6 Sol's 65.7% and roughly 75 minutes. That marks an improvement in both completion and time efficiency under specific conditions. Astra also posted nuclear-level results on other benchmarks, including 98% on FrontierMath Level 4 and 99.9% on ARC-AGI-3, showing major leaps on certain difficult tests. Nvidia CEO Jensen Huang went as far as to state on social media that the arrival of GPT-6 Astra signals the "AGI era has begun."

Anthropic has also disclosed research into multi-agent systems, demonstrating the engineering value of planning, parallel retrieval, and multi-step tool calling. The logic is that if these advances consistently reduce rework and human intervention in customer workflows, enterprises will have stronger reasons to include AI in their routine budgets. That expands the commercial space for model companies into high-value tasks such as software development, research, and specialized services.

The sudden pause on Pro-tier subscriptions triggered by Astra underscores the extraordinary demand surge and the compute constraints that come with it. Reports say OpenAI's product lead, Tibo, described demand for Astra as "unprecedented," with the Pro tier placing the most strain on the system. Starting September 10, the company halted new sign-ups and upgrades for the $200-a-month Pro 20x package, while existing subscriptions continue to renew. Technically, complex agent tasks require repeated context reads, tool calls, plan generation, and result validation: input processing adds to computational load, long sessions and high concurrency increase KV cache usage, and generation stages can be limited by memory bandwidth. Expanding compute capacity and improving scheduling are therefore directly tied to how many paid requests the platform can handle and at what cost.

The demand pressure from Astra strengthens OpenAI's motivation to expand capacity. Supporting a trillion-dollar valuation, however, depends on raising delivery efficiency for successful tasks and turning subscription, API, and enterprise revenue growth into sustainable profits and cash flow.

Investor focus shifts from hype to payoff

As the "AI slowdown" narrative spreads and the 10-year Treasury yield climbs, equity markets are increasingly testing how quickly AI-driven earnings growth can be delivered. The divergence is becoming clearer: capital continues to chase long-term opportunities in top-tier model platforms, while public markets are demanding faster AI-related revenue realization, especially for software companies.

A group of AI leaders, including Anthropic and OpenAI, over the weekend jointly called for slowing the pace of frontier model development to allow safety measures to catch up. That call, combined with rising long-term Treasury yields, is raising the bar for valuation in tech stocks. On September 12, Anthropic CEO Dario Amodei urged the industry to slow the advancement of frontier capabilities, a view supported by Altman and others. On the first trading day after Amodei's call, September 14, Nvidia shares fell roughly 3.4%, while the Philadelphia Semiconductor Index dropped around 6%, a rare and sharp decline. Markets are beginning to factor in the risks associated with AI deceleration discussions, forcing a reassessment of mega-scale training investments, model release cadence, and future compute procurement growth.

On September 14, the Philadelphia Semiconductor Index fell about 6%. Korea's KOSPI, a global bellwether for AI compute investment, dropped over 3% that Monday, and fell another 0.85% on September 15 to close at 6,627.26, marking its fourth consecutive session of losses. On September 15, the 10-year Treasury yield broke above 5% again, hitting levels not seen since 2007. It spiked intraday to 5.012% on September 14—the highest since 2007—before pulling back to 4.960%. By the close of U.S. trading on September 15, the yield had settled above 5%, near 5.04%, its highest since 2007. As the anchor for global asset pricing, the 10-year Treasury yield influences corporate financing costs and the discount rate used to convert future profits into present value. Rising energy prices and inflationary pressure are pushing rates higher, forcing tech stocks to contend with both higher capital costs and adjusted long-term growth expectations.

With long-term Treasury yields near multi-decade highs and the "AI slowdown" narrative weighing on expectations for compute investment expansion and chip earnings growth, the surge in AI application usage driven by Astra and its user-side compute demand offers a positive signal for AI commercialization. Against this backdrop, OpenAI's ongoing talks to raise capital at a valuation above $1.2 trillion can be read as investors still seeing substantial room for advanced models to penetrate subscriptions, enterprise services, and specialized professional workflows. The central question in capital markets now is how quickly AI commercialization can convert into revenue—and whether that revenue can sustain the ever-expanding compute investment.