OpenAI missed ChatGPT targets; $600B compute bills loom

OpenAI failed to reach internal targets for ChatGPT weekly active users and annual revenue, and Chief Financial Officer Sarah Friar warned company leaders that roughly $600 billion in precommitted data‑center contracts could exceed revenue and strain cash flow. Company insiders say OpenAI set an internal goal of one billion weekly active users by the end of last year; the company did not meet that milestone and did not announce it.

Friar privately raised concerns that the company’s multi‑year compute commitments, secured through large long‑term deals, may be unsustainable if subscription and customer revenue do not accelerate. Board members have increased scrutiny of the data‑center agreements and questioned why the company is seeking additional capacity amid slower growth.

OpenAI accumulated about $600 billion in future data‑center spending after several years of locking in capacity under the view that compute would be the main constraint on AI progress. Company executives say the contracts are long term and cover significant portions of OpenAI’s compute needs. Friar warned that payments tied to those deals could outpace the cash coming in, affecting near‑term liquidity for cloud and hardware bills and complicating longer‑term financial planning.

The fundraising and valuation backdrop for large‑language models has shifted. Recent secondary trading placed a rival AI developer at an implied valuation of roughly $1 trillion, above OpenAI’s implied valuation near $880 billion. Public prediction markets assign about a 64% probability that that rival will complete an initial public offering before OpenAI. OpenAI’s chief executive has pushed for a public listing this year, while Friar has cautioned that internal reporting controls are not yet ready for public‑market requirements.

Industry estimates show that by the end of 2025 about 84% of the world’s working‑age population had not used generative AI tools, and roughly 44.8 million people held paying AI subscriptions worldwide. These adoption figures help explain why revenue growth can lag technological capability even as firms commit to large compute contracts.

OpenAI is pursuing hardware initiatives while managing compute costs. An analyst reported that OpenAI is co‑developing a smartphone chip with Qualcomm and MediaTek, with Luxshare handling system co‑design and manufacturing and potential mass production targeted for 2028. The companies involved have not confirmed the plans.

Outside the finance and operations discussion, OpenAI apologized to a community in British Columbia after acknowledging it had banned a ChatGPT account tied to a suspect in a February mass shooting without notifying law enforcement.

Investment professionals offered varied perspectives on the company’s situation. Alice Li, an investment partner at Foresight Ventures, described the gap between expectations and outcomes as a recalibration rather than a systemic collapse and noted AI still requires human judgment and context for many tasks. Markus Levin, co‑founder of a distributed infrastructure network, argued that slow and uneven adoption of generative AI reflects early penetration rather than a broad market failure. Pavel Bezhin, a finance chief at an AI developer, compared the pattern to prior technology cycles in which breakthroughs were followed by market corrections.

Company leaders face competing demands: secure sufficient compute to train and run models, manage rising costs and strengthen finance and reporting systems if a public offering moves forward. The company has not provided public comment on the internal targets or the scale of its compute commitments.