OpenAI CFO defends spending: Revenue triples to $20B alongside compute growth

Throughout 2025, OpenAI has been busy signing several AI infrastructure deals worth hundreds of billions of dollars with NVIDIA, Microsoft, AMD, Oracle, and others. While OpenAI has consistently said it is short on compute to serve its users, critics have argued that the company cannot generate enough revenue to justify the scale of investment. Today, OpenAI CFO Sarah Friar published a blog post responding to those criticisms.

Friar said OpenAI’s revenue has tripled over the past year and is growing in proportion to its AI infrastructure capacity. She noted that the figures below cover the past three years since ChatGPT launched. OpenAI’s compute grew 3x year over year, about 9.5x from 2023 to 2025.

  • Compute: 0.2 GW in 2023, 0.6 GW in 2024, and approximately 1.9 GW in 2025.
  • Revenue: Growing 3x year over year, about 10x from 2023 to 2025 ($2B ARR in 2023, $6B in 2024, and $20B+ in 2025).

Based on this trajectory, Friar argued that having more compute during these periods would have accelerated customer adoption and monetization.

In 2023 and 2024, OpenAI relied exclusively on Microsoft for its infrastructure needs. After revising its longstanding partnership with Microsoft, OpenAI is now working with multiple AI infrastructure providers to meet demand. Friar said this shift, from a single provider to several, has enabled OpenAI to plan, finance, and deploy capacity with greater confidence.

OpenAI is now training frontier models on the latest hardware when performance matters most, while serving high-volume workloads on lower-cost infrastructure when efficiency is the priority. She also pointed to OpenAI’s recent partnership with Cerebras to enable ultra-fast inference for its Codex coding models.

Friar further noted that OpenAI now operates with several revenue models:

  • ChatGPT subscriptions, including consumer and Team plans.
  • Ad- and commerce-supported revenue from ChatGPT free users.
  • OpenAI’s model APIs for developers.

Looking ahead, she said OpenAI expects additional economic models, including licensing, IP-based agreements, and outcome-based pricing. She ended the post with the following message:

"Infrastructure expands what we can deliver. Innovation expands what intelligence can do. Adoption expands who can use it. Revenue funds the next leap. This is how intelligence scales and becomes a foundation for the global economy."

Friar’s blog post aims to reassure existing and potential new investors that OpenAI’s capital-intensive strategy will deliver proportional financial returns rather than unchecked burn.

Report a problem with article
Next Article

Anthropic targets massive $350 billion valuation in new funding push, Sequoia latest to join

Previous Article

Linus Torvalds signals likely delay for final Linux 6.19 release