OpenAI reveals its data center capacity tripled to 1.9GW in 2025 – News

OpenAI Group PBC on Sunday shared new information about its financial performance and data center construction efforts.

Chief Financial Officer Sarah Friar disclosed in a blog post that its annualized recurring revenue topped $20 billion last year. That’s up from $6 billion in 2024 and $2 billion the year before. In the same time frame, OpenAI grew its data center footprint from 200 megawatts to about 1.9 gigawatts.

That the company’s annualized revenue and computing capacity both grew about tenfold in three years is no coincidence. In today’s blog post, Friar disclosed that OpenAI ties data center investments to growth milestones. “Capital is committed in tranches against real demand signals,” she explained. “That lets us lean forward when growth is there without locking in more of the future than the market has earned.”

According to the executive, OpenAI is working to not only expand its data center infrastructure but also make it more cost-efficient. Friar divulged that the company has brought down its inference expenses to under $1 per million tokens. OpenAI achieved that partly by mixing and matching different types of data center hardware.

“We train frontier models on premium hardware when capability matters most,” Friar wrote. “We serve high-volume workloads on lower-cost infrastructure when efficiency matters more than raw scale.”

Friar didn’t specify what chips power OpenAI’s lower-cost infrastructure. It’s possible the hardware uses the same pricey, top-of-the-line graphics processing units as the company’s most advanced training clusters. The most advanced GPUs are often the most cost-efficient. Nvidia’s flagship Rubin graphics card, for example, can run some inference workloads at 1/10th the cost per token of its predecessor.

Finding ways to lower hardware expenses is likely to become an even bigger priority for OpenAI going forward. In September, sources told The Information that the company was on track to end 2025 with a $8 billion loss, $1.5 billion more than originally expected. OpenAI’s loss is reportedly set to more than double to $17 billion this year.

The sources stated that the company’s efforts to develop custom chips and data centers are part of its effort to cut infrastructure costs. Last year, OpenAI inked a $10 billion partnership with Broadcom Inc. to co-design AI accelerators. Separately, it’s working with SoftBank Group Corp.’s SB Energy business to build Stargate data centers based on a custom design.

Friar’s blog post contained hints about OpenAI’s long-term revenue growth plans. The executive detailed that she expects new monetization models to emerge in the artificial intelligence market. “Licensing, IP-based agreements and outcome-based pricing will share in the value created,” Friar wrote.

Ads are another component of OpenAI’s growth strategy. On Friday, the AI provider announced plans to display paid promotions below ChatGPT prompt responses. The company will test its advertising system with a limited number of users in the U.S. before rolling it out more broadly. 

Friar indicated that OpenAI’s development roadmap also prioritizes AI agents and other workflow automation tools. According to the executive, the company is focused on helping users automate tasks that span multiple applications. Another priority is equipping models with the ability to use context for extended periods of time.

Photo: Focal Foto/Flickr