OpenAI lifts its compute margin to 70% as it tries to reach profit

OpenAI’s compute margin is has hit 70%, according to The Information, which said the number rose from 52% at the end of 2024 and was roughly half of today’s level at the start of that same year.

The rally comes while the company still reports no profit, even after being valued at $500 billion in October. The business continues to hunt for ways to cover its large computing bill and build the infrastructure that supports its models.

The pressure on spending keeps getting louder as rivals push harder. Google’s Gemini model posted stronger benchmark results, and that result led Sam Altman, the company’s chief executive, to call a “code red” inside the organization.

Sam told teams to move their work toward ChatGPT upgrades. That move also delayed internal plans for an ad service. The company depends heavily on free use of ChatGPT, so it has been trying to grow paid business tools in industries such as banking and education, where it competes with Google and Anthropic.

Tracking higher margins and new revenue plans

The Information reported that OpenAI has better compute margins than Anthropic for paid accounts, but Anthropic spends less on servers.

That difference shows how each company handles its computing load as the cost of running large models grows. The company that pushed AI into the mainstream is still trying to reach profit while expanding the scale of its systems.

Sam shared several projections to justify the company’s spending plans.He said the company expects $20 billion in run-rate revenue by the end of this year and wants the number to grow to “hundreds of billion” by 2030.He shared the comments on social media, which he often uses to outline his view of the business.

Sam also listed new areas the company might enter next, like consumer devices, robotics work, and selling cloud compute space to other firms.

Responding to funding questions and bailout concerns

The company’s spending faced new questions when Sarah Friar, the chief financial officer, spoke at a tech event in Napa, California. The conversation shifted when Sarah used the word “backstop.”

Sarah said the company would use banks and private equity to support its trillion-dollar AI plan. Then she added that there might be a possible “governmental” step to “backstop the guarantee that allows the financing to happen.”

The comment drew strong reactions from people watching the industry. Some pointed to an interview from weeks earlier where Sam said the US government could serve as the “ultimate insurer” if AI was misused.

Sam later said that he was talking about harm caused by bad actors, not data center funding. The issue reached Washington, where White House AI and crypto chief David Sacks said, “There will be no federal bailout for AI.”