Sam Altman says OpenAI is preparing to test a new phase for ChatGPT, one that leans heavily on compute and stretches the economics of today’s large models.
In a new post on X, the OpenAI CEO says users will get to access ChatGPT’s new offerings soon.
“Over the next few weeks, we are launching some new compute-intensive offerings. Because of the associated costs, some features will initially only be available to Pro subscribers, and some new products will have additional fees.”
Altman also says the rollout comes as OpenAI experiments with making tools widely available while grappling with the expense of running large models.
“Our intention remains to drive the cost of intelligence down as aggressively as we can and make our services widely available, and we are confident we will get there over time.
But we also want to learn what’s possible when we throw a lot of compute, at today’s model costs, at interesting new ideas.”
Altman’s concerns about expenses highlight OpenAI’s financial commitments in the coming years, with current revenue making up just a fraction of the obligations.
The startup has committed to $328 billion in spending — $300 billion to Oracle for compute, $18 billion for a data-center venture, and $10 billion for Broadcom chips — while pulling in just $13 billion a year.
Losses already run into the billions each year, and chief executive Sam Altman has told investors they will stack to $44 billion by 2029, the first year OpenAI is projected to turn a profit.