The artificial intelligence startup forecast that its compute costs will total $1.4 trillion over the next eight years, the analysts said, according to the report.
“We update our OpenAI forecasts with our new compute capacity and rental cost schedule and conclude it would need [$207 billion] of new financing by 2030,” the analysts said, per the report. “One unknown parameter is the flexibility that OpenAI may have to adjust its commitment [versus] effective demand or financial capacity. Capital injections, debt issuance or higher revenue than in our model would help [close] the funding gap.”
OpenAI CEO Sam Altman said in an Oct. 29 post on social platform X that the company would like to build an AI factory to make 1 gigawatt of compute per week of new capacity.
The company would like to do so “at a greatly reduced cost relative to today,” Altman said.
Currently, OpenAI has committed to about 30 gigawatts of compute with a total cost of ownership at about $1.4 trillion, Altman said.
Advertisement: Scroll to Continue
“We are comfortable with this, given what we see on the horizon for model capability growth and revenue growth,” Altman said in the post.
PYMNTS reported Oct. 27 that investors were continuing to pour capital into AI startups and funding the companies that build the infrastructure supporting the creation and deployment of advanced models and the applications powered by those models.
On Nov. 18, Microsoft said it was joining Nvidia in a multibillion-dollar partnership with AI startup Anthropic.
Under the collaboration, Anthropic will scale its Claude AI model on Nvidia-powered Microsoft Azure. The AI startup has agreed to purchase $30 billion of Azure compute capacity and to contract additional compute capacity up to 1 gigawatt, while Microsoft and Nvidia have pledged to invest up to $5 billion and $10 billion, respectively, in Anthropic.
On Nov. 9, Meta said it will invest $600 billion in the United States by 2028 to build AI data centers.
For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.
