OpenAI Is Testing Features by Throwing ‘a Lot of Compute’ at Them

Sam Altman is bracing OpenAI customers for pricier, compute-heavy merchandise — and hinting at what’s subsequent.

The OpenAI CEO stated in a submit on X on Monday that the corporate is launching “new compute-intensive choices” over the following few weeks.

Altman stated due to the prices concerned, some options will initially be restricted to Professional subscribers, whereas sure new merchandise could have further charges.

Altman framed the push as an experiment in stretching AI infrastructure to its limits: “We additionally wish to study what’s attainable after we throw a whole lot of compute, at immediately’s mannequin prices, at attention-grabbing new concepts,” he wrote.

The announcement provides on to OpenAI’s balancing act: making superior AI accessible whereas protecting the steep costs of compute. Altman stated the corporate’s intention “stays to drive the price of intelligence down as aggressively as we will and make our companies extensively out there.”

“We’re assured we’ll get there over time,” he added.

Altman and OpenAI didn’t reply to a request for remark from Enterprise Insider.

The race for GPUs

OpenAI has been vocal about its insatiable demand for computing power.

“Each time we get extra GPUs, they instantly get used,” OpenAI’s chief product officer, Kevin Weil, stated on an episode of the “Moonshot” podcast printed final month.

Weil stated the necessity for compute is easy: “The extra GPUs we get, the extra AI we’ll all use.” He highlighted that including bandwidth made the explosion of video attainable.

Altman stated in July that the corporate will convey on greater than 1 million GPUs by the top of the 12 months.

“Very happy with the group however now they higher get to work determining easy methods to 100x that lol,” Altman wrote on X in July.

For comparability, Elon Musk‘s xAI disclosed that it used a supercluster of over 200,000 GPUs referred to as Colossus to assist practice Grok4.

Different tech giants have additionally been blunt about their urge for food for GPUs.

Mark Zuckerberg stated on an episode of the “Entry” podcast printed Thursday that Meta is making “compute per researcher” a aggressive benefit and is outspending rivals on GPUs and the customized infrastructure wanted to energy them.