As company releases its latest generative AI model
OpenAI has shared some details about its growing compute infrastructure alongside the launch of its latest model, GPT-5.
Compute and infrastructure staffer Anuj Saharan said that the company has increased its compute 15-fold since 2024.
On LinkedIn, he added: "[In the] last 60 days: built out 60+ clusters, a backbone moving more traffic than entire continents, and 200k+ GPUs to launch GPT-5 to 700m people - all while designing the next 4.5GW toward superintelligence."
Saharan then pointed to roles at OpenAI's Stargate data center venture: "We're a small team building planetary scale Al infrastructure at unprecedented pace; hiring across data, energy, data centers, capacity planning, biz dev, finance and more. Join us for the next 100x scale-up."
The company is hiring aggressively for the Stargate team, while primarily relying on Oracle for compute.
Last month, the two companies confirmed a 4.5GW deal for Stargate data center capacity in the US, expected to cost some $30 billion a year.
OpenAI is still planning to self-build its own data centers, the company's director of physical infrastructure told DCD in the latest issue of the magazine.
It also plans to develop a Stargate data center campus in the United Arab Emirates with Oracle, Nvidia, Cisco, SoftBank, and G42 involved. Over in Europe, it is working with Nscale on a data center in Norway.
It is searching for sites for other projects across the world.
