- OpenAI's 'Guaranteed Capacity' allows large enterprises to lock in up to three years of guaranteed compute
- In return, OpenAI gets predictable revenue to fund ongoing data center expansion
- Stargate also introduced new closed-loop water cooling system in major sustainability leap
ChatGPT maker OpenAI has announced a new ‘Guaranteed Capacity’ plan for enterprises who want to lock in and reserve long-term access to the company’s compute infrastructure.
This strategic shift comes as AI companies face increased pressure from rising enterprise AI adoption, which can lead to occasional downtime.
OpenAI recently revealed that around 40% of its $2 billion monthly revenue comes from enterprise customers – a figure it expects to continue to grow.
Guaranteed Capacity
The company recently revealed it had 900 million weekly ChatGPT users, but earlier WSJ reporting indicated that OpenAI failed to reach internal targets in 2025, including user numbers.
That article also revealed potential tension between CDO Sarah Friar, who advocates financial discipline, and CEO Sam Altman, who takes a ‘buy everything’ approach.
The Guaranteed Capacity program is designed specifically for large-scale applications and agentic automation that require continual and uninterrupted compute, rather than smaller companies that want to avoid the occasional outage.
Eligible companies can request upwards of 1B tokens per minute in terms of capacity, and pricing is tied directly to annual spending levels across one-, two- and three-year plans.
“Customers can draw down from this commitment across the portfolio of OpenAI products,” the company said.
Company CEO Sam Altman said in an X post that the update comes in response to customer feedback, and that it would be available “until we sell out of our current allocation for this program.”
We ‘will significantly expand our compute capacity’, OpenAI says
Altman also admitted that predictable revenue from large enterprise contracts and Guaranteed Capacity applicants would help the company to plan, giving it access to more consistent cash to build out future data centers.
“Our intention remains to build as much compute as fast as we can,” Altman added.
OpenAI’s headline-grabber is the Stargate project, which attracted collaboration from Microsoft, Oracle and others. When it was first announced in January 2025, the company targeted 10GW of US AI infrastructure by 2029 – a figure it had reached just over a year later. In April, the firm noted that it had also recently acquired a further 3GW of capacity.
With plans to “significantly expand [its] compute capacity in the years ahead,” OpenAI and partners are already scouting out future locations across the US, noting the progress data centers have made in terms of sustainability too.
Each building in its flagship Abilene, Texas campus requires the equivalent of two Olympic-sized swimming pools of water, but instead of traditional evaporative cooling towers, it uses a closed-loop system where water recirculates through sealed pipes.
“Annual water use for the entire cooling system at full buildout is expected to be comparable to a medium-sized office building, or about four average households,” the company declared.
By offering companies a way to secure up to three years of dedicated compute, OpenAI is ultimately building predictability into its revenue in order to sustain ongoing data center buildout and improve efficiency further.
Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds.