OpenAI introduces Guaranteed Capacity with 1-3 year token commits for reserved compute

• OpenAI's launch post introduced Guaranteed Capacity, a program that lets customers lock in long-term access to OpenAI compute for critical products, agents, and workflows.
• According to sama's thread, the commercial trade is simple: customers make 1 to 3 year commitments, and OpenAI offers discounted tokens plus capacity certainty.
• sk7037's post says the offering spans supported cloud providers and is aimed at eligible customers scaling production workloads, not general self-serve usage.
• OpenAI is framing the launch around scarcity: sama said the world will stay capacity constrained for some time, while gdb repeated that compute is becoming a hard limit as models get more useful.

You can read the official product page, skim OpenAI's announcement, and see Sam Altman's pricing and allocation details. The weirdly blunt part is that OpenAI is now selling reserved supply as a product, with a sales page that promises certainty for the workloads that matter most.

OpenAI's pitch is straightforward: guaranteed access to model compute, sold as a business offering rather than a best-effort API tier. The official page describes it as a way to guarantee long-term access for products, agents, and customer workflows.

The launch language is unusually explicit about why this exists. In OpenAI's post, the company says it has been investing in infrastructure, partnerships, and capacity planning so customers can scale reliably in a compute-constrained market.

The commercial terms surfaced first in sama's thread: OpenAI is offering discounted tokens in exchange for 1 to 3 year commitments. gdb's post repeated the same structure, adding that the deal is really about certainty on capacity availability.

That makes this less like a normal usage discount and more like reserved infrastructure. scaling01's screenshot and comment read it as a hedging instrument for businesses that do not want to live on shared capacity.

One detail that is easy to miss in the launch tweet is scope. sk7037's post says Guaranteed Capacity is for eligible customers and works across supported cloud providers, which puts it squarely in enterprise procurement territory rather than broad public access.

That also explains the landing page's CTA. The page linked in OpenAI's product page pushes buyers toward planning capacity and contacting sales, not flipping a dashboard toggle.

Altman added one more concrete constraint in the thread attached to his post: OpenAI will sell Guaranteed Capacity only until the current allocation for the program is gone. He also said the company will leave enough capacity for ChatGPT, Codex, and similar products, then offer the program again in the future as it builds more compute.