OpenAI sells/rents AI compute to external customers by end-2026?

Question

Resolution Criteria

Resolves YES if by 11:59 PM PT on December 31, 2026, OpenAI offers AI compute services (training or inference) to external customers through ANY of the following:

Public cloud service offering:

Publicly accessible platform where customers can purchase/rent compute

Website or portal allowing sign-ups for compute access

Published pricing for compute hours/tokens

Enterprise/Partnership deals:

Announced deals to provide compute to named companies

SEC filings or earnings reports mentioning compute service revenue

Customer testimonials or case studies on OpenAI channels

Official confirmation via:

OpenAI website, blog, or press releases

Sam Altman on social media explicitly confirming external compute sales

Earnings call statements from OpenAI executives

Resolves NO if no external compute offerings exist by deadline, or if OpenAI only uses infrastructure internally for its own AI/products.

Important clarifications:

✅ DOES count: Any external customer (not OpenAI/ChatGPT/other OpenAI products) purchasing compute; compute for training, fine-tuning, or inference; beta programs with paying customers

❌ Does NOT count: Free research partnerships; internal use by OpenAI products; only licensing models without compute infrastructure; traditional software/API licensing without compute component

Background

OpenAI may, in the future, offer AI infrastructure services. In an interview on August 20, OpenAI's CFO Sarah Friar said that, while OpenAI is not "actively looking" at the prospect currently, "I do think about it as a business down the line, for sure." OpenAI is currently focused on securing enough compute capacity for its own AI operations, be that from other suppliers or through its $500 billion Stargate project.

OpenAI's leadership has previously projected spending of up to $1.4 trillion to build 30 gigawatts of computing resources in the coming years. By October 2025 Stargate is a sprawling, multi-continental initiative to build roughly 10 GW of AI data center capacity (mostly U.S.-based initially) underwritten by an unprecedented $500 billion investment. It is led by OpenAI with SoftBank and Oracle.

Considerations

OpenAI currently operates as a buyer of compute rather than a seller. AWS will provide OpenAI with immediate access to its large-scale cloud infrastructure, including clusters of NVIDIA's newest GB200 and GB300 processors. The deployment involves hundreds of thousands of chips and will scale to tens of millions of CPUs by 2026. OpenAI signed a $4 billion cloud deal with CoreWeave in May 2025, building on an earlier $11.9bn deal with the neocloud. The company has also signed deals with Google Cloud and was confirmed in July to be behind a $30 billion-a-year cloud deal with Oracle.

OpenAI's CFO noted the company is "trying to be thoughtful" about alternative financing models beyond debt, suggesting infrastructure monetization could be explored. However, no concrete plans for external compute sales have been announced, and the company's stated focus remains on securing capacity for its own operations through 2026.

[tweet]Update 2026-10-17 (PST) (AI summary of creator comment): - Guaranteed Capacity does not count, as reserved throughput for OpenAI's models is considered API usage rather than compute rental.

Resolving YES requires OpenAI to sell or rent compute itself (e.g., renting out GPUs or Stargate capacity to a third party).

Manifold Markets · Answer

Unlikely — Manifold Markets prediction market estimates a 8% chance (26 traders, as of Oct 19, 2026).