FULL STORY

OpenAI Launches B2B Marketplace Amid Strategy Questions

OpenAI launched a B2B marketplace with third-party inference providers like Baseten onboarded for open models in Codex. Analysts quickly questioned the strategic implications of not mandating OpenAI's own inference.

2026-09-30 ~ 2026-10-01 · 2 episodes · 12 posts

Episode 1 · Baseten Becomes First Open-Model Inference Provider in OpenAI's B2B Marketplace (2026-09-30, 4 posts)

OpenAI launched a B2B marketplace with Baseten as its first open-model inference provider, letting enterprises run open-weight models via Codex and the Responses API — a shift from 'use our models' to capturing enterprise AI budgets.

Episode 2 · Analyst: OpenAI Marketplace Won't Require Its Own Inference Infrastructure (2026-10-01, 8 posts)

On October 1, analyst Matt Slotnick published a thread analyzing OpenAI's marketplace announcement. His core finding: listing on the marketplace apparently does not require using OpenAI's inference infrastructure—the participant list even includes an open-source model provider—and monetization is only a small commission on transaction volume. Combined with open models being wired into the harness, he speculates OpenAI may be pivoting toward "aggregating intelligence demand." Details remain incomplete, so this is a preliminary read, but the strategic divergence already raises meta-strategy questions.

Confirmed

  • Per the announcement, marketplace listing does not require OpenAI's inference services; an open-source model provider is among the participants.
  • The marketplace takes only a small percentage commission on transactions, rather than monetizing underlying compute or storage.

Unconfirmed

  • Full details are not yet public; Slotnick repeatedly cautions his conclusions are provisional.
  • The "aggregating intelligence demand" framing is analyst speculation, not an official statement.

Why it matters

Slotnick contrasts this with hyperscaler marketplaces, where value comes from co-selling customers onto your own infrastructure—run elsewhere and you lose marketplace access and co-sell eligibility entirely. By not using the store as an infrastructure lock-in lever, OpenAI appears to forgo compelling inference consumption through the marketplace, instead keeping the ecosystem open with a low take rate. He argues that focusing on deployment and becoming an aggregator of intelligence demand is a sound strategy for most companies—more scaled platforms should do it—but it is strikingly unusual for OpenAI, which has publicly staked its identity on model capability. Worth watching are developer adoption and the platform's long-term monetization structure.