OpenAI assembles ~26 gigawatts of compute across NVIDIA, AMD, and Broadcom
Beyond its own Jalapeño chip, OpenAI has stacked supply: NVIDIA investing up to $100 billion for 10 GW of infrastructure, AMD providing 6 GW, and the Broadcom custom accelerators adding 10 GW. The aggregate — roughly 26 gigawatts — is a multi-vendor compute strategy at a scale that tests the limits of AI funding itself.
The multi-vendor spread is deliberate de-risking. Committing to NVIDIA, AMD, and its own Broadcom silicon at once means OpenAI is not hostage to any single supplier's capacity, pricing, or roadmap. Diversifying compute the way a manufacturer diversifies its parts suppliers is what a company does when compute has become its most critical and most constrained input.
The 26-gigawatt aggregate is a staggering number in energy terms — comparable to the draw of tens of millions of homes — and it reframes OpenAI's constraint as power and land, not chips. Securing that many gigawatts is less a procurement problem than an infrastructure build spanning years, which is why the deals stretch to 2029 and involve energy as much as silicon.
The financial read is that these commitments test whether AI funding can underwrite infrastructure at this scale. A $100 billion NVIDIA investment plus multi-gigawatt AMD and Broadcom deals is capital allocation on the order of national infrastructure, and it concentrates an enormous bet on OpenAI's future revenue justifying the buildout. The compute is real; whether the economics close is the open question.
Data Center Dynamics — OpenAI partners with Broadcom for custom AI accelerators and Ethernet → · Memeburn — Nvidia OpenAI data center deal tests AI funding 2026 →