Broadcom reportedly won’t build OpenAI’s custom chip unless Microsoft buys 40 percent of them

What happened

Broadcom has paused building OpenAI’s custom AI chips because it demands that Microsoft commit to buying 40 percent of the total production before moving forward. Microsoft has not agreed to this deal yet. OpenAI’s internal team flags this arrangement as financially unattractive, noting that just the first phase of the chip’s production could cost about 18 billion dollars.

Why it matters

This deal stalls OpenAI’s efforts to control its AI hardware stack directly while making Microsoft the potential gatekeeper for chip volume and revenue. If Microsoft says no, or negotiations drag, Broadcom won’t start production, slowing OpenAI’s plans to optimize performance and cost with custom silicon. The financing structure shifts risk heavily toward Microsoft, exposing how expensive and challenging it remains to develop specialized AI chips independently. It also highlights how expensive building AI infrastructure can slow project timelines and reshape partnerships in the AI supply chain.

What changes in practice

Builders and founders targeting custom AI hardware face a clear signal: independent chip production requires massive upfront commitments or strong guarantees from major buyers. Startups and large companies alike must factor in higher capital needs and conditional vendor dependencies during chip development phases. For founders, this raises the bar for funding chip projects and increases reliance on deep-pocketed partners to secure volume deals upfront. Microsoft’s hesitation means OpenAI may have to delay hardware control ambitions, sticking longer to existing cloud providers and off-the-shelf chips, which raises ongoing operational costs.

Buyers like Microsoft gain leverage over deal terms and chip supply while smaller buyers or those without guaranteed volumes will struggle to lock early access or favorable pricing on customized AI chips. Investors watching AI hardware startups should expect cash burn targets to accelerate as producing and scaling chips demands billions before volume discounts. Security and risk teams within AI firms should note the vendor risk steepens with such large undisbursed commitments needed to greenlight production, posing exposure to supplier hold-ups or strategic shifts.

Who should pay attention

AI developers and companies building custom chips must watch this closely as it lays bare the enormous upfront capital and buyer assurance needed. Cloud providers and large AI infrastructure buyers like Microsoft are also in focus, since their purchasing decisions directly unlock or stall innovation at the silicon level. Investors interested in AI hardware startups have reason to reconsider risk and funding cycles in this space. Finally, founders pursuing vertical integration of AI stacks have to reassess timelines and funding strategies under these new cost dynamics.

What to watch next

The critical signals will be whether Microsoft publicly or privately commits to the 40 percent purchase volume. Watch for announcements from Broadcom about project delays, cost changes, or alternative buyers stepping in. Any disclosure of revised financing terms or OpenAI’s fallback plans can confirm if this standoff leads to long delays or a shift back to established cloud chips. Also tracking how this affects OpenAI’s AI hardware roadmap and discussions on supply chain risk in AI chip development will matter.

AI Quick Briefs Editorial Desk