Nvidia has been the dominant supplier in the AI chip market for years, but a period of near-total dependence appears to be shifting. Several major technology companies — including OpenAI, Google, Apple and SpaceX — are building or planning custom chips to reduce the risk of relying on a single vendor.
OpenAI’s Jalapeño: an inference chip built with Broadcom
OpenAI announced plans for Jalapeño, its own inference chip produced in partnership with Broadcom. The move signals that OpenAI is less focused on a clean break from existing suppliers and more on hedging: gaining more control over hardware and diversifying its supply chain.
Benefits of custom silicon
Proprietary chip designs bring several concrete advantages:
- Greater control over architecture and manufacturing, helping mitigate external supply risks.
- Hardware tuned to specific software or model requirements, which can deliver efficiency and performance improvements.
- A precedent in Apple’s transition off Intel: Apple achieved notable performance gains after moving to its own chips, a case often cited by other companies.
Overall, the objective for many firms is not an abrupt separation from current suppliers but a protective hedge against single-supplier exposure.
Industry discussion on TechCrunch’s Equity podcast
On TechCrunch’s Equity podcast, hosts Kirsten Korosec, Anthony Ha and Sean O’Kane discussed the custom-chip trend and highlighted a few deals worth watching. The episode emphasized that the trend is driven both by strategic caution and the desire to optimize performance for AI workloads.
Why this matters
Wider adoption of in-house chip development could reshape supply-chain dynamics and market competition: successful, AI-optimized chips from more companies could reduce the market share of dominant hardware vendors and increase options for users. At the same time, building custom silicon requires substantial investment and engineering resources, so many organizations are pursuing gradual diversification rather than an immediate break.
