OpenAI and Broadcom Strike Multibillion-Dollar Deal to Build 10 Gigawatts of Custom AI Chips

Broadcom (AVGO) shares soared Monday after OpenAI announced a new partnership with the semiconductor giant to co-develop and deploy up to 10 gigawatts of custom AI accelerators. The collaboration represents one of the largest infrastructure deals in the AI sector to date and highlights OpenAI’s push to expand its hardware ecosystem beyond Nvidia and AMD.

Under the agreement, OpenAI will design the chips and systems, while Broadcom handles development and large-scale deployment. The first chips are expected to begin rolling out in the second half of 2026.

Custom Silicon to Power OpenAI’s Expanding AI Infrastructure

The deal puts Broadcom as a major player in AI compute, a space traditionally dominated by Nvidia. OpenAI said the new accelerators will incorporate learnings from its latest models, allowing for more efficient performance tailored to its unique workloads.

Broadcom will integrate its networking and connectivity solutions across the systems, allowing for faster communication between data centers as OpenAI continues scaling its AI operations. Analysts say the 10-gigawatt deployment goal highlights OpenAI’s intent to gain greater control over the compute powering its rapidly growing suite of AI products.

A Strategic Shift Away from GPU Dependence

OpenAI’s Broadcom partnership follows recent multibillion-dollar agreements with AMD and Nvidia, signaling a strategic shift toward diversifying chip supply. Tech companies from Google to Amazon and Microsoft have already built proprietary AI chips, reducing their exposure to supply bottlenecks and cost pressures.

This diversification reflects a broader trend in Silicon Valley: major AI developers are increasingly prioritizing vertical integration — designing both their software and hardware stacks. The move could gradually reshape the competitive landscape for Nvidia, which currently dominates the market for AI accelerators.

Financial Scale and Market Impact

While the companies didn’t disclose financial terms, analysts estimate the total cost of deploying 10 gigawatts of AI systems could reach tens of billions of dollars. The deal adds to OpenAI’s growing list of infrastructure partnerships, which already include Nvidia, AMD, Oracle, and CoreWeave.

Broadcom’s stock surged more than 9% following the announcement, reflecting optimism that the collaboration could deliver significant long-term revenue and cement its position in the AI hardware ecosystem. However, questions remain about how sustainable OpenAI’s rapid expansion will be, given that the company has yet to turn a consistent profit despite its massive valuation.

Looking Ahead

The OpenAI–Broadcom partnership escalates the global race to build next-generation AI infrastructure. If successful, it could help OpenAI gain greater control over performance and costs while giving Broadcom a powerful foothold in one of tech’s most lucrative growth sectors. Investors will be watching for updates on deployment milestones in 2026 and beyond, as well as whether OpenAI can balance its staggering infrastructure ambitions with profitability. For now, the deal signals that the AI hardware arms race is only accelerating — and Broadcom is now firmly in the spotlight.