Chipmaker's Shares Soar Following OpenAI - Broadcom Agreement

OpenAI has entered a multiyear partnership with Broadcom Inc. to co-develop custom chips and networking gear, marking another major move in the AI company’s push to expand its computing infrastructure. The announcement sent Broadcom shares soaring as investors welcomed the deepening relationship between two key players in the AI ecosystem.

‍

According to a joint statement released Monday, OpenAI will design specialized hardware in collaboration with Broadcom, with plans to build and deploy up to 10 gigawatts of AI data center capacity. The rollout of servers equipped with the new technology is expected to begin in the second half of 2026.

‍

OpenAI said that by tailoring the chips to its specific needs, the company can integrate insights gained from developing its AI models and services “directly into the hardware,” resulting in more capable and efficient systems. The companies aim to complete the hardware rollout by the end of 2029.

‍

For Broadcom, which supplies components for everything from iPhones to data centers, this deal marks a deeper foray into the rapidly expanding AI chip market. The company’s CEO, Hock Tan, had hinted at this partnership during an earlier earnings call. Following confirmation of the deal, Broadcom’s stock climbed more than 12% in premarket trading in New York.

‍

This collaboration is just the latest in a string of high-profile infrastructure partnerships for OpenAI, the creator of ChatGPT, as it races to alleviate computing constraints. Nvidia Corp., the leading AI chipmaker, announced last month that it will invest up to $100 billion in OpenAI to help build new infrastructure supporting at least 10GW of capacity. Similarly, Advanced Micro Devices Inc. (AMD) recently signed a multiyear agreement with OpenAI to provide 6GW of processing power.

‍

While OpenAI continues to source chips from external suppliers, it has also been quietly designing its own semiconductors, primarily optimized for the inference stage of AI the process that powers responses after models have been trained.

‍

CEO Sam Altman revealed that OpenAI has been working alongside Broadcom for the past 18 months, reimagining technology from the transistor level up to the user experience with ChatGPT. “By optimizing the entire stack,” Altman explained on a company podcast, “we can achieve massive efficiency gains faster, cheaper, and more powerful models.”

‍

Tan echoed this sentiment on the same podcast, saying, “If you build your own chips, you control your destiny.” The partnership positions Broadcom as a key beneficiary of the AI investment boom, which has already lifted its stock by 40% this year, outperforming the 29% gain of the Philadelphia Semiconductor Index. Meanwhile, OpenAI’s valuation has surged to $500 billion, making it the most valuable startup in the world.

‍

The collaboration also reflects OpenAI’s strategy to diversify its hardware ecosystem. By leveraging Broadcom’s Ethernet-based networking technology, OpenAI reduces its reliance on Nvidia’s proprietary interconnect systems. As part of this effort, OpenAI will also design custom gear to accompany the new chips.

‍

Notably, Broadcom won’t operate data centers itself. Instead, it will install server racks containing the specialized hardware in facilities managed by OpenAI or its cloud-computing partners.

‍

As the race to expand AI infrastructure accelerates, industry observers have raised questions about how such large-scale projects are being financed and whether overlapping partnerships could inflate a potential AI bubble.

‍

Unlike its agreements with Nvidia and AMD, OpenAI clarified that this latest deal does not include an equity or investment component. While the company didn’t disclose specific financing details, it emphasized that greater computing capacity will enable it to scale operations and sell more services.

‍

To put the scope of OpenAI’s plans in perspective, one gigawatt roughly equals the energy capacity of a nuclear power plant. Yet even with 10GW of computing power, OpenAI President and co-founder Greg Brockman said the company is still far from reaching its artificial general intelligence (AGI) ambitions. “That’s just a drop in the bucket compared to where we need to go,” Brockman noted.

‍

Charlie Kawwas, Broadcom’s president of semiconductor solutions, added that achieving such scale will take time. “Think about railroads they took a century to become critical infrastructure. The internet took about 30 years. This won’t happen in five,” he said.

‍

Through this Broadcom alliance, OpenAI is signaling that it intends to build the foundation for the next era of AI computing one where efficiency, scalability, and independence are just as important as raw processing power.

‍