OpenAI and Broadcom Agree to Jointly Develop Custom AI Chips
The two tech giants are joining forces to develop specialized AI chips and networking equipment, aiming to deploy a staggering 10 gigawatts of computing power by 2029. This collaboration marks a significant move to create alternatives to Nvidia's dominant hardware.
A Strategic Collaboration
- OpenAI will design custom chips tailored for training and running advanced AI models.
- Broadcom will handle the manufacturing, leveraging its proprietary technology.
- While the companies have been working together for 18 months, they have only now publicly announced their partnership.
Timeline and Market Impact
- Initial hardware shipments are scheduled to begin in 2026.
- The goal is to deploy equipment with a total capacity of 10 GW by 2029.
- Following the announcement, Broadcom's stock jumped by over 9% to $354.49 per share.
The Staggering Cost of AI Ambition
According to a Bloomberg report, the estimated cost for 10 GW of AI hardware is a colossal $350 billion.
This massive investment is just one piece of OpenAI's strategy. The company also has agreements with Nvidia and AMD to procure an additional 16 GW of chip capacity, pushing its total infrastructure spending into the hundreds of billions.
To justify such monumental costs, the AI market's revenue will need to explode. A Bain & Co. analysis cited by the WSJ suggests the market must reach $2 trillion by 2030. For perspective, that figure surpasses the entire 2024 combined revenue of Amazon, Apple, Alphabet, Microsoft, Meta, and Nvidia.