TLDR:
- OpenAI is developing its first custom AI chip with Broadcom and TSMC, planned for 2026
- The company has assembled a 20-person chip team led by former Google TPU engineers
- OpenAI is adding AMD chips to diversify supply beyond Nvidia
- The company dropped plans to build its own chip manufacturing network due to costs
- OpenAI’s compute costs are its largest expense, with projected $5B loss in 2024 on $3.7B revenue
OpenAI, the company behind ChatGPT, is taking steps to develop its first custom AI chip in partnership with Broadcom and Taiwan Semiconductor Manufacturing Company (TSMC). The project marks a major move by the AI leader to reduce its reliance on existing chip suppliers and control its computing infrastructure costs.
The company has built a specialized team of approximately 20 engineers, including Thomas Norrie and Richard Ho, who previously developed Tensor Processing Units at Google. This team is focusing on creating an AI inference chip, which processes new data through trained AI models, with manufacturing expected to begin in 2026.
Through its partnership with Broadcom, OpenAI has secured manufacturing capacity at TSMC. Broadcom’s role includes helping OpenAI optimize chip designs for manufacturing and providing components that manage data movement between chips, a crucial factor in AI systems where thousands of processors work together.
OpenAI initially considered building its own network of chip manufacturing factories but dropped these plans due to the high costs and time requirements. Instead, the company is concentrating on chip design while working with established manufacturers.
The company’s move comes as it faces mounting infrastructure costs. OpenAI is projected to lose $5 billion this year on revenue of $3.7 billion, with computing expenses representing its largest cost category. These expenses include hardware, electricity, and cloud services needed to process large datasets and develop AI models.
While developing its custom chip, OpenAI is also diversifying its chip supply by adding Advanced Micro Devices (AMD) processors to its infrastructure. AMD’s new MI300X chips will complement the Nvidia GPUs that OpenAI currently uses. This addition shows how AMD is working to gain market share in a field where Nvidia currently holds over 80% of the market.
OpenAI has maintained a careful approach in its relationship with Nvidia, avoiding aggressive recruitment of Nvidia talent to preserve good relations. This strategy helps ensure continued access to Nvidia’s latest technology, including their new Blackwell chips.
The development of custom chips focuses specifically on inference operations, which industry analysts predict will eventually surpass training chips in demand. While AI model training currently requires more computing power, the deployment of AI applications across various sectors is expected to drive up the need for inference processing.
OpenAI’s chip strategy mirrors similar moves by other tech giants. Companies like Amazon, Meta, Google, and Microsoft have all pursued various combinations of custom chip development and diverse supplier relationships to manage their AI computing needs.
The company is still evaluating whether to develop or acquire additional elements for its chip design. OpenAI may bring in more partners as the project progresses, showing the complex nature of chip development and manufacturing.
AMD has projected $4.5 billion in AI chip sales for 2024, following the launch of their MI300X chips in late 2023. This projection indicates the growing market for AI processors beyond Nvidia’s offerings.
The shortage of AI chips and their rising costs have pushed major technology companies to seek alternatives. OpenAI’s approach combines developing custom solutions while maintaining relationships with multiple suppliers.
In September 2023, OpenAI released new “reasoning” AI models designed to process information more thoroughly before responding. The company recently completed a $6.6 billion funding round, achieving a valuation of $157 billion.
The timing of OpenAI’s chip manufacturing through TSMC may change, as semiconductor production schedules often face adjustments due to technical and market factors.
Stay Ahead of the Market with Benzinga Pro!
Want to trade like a pro? Benzinga Pro gives you the edge you need in today's fast-paced markets. Get real-time news, exclusive insights, and powerful tools trusted by professional traders:
- Breaking market-moving stories before they hit mainstream media
- Live audio squawk for hands-free market updates
- Advanced stock scanner to spot promising trades
- Expert trade ideas and on-demand support