OpenAI Collaborates with Broadcom and TSMC on AI Chip Development

OpenAI is partnering with Broadcom and TSMC to develop its own AI chip, according to Reuters, citing sources.

The AI startup has reportedly considered various options to diversify its processor supply, including:

• raising capital for an expensive project to build a network of factories and establish a complete production process;
• developing its own processors.

The company has temporarily shelved its ambitious plan to create foundries due to high financial and time costs and plans to focus on designing chips independently.

OpenAI has been collaborating with Broadcom for several months to create a chip aimed at supporting the operation of already trained AI. Currently, the demand for processors to train neural networks is higher, but the need for chips for inference may grow as more applications enter the market, journalists noted.

Sam Altman’s company is still considering developing or purchasing components for its chip and may bring in additional partners, the agency noted. A team of about 20 people has been formed to create the processor, led by top developers from Google, including Thomas Norrie and Richard Ho.

Through Broadcom, OpenAI has secured TSMC’s manufacturing capabilities, which are ready to produce the first chips in 2026. The timeline may change, Reuters emphasized.

Journalists highlighted that the firm’s drive to modernize its AI processor supply system is linked to the high costs of these processors and maintaining AI operations.

Earlier reports indicated that OpenAI could face a potential loss of $5 billion in 2024. Altman’s startup plans to spend about $7 billion on maintaining and training artificial intelligence, with another $1.5 billion allocated for personnel.

In early October, OpenAI raised $6.6 billion at a valuation of $157 billion.

Back in September, journalists learned of Altman’s plans to undertake a large-scale project to create AI hardware and systems worth tens of billions of dollars.