Behind OpenAI & TSMCâs First Custom Chip Design
The semiconductor industry faces a critical challenge in meeting the processing requirements of AI systems as AI accelerates all over the world.
Whilst AI models grow in size and complexity, they demand increasing amounts of computing power, driving competition for specialised chips.
This demand has created supply constraints and raised concerns about reliance on single suppliers in a market valued at hundreds of billions of pounds, such as Nvidia, who dominates the market, leading companies to seek alternatives through partnerships or by developing their own AI chips.
As a result, according to Reuters, OpenAI has reportedly said it will finalise the design of its first custom chip in the coming months, with plans to manufacture it at Taiwan Semiconductor Manufacturing Company (TSMC), the world's largest contract chip manufacturer.
OpenAI and Broadcom develop new processor architecture
The chip development programme is led by Richard Ho, who joined OpenAI from Google, where he headed the search company's custom AI chip initiative.
His team has doubled in size through collaboration with Broadcom, yet the team remains smaller than chip development programmes at technology companies like Google and Amazon.
Reuters reports that industry sources indicate that developing a new chip design could cost US$500m for a single version, with costs potentially doubling when including necessary software and peripheral systems.
This means that sources are indicating that OpenAI would need to hire hundreds of additional engineers to match the scale of chip development programmes at other technology companies.
TSMC applies advanced manufacturing process
The processor will reportedly be manufactured using TSMC's 3-nanometer process technology, referring to the size of transistors on the chip; and the design incorporates a systolic array architecture with high-bandwidth memory, a configuration also used in Nvidia's processors, along with networking capabilities.
- OpenAI to complete first custom chip design in 2025
- OpenAI's chip aims to reduce dependency on Nvidia
- TSMC to use 3-nanometer manufacturing process
- OpenAI's chip team is led by Richard Ho, formerly from Google
The chip will support both training AI models and inference operations, which refers to running existing models and sources indicate the processor will be deployed primarily for inference operations within OpenAI's infrastructure.
Inside OpenAI, the training-focused chip is viewed as a strategic tool to strengthen negotiating leverage with other chip suppliers.
Engineers plan to develop increasingly advanced processors with broader capabilities in subsequent iterations.
Technology sector increases AI infrastructure spending
The development comes as other technology companies increase investment in AI infrastructure, such as Meta, which plans to spend US$60bn on AI systems next year.
Meanwhile Microsoft has announced US$80bn in AI infrastructure spending for 2025 and OpenAI's participation in the US$500bn Stargate infrastructure programme, announced by the US government, demonstrates the scale of investment in AI computing resources.
Manufacturing process presents technical hurdles
Successful development of OpenAI’s processor would require multiple stages.
For instance, a typical "tape-out" – the process of sending a first design to a chip factory – costs tens of millions of pounds and takes approximately six months to produce a finished chip, yet this timeline could be shortened with additional investment in expedited manufacturing.
The initial processor may require multiple iterations to achieve functionality and if successful, could enable OpenAI to test alternatives to Nvidia's processors later this year.
Other technology companies have faced challenges in developing custom chips, such as Microsoft and Meta's efforts to produce satisfactory processors having encountered difficulties despite years of development.
Yet recent market developments, including DeepSeek's announcement about reduced chip requirements for AI model development, have raised questions about future processor demand in the AI game.
Explore the latest edition of Technology Magazine and be part of the conversation at our global conference series, Tech & AI LIVE.
Discover all our upcoming events and secure your tickets today.
Technology Magazine is a BizClik brand
