OpenAI appears to be taking the next decisive step towards hardware autonomy. According to a report in the Financial Times , OpenAI, the company that brought Generative AI to the fore with the ChatGPT chatbot , is preparing to build its first in-house artificial intelligence (AI) chip , in partnership with American semiconductor giant Broadcom .
From dependence to independence
To date, OpenAI has relied heavily on external vendors, most notably Nvidia, to meet its massive computing needs. Training and running large language models requires thousands of specialized processors, making cost and supply constraints a critical issue.
The new partnership with Broadcom shows OpenAI’s intention to reduce its reliance on Nvidia and gain greater control over its supply chain. This move is not an isolated phenomenon. Rather, it is part of a broader industry strategy: major tech companies such as Google, Amazon, and Meta have already developed their own custom chips to serve their needs internally.
See also: OpenAI: Tests “Thinking effort picker” on ChatGPT
The production plan
According to the information, the chip being designed by OpenAI is not intended to be made available to third-party customers. It will be used exclusively for internal needs, strengthening the infrastructure required for training and developing new versions of ChatGPT and other models.
Broadcom has reportedly already secured orders worth more than $10 billion from a new , unnamed major AI customer. Many analysts believe that OpenAI is behind this deal, which will significantly boost Broadcom's future revenue
The manufacturing process is expected to be carried out in collaboration with Taiwan Semiconductor Manufacturing Company (TSMC), the world's leading semiconductor manufacturer. TSMC's name was already mentioned in previous Reuters reports, which spoke of the development of OpenAI's first in-house chip.
The battle for computing resources
OpenAI's decision reflects the enormous infrastructure challenges facing the AI industry. The explosive demand for AI services has stretched available data centers and increased competition for access to high-performance GPUs.
See also: Parents sue OpenAI over their son's suicide
Nvidia continues to dominate the market with the powerful A100 and H100, but delivery times remain long and prices are particularly high. For companies like OpenAI, which operate at massive scale, this cost is unbearable and simultaneously limits the ability to develop new models.
The strategic choice for «home‑grown» chips can offer:
- Cost reduction in the long term through supply chain control.
- Performance optimization, as custom chips are designed exclusively for specific workloads.
- Strategic independence from suppliers who currently hold an almost monopoly position.
The competitive landscape
OpenAI's move is not surprising given the competition. Google has been developing Tensor Processing Units (TPUs) for its services since 2016. Amazon has launched the Inferentia and Trainiumspecifically for AWS. Meta is also investing in developing its own chips for AI workloads.
In this context, OpenAI is following an expected trajectory, confirming that the major players cannot rely forever on third‑party companies, when the future of innovation depends entirely on access to specialized computing power.
See also: OpenAI prepares new open weight models alongside GPT-5
Challenges and uncertainties
Despite the advantages, developing custom chips is an extremely expensive and complex process. It requires huge investments, high know-how and several years of maturation before it bears fruit. It is no coincidence that even giants like Google took years to bring their solutions to a commercial level.
Additionally, OpenAI will need to ensure that the new chip architecture is fully compatible with its existing software systems and platforms, a transition that is not straightforward and carries the risk of delays or technical hurdles.
The question that remains open is whether OpenAI will manage to build a chip that will rise to the occasion and give it the comparative advantage it seeks in the \"GPU war\". In any case, the next two years will prove decisive, not only for OpenAI but also for the future of the entire artificial intelligence industry.
