OpenAI reportedly plans to buy Google’s artificial intelligence chips, according to a report by The Information cited by Reuters. The potential deal would bring two major AI rivals closer together as demand for computing power continues to rise. Neither company confirmed the reported purchase, and financial terms were not disclosed. If completed, the arrangement could give OpenAI another source of hardware beyond Nvidia’s widely used graphics processors.
OpenAI weighs Google TPU supply
Google’s Tensor Processing Units, or TPUs, are custom chips designed to accelerate machine-learning workloads. Google Cloud describes them as processors built for training and running AI models. Unlike Nvidia GPUs, which are sold widely across the industry, TPUs have historically been closely associated with Google’s own infrastructure and cloud services.
Reuters reported that OpenAI plans to buy Google’s chips, citing a person familiar with the matter. Reuters’ report said the companies did not immediately provide comments. The report did not specify which TPU generation OpenAI would use, how many chips it could acquire or whether the deal would involve direct purchase or a broader cloud arrangement.
The deal would deepen a surprising partnership
OpenAI and Google compete in consumer and enterprise AI products. Google develops Gemini and controls one of the world’s largest search and cloud businesses, while OpenAI operates ChatGPT and sells access to its models. Their commercial relationship has nevertheless expanded, with OpenAI recently making its frontier models available through Google Cloud.
Using Google’s chips would show how competition and cooperation are developing at the same time. AI companies need enormous amounts of hardware, data-centre capacity and electricity. That pressure can encourage rivals to become suppliers, customers or infrastructure partners even when they compete for users and developers.
Nvidia remains central to AI computing
OpenAI is already one of Nvidia’s largest customers. Its models require processors for both training, which teaches systems from large datasets, and inference, which generates answers for users. Nvidia’s GPUs remain the dominant option for many AI workloads, but buyers are seeking alternatives that could improve availability, cost control and negotiating power.
The reported Google deal would not necessarily replace Nvidia hardware. Instead, it could give OpenAI a more diverse computing portfolio. Google Cloud’s TPU documentation says the chips are built for machine-learning workloads, while Reuters reported that companies such as Anthropic have also increased their use of Google’s processors.
OpenAI is building custom chips too
OpenAI has also pursued its own hardware strategy. In June, the company announced a custom inference chip with Broadcom, designed to improve the speed and efficiency of running large language models. OpenAI and Broadcom’s announcement said the chip programme is intended to support performance and scale as demand grows.
Developing custom hardware can reduce dependence on outside suppliers, but it requires time, engineering talent and manufacturing capacity. Buying or renting chips from other providers can meet immediate needs while internal designs mature. OpenAI may therefore use several approaches rather than rely on one processor family.
AI infrastructure spending keeps climbing
The reported plan arrives as technology companies commit extraordinary sums to AI infrastructure. Reuters said Amazon, Microsoft and Google’s parent Alphabet planned to spend as much as $600 billion combined this year. Investors are watching whether those investments produce enough cloud revenue, productivity gains and new AI services to justify the cost.
The spending race includes chips, data centres, networking equipment and power supplies. It also creates pressure to improve energy efficiency because model training and inference can consume large amounts of electricity. For OpenAI, access to additional hardware could help it serve more users and develop larger systems, but it would also increase operating and capital costs.
Diaspora Digital Media’s report on AI models challenging a core computing theory provides related coverage of the industry’s fast-moving AI research landscape.
Ultimately, OpenAI’s reported TPU plan remains unconfirmed. The story nevertheless highlights a major shift in AI competition, where rivals increasingly depend on one another’s infrastructure. Google could gain chip revenue and cloud relevance, while OpenAI could gain another route to computing capacity. The importance of the deal will depend on its size, timing and whether the companies turn a reported plan into a completed transaction.
