Google and Amazon are moving to sell their in-house AI chips to external companies, stepping beyond using them solely in their own cloud services. Google has offered a $3.2 billion financial guarantee to supply its TPU chips for a New York data centre project, while Amazon is in active discussions to sell its Trainium chips to outside operators.
Google uses financing to push TPU adoption
Google is providing its Tensor Processing Units through financing support to the "Lake Mariner" project, an AI data centre cluster under construction in western New York state. The $3.2 billion financial guarantee backs a deal where project developers will supply Anthropic with computing power built on thousands of Google TPUs, according to the Wall Street Journal. This financing tactic mirrors a strategy Nvidia has long used to expand demand for its GPUs.
Internal pressure at Google's cloud business unit to move faster on AI chips has grown following an executive reshuffle. The company also recently announced an AI cloud joint venture with Blackstone using TPUs. The venture aims to bring 500 megawatts of capacity online in 2027 - roughly the electricity supply of a small-to-mid-sized city - with plans to increase capacity sharply afterward. Google will supply hardware, software, and services, including its in-house TPU chips.
Google has already signed large-scale TPU supply deals with Anthropic and Meta. It first used TPUs internally to train models and develop services, then sold them through Google Cloud, and now plans to sell directly to companies operating infrastructure outside its cloud. In April, Google also unveiled a new processor built for AI inference.
Amazon sees a $50 billion chip opportunity
Amazon Web Services is discussing ways to sell its Trainium AI chips to other companies, Bloomberg reported. CEO Andy Jassy said in his annual shareholder letter that demand for the company's AI chips was very high. "If the chip business, as an independent business entity, sells the chips produced this year to AWS and outside companies, annual revenue would reach about $50 billion," Jassy said. "Demand is so high that it is highly likely we will sell chips to outside companies going forward."
That $50 billion figure falls well short of Nvidia's revenue but matches Intel's annual revenue. It signals how seriously Amazon views external chip sales as a standalone business line, not just a cost-saving measure for its own cloud.
The Nvidia dynamic
Some companies have hesitated to publicly challenge Nvidia's dominance. Adam Fisher, a partner at Bessemer Venture Partners, described a phenomenon he called "Jensen jail," where emerging cloud companies worry they could see reduced allocations of Nvidia GPUs if they buy competing hardware. "Some emerging cloud companies cannot break away from buying Nvidia full-stack hardware because they worry they could be disadvantaged in allocations of Nvidia GPUs or see quotas reduced," Fisher said.
As the shortage of computing resources has worsened, however, more players are working with alternatives. Blackstone's joint venture with Google is one example of a large infrastructure player moving beyond that concern.
Why this matters for sales professionals
When hyperscalers like Google and Amazon start selling their own silicon externally, it reshapes the competitive landscape for anyone selling cloud services, data centre capacity, or AI infrastructure. A second source of advanced AI chips means pricing pressure on Nvidia, more options for customers, and new partnership models - like the financing-backed deals Google is structuring. Sales teams who understand the chip-level dynamics behind AI pricing and availability will have an edge positioning their own offerings. For a deeper dive into how AI shifts sales strategy, explore the AI Learning Path for Sales Representatives.
Your membership also unlocks:
