‘Our payback period on AI servers, in aggregate, is less than two years. And on our own silicon—it’s half of GPUs,’ says Google Cloud CEO Thomas Kurian.
CEO Thomas Kurian said Google Cloud’s TPU accelerator chip business is doubling that of its hyperscaler competitors, such as AWS and Microsoft, with Google TPUs’ return on investment better than GPUs in the AI era.
“The size of our accelerator business, our TPU business, is more than twice that of the next largest hyperscaler,” said Kurian recently at the Goldman Sachs Communacopia Technology Conference.
The CEO said Google has a two-year AI server payback period, while Google TPUs have a much “faster” expected payback period compared with GPUs.
“Our payback period on AI servers, in aggregate, is less than two years. And on our own silicon—it’s half of GPUs’ [payback period],” Kurian said.
“We have strong payback periods, and the majority of our infrastructure contracts, the total contract value of the infrastructure, is long-term committed five-year contracts,” he added.
Google Cloud’s TPU Strategy: ‘One Type Of Accelerator’ Doesn’t Solve Everything
Mountain View, Calif.-based Google has been building TPU accelerators for 12 years.
The cloud and AI giant offers a variety of different types of TPU accelerators for different types of models including its Trillium, Ironwood and new TPU-8t accelerator for large-scale workloads that scales to 9,600 chips in a single pod.
“You can have sparse models, dense models, mixture of experts. You need sparse cores for certain things. So one type of accelerator definitely does not solve everything that customers need,” Kurian said.
John Pettit, CTO of Google Cloud Premier partner Promevo, said Google’s ability to provide customers with specific types of TPUs is a major differentiator in the market.
“Google just offers a huge range of configurations both on GCP and in clients’ data centers because some want it to close to their data,” he said. “And what they’re doing with Gemini Enterprise is just amazing.”
Gemini Enterprise With Google TPUs A Major Agentic AI Hit
Google’s flagship agentic AI platform, Gemini Enterprise, runs on Google TPUs.
Kurian said Gemini Enterprise is now used by over 90 percent of Fortune 100 companies and “thousands” of small businesses.
The platform is resonating with customers as it gives users the ability to quickly build custom AI agents for various use cases, according to Pettit.
“Gemini Enterprise has made it possible for customers to build no-code agents. So they can start to prototype and build out some of those workflows,” he said. “They’ve also brought in a lot of first-party connectors to your local Google Drive data, as well as the third-party connectors to other internal systems.”
Pettit said Gemini Enterprise can “create secure landing zones for employees to use AI against their corporate data, which is not something that you see through all the commercial offerings without building up a lot of security scaffolding or other things like that to keep the data protected.”
“So it creates an interesting opportunity for people to roll it out to all of their employees and still feel safe,” he said.
Louisville, Ky.-based Promevo, for its part, recently launched a new Google Gemini Enterprise Accelerator program that guides customers through a full Gemini Enterprise deployment in as few as eight weeks.
Google Winning New Customers ‘Twice As Fast As A Year Ago’; ARR Reaches Nearly $100B
Overall, Kurian said Google Cloud has witnessed strong growth over the past several quarters thanks, in part, to its TPU innovation and enterprise momentum.
“We’re winning new customers roughly twice as fast as a year ago, deepening our relationship with existing customers, capturing market share and, as a result, growing both top-line and operating income,” Google Cloud’s CEO said.
“On top of that, we build our own frontier models, and we’ve seen strong growth with Gemini in large companies, as well as in small companies, and small and medium enterprises,” Kurian said.
Google Cloud generated $24.8 billion in sales during the second quarter of 2026, representing an 82 percent growth increase year over year.
Google Cloud currently has an annual run rate of nearly $100 billion.
