Google invests heavily in developing proprietary AI chips, directly challenging NVIDIA's market monopoly.
Google is expanding its custom artificial intelligence chip business as competition in the global data center market intensifies. The company is leveraging financial guarantees, manufacturing agreements, and software tools to attract more external customers to use its Tensor Processing Units (TPUs).
According to reports, Google provided $3.2 billion in financial guarantees for the Lake Mariner data center complex located in western New York. TeraWulf operates the site, which will deploy computing systems based on Google TPU chips. The project will provide computing capacity to Fluidstack, which will subsequently provide resources for work on Anthropic's Claude AI model.
This arrangement expands Google's role beyond chip design and cloud services. It also positions the company as part of the infrastructure financing structure supporting the use of its hardware.
NVIDIA has provided similar support for GPU projects by helping customers obtain leasing and financing for large-scale computing clusters. Google is now leveraging financial support to increase demand for its processors.
Google initially developed TPUs for its internal services. The company subsequently opened access through Google Cloud and is now seeking more customers outside its operations. Broadcom CEO Hock Tan stated that the TPU business has generated "tens of billions of dollars" in revenue, although Google has not yet released separate financial data.
Reports predict that Google TPU shipments could reach 4.3 million units in 2026 and potentially increase to 35 million by 2028. These estimates remain forecasts rather than confirmed sales data. TrendForce expects that custom AI chip sales in 2026 will grow faster than standard GPU sales, as cloud companies build processors for specific workloads and seek greater cost control and supply chain flexibility.
Meanwhile, Google is expanding its production network. According to reports, Google has placed orders with Intel for over 3 million TPUs for delivery in 2028. Google is also reportedly negotiating with Marvell Technology regarding new custom chip designs. These plans will add suppliers while increasing production capacity.
Hardware alone is insufficient to determine the AI chip market. NVIDIA's CUDA software platform supports AI tools and developers. Google has launched TorchTPU to help PyTorch developers run workloads on TPUs. The software aims to reduce the effort required to migrate certain AI tasks away from NVIDIA systems.
Google is also associated with a $35 billion infrastructure arrangement involving Broadcom, Anthropic, Apollo Global Management, and Blackstone, which covers five U.S. data centers. Reports describe the plan as a financing structure for sites using Google chips. Complete commercial terms have not been disclosed, so the final scale and timeline remain to be determined by agreements between the companies.
Other technology companies are also developing custom AI processors. Amazon, Microsoft, and Meta have invested in their own chips to obtain additional supply options. However, these companies use NVIDIA GPUs for AI workloads. Reports that Google and Amazon notified NVIDIA CEO Jensen Huang of their plans indicate that custom chips and NVIDIA hardware remain part of their current strategy.
Google's approach combines chip design, cloud services, software support, and project financing. The company is leveraging these tools to deploy TPUs across more data centers and attract customers requiring substantial AI computing capacity.