Google has strengthened its position in artificial intelligence infrastructure through its homegrown tensor processing units (TPUs), which power the company's Gemini chatbot and represent an integral part of its fast-growing cloud-computing business. Wall Street projects Google Cloud revenue to surge roughly 64% this year to $96 billion, according to FactSet, with analysts modeling growth above 50% continuing in 2027. The TPUs deliver cost advantages by consuming 20% to 40% less energy than Nvidia processors, allowing Google to charge about 20% to 30% less for excess compute capacity, according to William Blair analyst Ralph Schackart. This positions Alphabet as a major force in AI infrastructure even as Google Cloud still trails Amazon Web Services and Microsoft Azure in revenue. The company's custom silicon approach addresses surging demand for AI computing power while reducing operational costs, with CEO Sundar Pichai pointing to a 78% reduction in Gemini serving unit costs across 2025.
Google TPUs Deliver Cost Efficiency Through Specialized Design
Google's TPUs belong to a class of chips called application-specific integrated circuits (ASICs), which are designed specifically for machine learning tasks like training models and running them in real time through inference. Brad Gastwirth, global head of market research and market intelligence at Circular Technology, likened ASICs to a custom suit tailored for certain tasks rather than a person's body. Google co-designs the chips with Broadcom.
The specialization enables TPUs to deliver more computing output with less power. "Most ASICs consume 20% to 40% less energy than Nvidia processors, allowing for greater performance-per-dollar," William Blair analyst Ralph Schackart stated. These cost advantages allow Google to charge about 20% to 30% less for excess compute capacity, attracting AI companies to Google's cloud business and enterprise services.
The latest eighth-generation TPUs, announced in late April at the Google Cloud Next conference, mark the first time Google has split its chip lineup into two specialized variants: the TPU 8t for model training and the TPU 8i for inference. Google stated the chips are up to three times faster for AI model training, offer 80% better performance per dollar, and can run more than 1 million TPUs in a single cluster. "This gives us the ability to create the largest training cluster in the world," Pichai said at Google's I/O developer conference last month.
Nvidia Maintains Dominant Position in AI Computing Market
Nvidia remains the biggest player in AI compute, with its GPUs holding a dominant position in training AI models and day-to-day inference. The company's GPUs offer more flexibility than an ASIC such as a TPU, as they were originally designed to render 3D computer graphics before their processing power was harnessed for AI. Nvidia holds a major advantage with its CUDA software system, which developers have built around for years. CEO Jensen Huang argued on an earnings call last year that "the reason why developers love us is because we're literally everywhere."
Analysts at Stifel wrote in a May research note that Nvidia remains the "broad ecosystem leader," with its dominant market share insulated in the near future. However, they argued Nvidia's "moat is increasingly being tested." The analysts said the market is shifting from a "training-led regime toward inference-led regime by the end of 2026," placing greater focus on compute costs and return on investment. This evolution is accelerating hyperscalers' interest in homegrown ASICs and alternative AI chips.
Anthropic and Meta Sign Multi-Billion-Dollar TPU Deals
Anthropic has committed to using multiple gigawatts of Google TPUs to increase its computing resources as demand for its models and services surges. Meta Platforms signed a multi-billion-dollar deal with Alphabet in February to use Google's TPUs. Customers rent access to the chips through Google's cloud business, and in some cases can now buy TPUs for their own data centers.
Google Cloud CEO Thomas Kurian noted on the Future Forward podcast on April 25 that he is seeing TPU demand beyond AI labs, into market segments like finance and energy. Financial firm Citadel Securities is using Google's TPUs for high-performance financial modeling, and all 17 U.S. Department of Energy national laboratories use AI co-scientist software developed by Google and powered by Gemini, built on the chips.
Google Cloud Backlog Reaches $472 Billion Driven by TPU Demand
Alphabet CFO Anat Ashkenazi said Google Cloud backlog nearly doubled sequentially to $472 billion by the end of the first quarter, driven by strong demand for enterprise AI offerings and the inclusion of TPU hardware sales for customers' own data centers. Analysts at Citizens forecasted in a note last month that Google will generate about $3 billion of revenue from TPU-related infrastructure in 2026, before jumping to $25 billion in 2027. "Importantly, we believe TPU monetization is not fully reflected in current consensus estimates, indicating meaningful upside potential," analysts wrote in early May.
Kurian explained in an April Future Forward podcast interview that "we make great margins no matter which way we're selling it because we own our own IP." He added that since chip demand is likely to exceed supply for years in an already capacity-constrained environment, "unit economics get more expensive, and in our case, because we control our chip, the unit economics remain attractive."
Blackstone Commits $5 Billion to Joint TPU Cloud Venture
Google established a new AI compute venture with asset management giant Blackstone, built around the TPU. Blackstone is committing $5 billion in initial equity to the venture, with plans to bring 500 megawatts of capacity online by 2027 and scale from there. Google will supply the hardware, software, and infrastructure expertise. A job posting on LinkedIn is currently available for the chief operating officer of the "Blackstone and Google TPU Cloud Company."
Piper Sandler wrote last month in a research note that the joint venture with Blackstone is "another vote of confidence in TPUs and allows Google to increase its commitment to Cloud without the significant capital requirements." Analysts called it a "capital-light way for Google to keep driving TPU momentum."
Alphabet shares are down 16% from their early-May peak, coinciding with a broader period of weakness among hyperscalers. For the year, Alphabet shares are still up about 8%, outperforming Microsoft, Amazon, and Meta Platforms.
FAQ
What are Google's TPUs and how do they differ from Nvidia's GPUs?
Google's tensor processing units (TPUs) are application-specific integrated circuits (ASICs) co-designed with Broadcom and optimized specifically for machine learning tasks like training models and inference. They consume 20% to 40% less energy than Nvidia processors, allowing Google to charge about 20% to 30% less for excess compute capacity, according to William Blair analyst Ralph Schackart. Nvidia's GPUs offer more flexibility as general-purpose processors originally designed for 3D graphics rendering, and hold a dominant market position with advantages including the CUDA software system that developers have built around for years.
How much revenue is Google projected to generate from its cloud business this year?
Wall Street projects Google Cloud revenue to surge roughly 64% this year to $96 billion, according to FactSet. Analysts model growth above 50% continuing in 2027. Alphabet CFO Anat Ashkenazi reported that Google Cloud backlog nearly doubled sequentially to $472 billion by the end of the first quarter, driven by strong demand for enterprise AI offerings and TPU hardware sales. Analysts at Citizens forecasted that Google will generate about $3 billion of revenue from TPU-related infrastructure in 2026, before jumping to $25 billion in 2027.
Which major companies have signed deals to use Google's TPUs?
Anthropic has committed to using multiple gigawatts of Google TPUs to increase its computing resources as demand for its models surges. Meta Platforms signed a multi-billion-dollar deal with Alphabet in February to use Google's TPUs. Blackstone is committing $5 billion in initial equity to a joint TPU cloud venture with Google, with plans to bring 500 megawatts of capacity online by 2027. Additionally, financial firm Citadel Securities is using TPUs for high-performance financial modeling, and all 17 U.S. Department of Energy national laboratories use AI software powered by Gemini and built on the chips.
