Google will begin selling its custom Tensor Processing Units (TPUs) to select cloud customers, Alphabet revealed during its Q1 2026 earnings call on April 29, marking a significant shift in how the company monetises its proprietary AI hardware. CEO Sundar Pichai also offered the first hints of what to expect at Google I/O next month, as YouTube Premium continued to record subscriber growth.
Google Cloud will open access to its custom-built Tensor Processing Units for external customers, the company confirmed during Alphabet's first-quarter 2026 earnings call. The move represents a notable strategic pivot: TPUs have long been a competitive advantage Google kept largely in-house, powering its AI research and cloud services. Now, customers who want access to Google's AI silicon — alongside conventional GPUs — will be able to purchase it directly.
According to The Register, the decision is partly demand-driven. Cloud customers have been requesting access to TPUs, and Google sees an opportunity to diversify its hardware revenue beyond advertising and subscription services. The offering will initially be available to select customers, suggesting a phased or invitation-based rollout rather than an immediate open market.
Alphabet CEO Sundar Pichai used the earnings call to set expectations for Google I/O 2026, the company's annual developer conference scheduled for next month. While specific product announcements were not disclosed, Pichai's tease signals that Google is preparing a significant slate of reveals — likely spanning AI, Android, and cloud services — as competition with Microsoft, Apple, and Amazon intensifies.
On the consumer side, YouTube Premium continued its growth trajectory during the quarter. Alphabet did not release specific subscriber figures in the earnings summary, but the service's expansion adds another pillar to Google's diversified revenue strategy as the company seeks to reduce its reliance on advertising income.
The earnings call also highlighted that AI is increasingly driving both search volume and advertising revenue — a trend that has become central to Alphabet's financial narrative over the past two years. Google has been integrating AI-generated summaries and responses into its core search product, and the company indicated this is contributing positively to user engagement and ad performance.
The TPU announcement positions Google more directly against Nvidia, which dominates the AI chip market, as well as Amazon Web Services and Microsoft Azure, both of which offer custom silicon to cloud customers. By commercialising its TPUs, Google gains a new competitive lever while potentially recouping some of its substantial investment in custom hardware development.
Analysis
Why This Matters
- Revenue diversification: Selling TPUs opens a new hardware revenue stream for Alphabet, reducing dependence on advertising and strengthening its cloud business against AWS and Azure.
- AI chip competition: Google entering the external AI chip market challenges Nvidia's dominance and signals that proprietary silicon is becoming a commercial product category, not just an internal tool.
- Developer ecosystem: If I/O 2026 delivers major AI and platform announcements, it could shift developer priorities and accelerate adoption of Google's tools heading into the second half of 2026.
Background
Google developed its first TPU internally around 2015 to accelerate machine learning workloads, particularly for its TensorFlow framework. For years, TPUs were available only indirectly through Google Cloud's hosted services — customers could rent TPU-powered virtual machines but could not purchase the chips outright. This kept Google's custom silicon at arm's length from the open market.
As AI infrastructure spending surged from 2023 onwards, cloud customers began demanding more flexibility over their hardware stack. Amazon and Microsoft have both moved toward custom silicon offerings — AWS with its Trainium and Inferentia chips, Microsoft with its Maia AI accelerator — intensifying pressure on Google to commercialise its own hardware more aggressively.
Alphabet's Q1 2026 results reflect a company navigating a complex transition: maintaining its dominant search and advertising business while investing heavily in AI infrastructure and cloud services to keep pace with rivals.
Key Perspectives
Google/Alphabet: The company frames TPU sales as a response to genuine customer demand, while also acknowledging the strategic benefit of revenue diversification. Selling TPUs directly could help Google monetise the billions it has spent developing custom AI hardware.
Cloud Customers: Enterprises and AI developers gain access to hardware that has previously been inaccessible, potentially offering performance advantages for specific machine learning workloads compared to GPU-based alternatives.
Critics/Skeptics: Analysts may question whether Google can scale TPU production for external demand without compromising its own internal AI research needs. There are also questions about whether TPUs — optimised for TensorFlow and Google's own frameworks — will find broad appeal among customers already invested in Nvidia's CUDA ecosystem.
What to Watch
- TPU availability details: Which customer segments gain early access, pricing tiers, and whether availability expands beyond the initial select group.
- Google I/O 2026 announcements: The developer conference next month will likely reveal the full scope of Google's AI and product roadmap for the year.
- Alphabet revenue breakdown: Future earnings calls will show whether TPU sales and YouTube Premium growth meaningfully reduce Alphabet's dependence on advertising revenue.