Google Cloud’s 82% Growth and TPU Push: A Real Threat to AWS and Azure?

Google Cloud’s latest results change the competitive conversation: it is no longer simply the smaller third hyperscaler growing from a low base. Its rapid revenue growth, sharply higher profit, and commercialization of Tensor Processing Units (TPUs) make it a credible destination for new AI training and inference workloads. But AWS and Microsoft Azure retain substantially larger installed businesses and enterprise relationships. The evidence points to a stronger challenger—not a completed reshuffling of cloud leadership. 46

The numbers that changed Google Cloud’s position

In the second quarter of 2026, Google Cloud reported revenue of $24.8 billion, up 82% year over year. Operating income more than tripled to $8.8 billion, lifting the segment operating margin to 35.6% from 20.7% a year earlier. Google also reported a $514 billion cloud backlog, or contracted work not yet recognized as revenue. 435

That combination matters. Fast revenue growth alone can result from spending heavily to secure scarce capacity; growth alongside a higher operating margin suggests Google Cloud was also converting demand into meaningful segment profit. The backlog provides visibility into contracted demand, though it is not the same as near-term revenue and must still be delivered over time. 433

Google’s growth was also accelerating: Google Cloud had grown 63% in the first quarter of 2026 before reaching 82% in the second quarter. 34

Google is growing faster, but AWS and Azure remain larger

AWS generated $42.23 billion in quarterly revenue, up nearly 37% year over year, and reported $16.6 billion in operating income. That keeps AWS well ahead of Google Cloud in absolute quarterly sales and profit, even as Google grew at more than twice AWS’s rate. 61

Azure and other cloud services grew 43% in constant currency in Microsoft’s fiscal fourth quarter, while Microsoft reported that Azure had surpassed $100 billion in annual revenue. 1

The practical comparison is straightforward:

• Google Cloud: fastest reported growth among the three, with improving profitability. 14
• AWS: the largest quarterly revenue base and the largest disclosed quarterly operating income in this comparison. 61
• Azure: a much larger established cloud business with strong growth and deep enterprise distribution. 12

Google therefore appears to be winning a larger share of incremental AI infrastructure demand. It has not overtaken AWS or Azure in total cloud scale.

Why external TPU sales matter

TPUs are Google-designed AI accelerators that were historically used for Google’s own services and cloud offerings. In the second quarter, Alphabet began recognizing revenue from TPU systems delivered into customer data centers, creating a new commercial route beyond renting compute through Google Cloud. 3743

Google now describes three ways to monetize TPUs:

1. Customers can rent TPU compute through Google Cloud.
2. Google can sell TPU systems for deployment in customer data centers; Anthropic has been identified as one such customer.
3. Google can sell TPUs through a cloud-computing joint venture with Blackstone. 2027

Reports also identify Anthropic and Meta as external TPU customers. 20

This changes Google’s strategic position in two ways. First, TPU capability can attract customers that want AI infrastructure without relying exclusively on Nvidia-based systems. Second, direct system sales expand Google’s possible revenue pool beyond workloads that run inside GCP. That said, the new reporting treatment also means TPU system sales contributed to the reported cloud growth, so readers should not treat the entire 82% increase as recurring cloud-consumption revenue. 3337

How strong is Google’s accelerator advantage?

Google Cloud CEO Thomas Kurian said Google’s accelerator-chip business is more than twice the size of any rival hyperscaler’s accelerator business. That is a strategically notable claim, particularly as other large cloud providers develop their own custom AI silicon. But it is a company statement, not an independently standardized market-share measure, so it should not be read as definitive proof of a sector-wide lead. 34

The more verifiable point is that Google has moved from internal TPU use toward external monetization. Its competitive advantage will depend not only on chip supply, but also on price-performance, software tooling, customer support, and how easily enterprises can move AI workloads onto TPUs.

The projections are large—but they are projections

Morgan Stanley estimated TPU-related Google Cloud revenue of $84 billion in 2027 and $108 billion in 2028. Its analysis raised the assumed value of external TPU sales to about $27 billion per gigawatt and increased its assumed gross margin to 30% from 20%. 1829

Those estimates show why investors see TPUs as a potentially major new growth engine. They are not Alphabet revenue guidance, nor are they equivalent to signed, recognized revenue. They rely on assumptions about deployment volumes, pricing, and margins that could change as competitors add capacity and customers gain more alternatives. 1817

Market share: fast gains from a smaller base

Market-share estimates vary based on the definition of cloud infrastructure spending and the measurement date. One estimate placed Google Cloud at roughly 15%, while another cited estimates closer to 11%–12%, compared with approximately 31% for AWS and 23%–25% for Azure. 117

The exact percentage is less important than the direction of the comparison: Google remains materially smaller than AWS and Azure, and its high percentage growth partly reflects that smaller starting base. A durable shift would require Google to keep winning workloads as its revenue base grows—not merely post a few exceptional quarters during an AI capacity shortage.

Is this a durable enterprise AI shift?

The case for a durable shift

Google has several concrete signals in its favor:

• Cloud revenue, operating income, and margin all increased sharply at once. 4
• The $514 billion backlog indicates significant contracted demand, even though timing and conversion remain uncertain. 433
• TPUs now reach customers through cloud rental and direct system sales, broadening the business model. 20
• Large external AI users provide evidence that TPUs can be relevant beyond Google’s own infrastructure. 20

Together, these factors make Google a more serious option for organizations choosing infrastructure for new AI deployments.

The case for caution

AWS and Azure still benefit from larger enterprise footprints, existing contracts, data estates, identity systems, and operational familiarity. Those are meaningful switching barriers for broad enterprise workloads. 2

There is also a cyclical risk. AI-compute capacity has been scarce, and direct TPU system sales are newly included in cloud revenue. As supply expands, Google will need to demonstrate that customers keep consuming or buying TPU capacity at attractive prices and margins. 3337

What to watch next

The next phase of the contest is less about one headline growth rate and more about execution. The key indicators are whether Google Cloud can sustain high utilization and margins, convert backlog into revenue, expand repeat TPU deployments, and make TPU-based workloads easy for enterprise developers to adopt.

Google Cloud has clearly strengthened its position against AWS and Azure in the AI infrastructure race. Its TPU business gives it a differentiated path to compete for new workloads. But AWS remains the scale leader, Azure remains deeply embedded in enterprise technology stacks, and Google still has to prove that today’s AI surge can become a durable market-share gain. 614