Cloud Infrastructure Divergence: Google Cloud's Massive TPU Hardware Push vs. Azure and AWS

Cloud Infrastructure Divergence: Google Cloud's Massive TPU Hardware Push vs. Azure and AWS

The cloud infrastructure market has entered a phase of profound divergence. While Microsoft Azure and Amazon Web Services (AWS) continue to act primarily as infrastructure-as-a-service (IaaS) providers heavily reliant on Nvidia hardware, Google Cloud (Alphabet) is executing a major strategic pivot to become a direct hardware vendor, utilizing its custom Tensor Processing Units (TPUs) to challenge Nvidia's dominance.¹

Google's Massive TPU Inventory Build-Up

Google's latest financial reports reveal a staggering, rapid accumulation of hardware inventory designed for direct sale to enterprise customers. Google's inventory has ballooned over the past 12 months:

• June 30, 2025: $628 million in inventory.
• June 30, 2026: $7.74 billion in inventory.

According to Google's 10-Q filing, this inventory consists:

"...primarily of hardware related to TPU systems for sale to enterprise customers and devices," including "raw material components purchased directly from our suppliers; work-in-process inventory undergoing conversion into finished products; and fully assembled finished goods."

This massive $7.1 billion inventory build-up represents a deliberate strategy to sell complete TPU systems directly to select customers, rather than just leasing TPU compute time on Google Cloud Platform (GCP). Sundar Pichai clarified this during the earnings call, stating that TPU system sales are included in Google's $514 billion cloud backlog, and that building this inventory ahead of deliveries has impacted Google's operating cash flows.

The $5B Blackstone TPU Cloud Joint Venture

The cornerstone of Google's direct hardware sales strategy is its landmark partnership with Blackstone (BX). Announced in May 2026, Blackstone and Google have formed a joint venture to build a new U.S.-based AI infrastructure company:

• Blackstone Commitment: Blackstone, the world's largest private owner of data centers, has committed $5 billion in equity capital to fund the physical data center buildout.
• Google Commitment: Google will supply the joint venture with its proprietary TPU hardware, software, and services.
• Capacity Target: The partnership plans to bring the first 500 megawatts of compute capacity online by 2027, with plans to scale significantly over time.

This JV represents a direct, major challenge to Nvidia's hardware monopoly. By partnering with Blackstone, Google is creating a massive TPU-only cloud platform that bypasses Nvidia's supply chain entirely, providing enterprise customers with an alternative, high-performance AI compute ecosystem.

Relative Performance of the Cloud Giants

This hardware pivot comes amid strong overall growth for Google Cloud, which is accelerating faster than its peers:

• Google Cloud: Q2 2026 revenue rose to $24.8 billion, up a staggering 82% year-over-year, heavily supported by the initial recognition of TPU system sales and robust GCP agreements.
• Microsoft Azure & AWS: While both continue to grow, they are facing severe capital constraints and margin pressures as they spend heavily on Nvidia's Blackwell chips. Microsoft's capex reached $35.80 billion (39.8% of revenue), and Amazon's capex hit $54.21 billion (27.0% of revenue), pushing both companies into negative free cash flow or lower cash generation.

Google's ability to manufacture its own high-performance AI silicon (TPUs) and sell it directly to enterprise partners like Blackstone gives it a structural cost and supply advantage that Microsoft and Amazon currently lack, fundamentally altering the competitive dynamics of the cloud infrastructure market.