Big Tech's AI Capex Surpasses $800 Billion in 2026 on Track for $1.3 Trillion in 2027 as Nvidia Projects 70% Growth
Nvidia's Q2 FY2027 earnings results (reported August 26, 2026) have firmly put to rest any lingering skepticism regarding the longevity and scale of the artificial intelligence capital expenditure cycle. The AI infrastructure build-out has transitioned from a speculative sprint into a multi-year, multi-trillion-dollar industrial project, backed by concrete cloud backlogs and expanding supplier commitments.
Blockbuster Hyperscaler Capex Guidance and Cloud Backlog
Nvidia CFO Colette Kress stunned Wall Street by revealing that top-five hyperscaler capex is projected to hit nearly $800 billion in 2026 and surge to $1.3 trillion in 2027, driven by a massive, contracted pipeline of cloud demand.
"With cloud industry backlog now greater than $2 trillion, CapEx by the top 5 hyperscalers is expected to reach nearly $800 billion in 2026 and $1.3 trillion in 20271." — Colette Kress, Nvidia CFO, quoted in Yahoo Finance
This $1.3 trillion projection for 2027 represents an unprecedented level of capital intensity. The $2 trillion backlog represents the contracted pipeline that cloud providers are racing to build out, ensuring that every GPU Nvidia can produce has a guaranteed, billable home.
Demand Remains Unconstrained, Growth Capped by Supply at ~70%
For fiscal year 2028 (roughly calendar year 2027), Nvidia has guided to revenue growth of approximately 70%. Crucially, both management and analysts emphasize that this growth rate is entirely supply-constrained rather than demand-constrained. Customer demand points to a doubling of growth, but Nvidia's tight supply chain—specifically HBM memory and TSMC CoWoS capacity—caps their near-term execution.
"even though our demand is much greater than 70%, our supply allows us to confidently deliver 70%." — Jensen Huang, Nvidia CEO, quoted in Yahoo Finance
Broadening Customer Base Combats Concentration Fears
A key bear thesis has been Nvidia's extreme customer concentration among a handful of US hyperscalers. However, Q2 FY2027 results showed a dramatic broadening of the buyer base. Data Center revenue grew 117% YoY to $89.0 billion. Within this:
- Hyperscalers generated $48.7 billion (up 101.2% YoY), representing about 55% of the total.
- ACIE (AI clouds, industrial/enterprise, sovereign AI) generated $40.3 billion (up 138.5% YoY), representing 45% of the total.
This rapid expansion of the non-hyperscaler bucket demonstrates that enterprise, sovereign nations, and specialized AI clouds are absorbing massive volumes of compute, diversifying Nvidia's revenue stream.
AWS GPU Commitment Expansion
To cement this multi-year runway, AWS has expanded its commitment to Nvidia's hardware. AWS is deploying an additional 2 million GPUs starting Q2 FY2027 through Q2 FY2029, integrating Nvidia's Vera CPUs and Rubin platforms.
-
An instance of Generational compute capacity is fully monetized and locked in before the silicon even exists. — Trillions of dollars in future cloud infrastructure revenues are locked in and monetized via contracted backlog years before the hardware is even deployed. ↩︎