Amazon and Alphabet to Deploy $400B in AI Infrastructure in 2026

Amazon and Alphabet are collectively on track to deploy approximately $400 billion in AI infrastructure during 2026, according to their second-quarter fiscal results. The buildout represents the largest capital commitment by any two companies in computing history, yet executives at both firms cautioned that returns will take years to materialize.

AWS reported $42.23 billion in revenue for the quarter, marking 37% year-over-year growth—its fastest expansion in 18 quarters. Operating margin reached 39.4%, and the division's backlog climbed to $496 billion. CEO Andy Jassy disclosed that both the AI services business and the custom chip business have each surpassed a $25 billion annual run rate, with triple-digit growth rates.

Google Cloud posted even steeper percentage gains, jumping 82% year-over-year to $24.77 billion in revenue with a $460 billion backlog. CEO Sundar Pichai said nearly 90% of Fortune 100 companies now use Gemini Enterprise, and the platform processes 22 billion API tokens per minute. Alphabet's Search business generated $63.27 billion in the quarter, up 17%, providing the cash flow to fund the cloud expansion.

Capital Expenditure at Historic Scale

Amazon guided to approximately $200 billion in capital expenditure for the full year 2026, with $54.21 billion deployed in the second quarter alone. Alphabet projected $175 billion to $185 billion for the year, having spent $44.92 billion in Q2. Despite the enormous outlay, Amazon reported negative $7.6 billion in trailing free cash flow, while Alphabet trades at a forward price-to-earnings ratio of 23, with its legacy advertising businesses funding the infrastructure push.

Jassy warned analysts that data center servers require nearly three years to reach breakeven, meaning the bulk of 2026's capital deployment will not fully monetize until 2028. The timeline creates a multi-year gap between investment and return that both companies are betting their existing businesses can bridge.

Divergent AI Strategies

The two hyperscalers are pursuing different paths to AI revenue. Amazon positions its Bedrock platform as a model-agnostic service, hosting GPT-5.6, Claude Opus 5, Gemma 4, and Grok 4.3. The company has secured multi-gigawatt compute commitments from Anthropic—up to 5 gigawatts—and OpenAI, which will begin drawing approximately 2 gigawatts starting in 2027. Amazon's Trainium custom chips anchor these partnerships. Jassy acknowledged Amazon is also developing its own frontier model but emphasized customer choice as the primary value proposition.

Google Cloud, by contrast, leads with its proprietary Gemini models and tightly integrated AI services across its product stack.

Why it matters

The $400 billion deployment sets the scale for enterprise AI adoption over the next decade, but the three-year breakeven timeline means investors and customers are operating on fundamentally different clocks. Companies signing cloud contracts today are locking in infrastructure that won't reach full economic efficiency until 2028, creating execution risk for both hyperscalers if demand patterns shift or if newer, more efficient architectures emerge before the current generation pays off.

These details were first reported by AI Watch.