Earnings Call Highlights
Alphabet's Q2 2026 call was defined by an unmistakable tone of aggressive confidence, with management projecting bullishness on AI infrastructure demand that has only intensified over the past year. Google Cloud's 82% revenue growth and a $514 billion backlog signal that enterprise AI adoption is accelerating well beyond what the market anticipated, and Sundar Pichai repeatedly framed the current moment as 'very early innings' to justify continued massive capital deployment. Analysts probed hard on model competitiveness, TPU strategy, and the sustainability of CapEx returns, and management pushed back with conviction rather than defensiveness. The dominant theme was supply constraint as a signal of strength, not weakness, with Anat Ashkenazi raising full-year CapEx guidance to $195–$205 billion and flagging that 2027 spend will increase further still. The call will likely be read as a strong validation of the AI infrastructure buildout thesis, with Google positioning itself as a full-stack competitor across chips, models, and enterprise platforms.
- Google Cloud revenue surged 82% year-over-year to $24.8 billion, with cloud backlog reaching $514 billion after growing more than $50 billion sequentially, driven by enterprise AI demand; Anat Ashkenazi noted that 'cloud revenue growth accelerated meaningfully even after excluding the impact of TPU system sales,' signaling the underlying business is robust independent of the new TPU hardware revenue line.
- Alphabet raised its full-year 2026 CapEx guidance to $195–$205 billion from $180–$190 billion, with Ashkenazi explicitly stating 'we continue to expect our CapEx to increase significantly in 2027,' and confirming the company is supply-constrained and plans to use third-party capacity as a bridge in Q3, which will create 'modest margin pressure in the near term.'
- TPU system sales to customer data centers began generating revenue for the first time in Q2, with Alphabet disclosing it expects the 'vast majority of revenues from these agreements will be realized in 2027,' and Pichai described a strategy of placing TPUs in customer and third-party data centers like the Blackstone project to balance external demand against internal frontier model training needs.
- Pichai disclosed that Alphabet has begun 'our most ambitious pretraining run yet for Gemini 4,' describing it as a larger base model designed to compete at the frontier, while also committing to 'almost a monthly cadence' of model releases on top of that base, and noting that 3.6 Flash improved over 10 points on the DeepSeek benchmark versus 3.5 Flash in just six weeks.
- Token consumption metrics underscore the scale of enterprise AI adoption: model APIs are processing approximately 22 billion tokens per minute, up from 16 billion just one quarter ago, and nearly 500 cloud customers have each processed more than 1 trillion tokens in the last year, with over 2,000 enterprises consuming more than 100 billion tokens over the trailing 12 months.