Google: Cloud Order Backlog Nearly Doubled From Previous Quarter, Gemini 3.5 Pro to Launch in June
nashnova research
Alphabet disclosed Wednesday that its cloud backlog topped $460 billion, nearly doubling from last quarter; AI products are scaling across the board and Gemini 3.5 Pro is expected in June — signaling AI is shifting from experimentation into revenue delivery.
What does a $460 billion backlog actually tell us?
Alphabet's cloud backlog nearly doubled quarter-on-quarter, surpassing $460 billion.
The company expects over 50% of that backlog to convert into recognized revenue within 24 months.
This means → these aren't just paper contracts. More than half will turn into real revenue within two years, showing clients are deploying, not window-shopping.
In plain terms = cloud has moved from "impressive bookings" to "confirmed cash flow."
How fast is AI usage actually growing?
Alphabet now processes over 3 trillion tokens per day; its products and platforms handle 3.2 quadrillion tokens per month.
Tokens — the smallest units a large language model processes, roughly equivalent to a short chunk of text — are the core measure of how much AI is actually being used.
This means → usage has left the experimental tier and entered industrial scale. Google's AI infrastructure is being called on massively.
User growth and cost cuts — which signal matters more?
AI Overview now has over 2.5 billion monthly active users; Gemini app MAUs topped 900 million in May, up from 400 million a year ago.
AI subscription plans are performing "exceptionally well," while Gemini's serving cost has fallen 78%.
This means → users doubling while costs drop sharply is the classic formula for AI profitability — the larger the scale, the lower the unit cost.
In plain terms = more people are using it, and each one costs less to serve. That is the healthiest growth pattern for a platform business.
Gemini 3.5 Pro launching in June — what does that signal?
Alphabet expects Gemini 3.5 Pro to launch in June.
This reflects Google maintaining a rapid iteration pace on large models, keeping step with OpenAI and Anthropic release cycles.
This means → for cloud customers, choosing Google Cloud locks in continuously upgrading AI capabilities. For rivals, the model arms race shows no sign of slowing.
市场有风险，内容仅供研究参考，不构成投资建议。