Product1 publisher3 min readPublished
Google Cloud booked $460 billion of backlog against a $20 billion quarter
On Alphabet's Q1 2026 call Sundar Pichai put cloud backlog above $460 billion, close to six years of cloud revenue at the current rate, while giving Gemini Enterprise's paid user base as a growth rate alone.
The Product Desk · Product desk
What happened
- Google Cloud revenue grew 63% in the quarter and exceeded $20 billion for the first time, Sundar Pichai told Alphabet's Q1 2026 earnings call.
- Google's first-party models now process more than 16 billion tokens a minute through direct customer API use, up from 10 billion the previous quarter.
- Google said the eighth-generation TPU 8i delivers 80% better performance per dollar for low-latency inference than the generation before it.
Compiled by The Product DeskSomething wrong?How this is made
Why it matters
- constraint A signed book worth about 23 quarters of cloud revenue puts every new capacity request behind commitments already made. A mid-sized customer has less to win on delivery dates.
- cost A team that defers migrating off the prior inference generation pays roughly 1.8 times per unit of performance, on Google's own comparison, every month it keeps serving traffic there.
- decision With enterprise adoption disclosed as a growth rate, anyone weighing a Gemini Enterprise expansion has to build the peer evidence from internal active-user counts instead of citing the vendor.
- capability The 60% quarter-over-quarter jump in API token throughput gives buyers a production-usage figure to argue from when a sales team talks about seats and adoption.
Half of $460 billion is $230 billion. Pichai said the backlog nearly doubled quarter-on-quarter to over $460 billion [2], and cloud revenue in the same quarter exceeded $20 billion for the first time [1]. So the book grew by something close to $230 billion in three months, roughly eleven quarters of cloud revenue at that rate, and the whole backlog is about 23 times one quarter, near six years [2] [1].
For anyone renewing a committed-use contract, that multiple is the context for every capacity conversation. New requests queue behind a book already signed. Pichai's remarks do not break the backlog out by contract length or customer concentration. A buyer cannot tell whether the queue ahead of them is thousands of three-year enterprise commitments or a handful of very large deals. Pichai said Cloud "accelerated again this quarter due to strong demand for our AI products and infrastructure" [10].
The figure that would tell an operator whether comparable companies have actually rolled enterprise AI out to staff is the Gemini Enterprise paid user base. What Google gave was 40% growth quarter-over-quarter in paid monthly active users [3]. Paid MAU is the better of the two available metrics, since it counts people who opened the thing. With no base, 40% fits a few thousand users and it fits several million. The consumer side got an absolute number in the same set of remarks: 350 million paid subscriptions, with YouTube and Google One named as the drivers [5].
Direct API traffic is the closest thing on offer to work actually done. First-party models now process more than 16 billion tokens a minute through customer API calls, up from 10 billion the prior quarter [6]. That is 60% growth, faster than the 40% on Gemini Enterprise paid users [3] [5].
Pricing is where a buyer has something to push on. Google says TPU 8i delivers 80% better performance per dollar for low-latency inference than the prior generation [7]. Take that at face value and a dollar buys 1.8 times the performance, so a workload left on older silicon is paying about 1.8 times as much per unit of inference served [4]. TPU 8t is the training part, at three times the processing power of Ironwood and two times the performance [8]. Google says it will be among the first to offer NVIDIA's Vera Rubin NVL72 alongside Blackwell- and Hopper-based instances already available [9].
For the generation you are actually buying, what is the dated availability in your region, written into the contract. And what is your own monthly count of active users against seats purchased, tracked every month, since Google chose to report its version of that ratio as a percentage. On Alphabet's own use of the tools, Pichai said engineers are "now orchestrating fully autonomous digital task forces, and building at a faster velocity" [11].
What to watch
- Whether Alphabet's quarterly filing breaks the $460 billion backlog down by contract duration or customer concentration.
- Whether Google publishes an absolute paid monthly active user count for Gemini Enterprise next quarter.
- Whether TPU 8i and Vera Rubin NVL72 capacity appears with dated regional availability in committed-use contracts.