Google Cloud booked roughly three dollars of new commitments for every dollar it recognised

Remaining performance obligations rise with what gets signed and fall with what gets recognised. Google Cloud recognised $24.8 billion in the quarter and still added more than $50 billion to the balance ^[1][4]. New commitments signed in those three months were therefore near $75 billion, about three times the revenue recognised ^[1].

Half of $514 billion is $257 billion, and over eight quarters that averages $32.1 billion a quarter, roughly 30 percent above the current run rate ^[3]. At $24.8 billion a quarter the whole backlog is about 5.2 years of revenue ^[2].

Sundar Pichai attributed the increase largely to multi-year commitments tied to AI workloads, according to Crypto Briefing's account of the quarter ^[8][13]. The same account has him putting Gemini Enterprise inside nearly 90 percent of the Fortune 100 ^[9]. That measures how many customers use it. It says nothing about how much they pay.

Google Cloud is 20.7 percent of Alphabet's $119.8 billion of quarterly revenue ^[6][2] and delivered roughly two-thirds of the group's incremental operating profit, with segment operating income tripling to $8.8 billion at about a 35.6 percent margin ^[6][5]. Run the revenue side backwards and the account's claim that cloud drove nearly half of group growth holds: 24 percent on $119.8 billion implies a group increment near $23.2 billion, and 82 percent on $24.8 billion implies $11.2 billion of it, or 48 percent ^[7].

On this evidence, the comparison with AWS at 37 percent growth and Azure at 43 percent is a comparison of rates ^[11][12].

The $195 billion to $205 billion of capital expenditure guidance is Alphabet-wide, not a cloud line ^[10]. At the $200 billion midpoint it is about twice Google Cloud's annualised revenue of $99.2 billion ^[5], and about 78 percent of the $257 billion of contracted revenue due over the next two years ^[9]. The backlog covers roughly 2.6 years of spending at that rate ^[8].

Watch the next two quarters. If cloud revenue averages under $32 billion a quarter while the backlog keeps climbing, customers are signing and not drawing. If the share expected to convert within 24 months slips below half, the commitments are lengthening at the far end, where the capacity has already been paid for. In my view the contracted position is genuine and concentration is the open question: one large AI customer renegotiating would move more of the $514 billion than any price cut out of AWS.