Goldman turns Alphabet's $514 billion backlog into a $300 billion a year bill

Google Cloud's $514 billion order book sits inside an industry pile above $1.5 trillion. Goldman's math says hyperscalers need $300 billion a year of AI revenue to break even, and $1 trillion of user spending to earn real returns.

In this storyGOOGLGS
Vincent JiangVincent Jiang · 3 min read
Share
Sundar Pichai, Alphabet's chief executive, at a European Commission event in 2023
1 / 6Slide 1 of 6
Sundar Pichai, Alphabet's chief executive, whose July earnings call disclosed Google Cloud's $514 billion backlog.

The charge: one order book holds a third of the promises

On 22 July, Sundar Pichai told Alphabet's earnings call that Google Cloud revenue grew 82% year over year and its backlog hit $514 billion 1. That order book now exceeds the $446 billion Alphabet sold across the whole trailing twelve months 1. It swelled by more than $50 billion in a single quarter on multi-year AI commitments 2.

Nine weeks later, Goldman Sachs priced the other side of the promise. The largest hyperscalers need roughly $300 billion a year of AI revenue just to break even, against more than $1.5 trillion of signed backlog across the industry 3. Alphabet's book alone is roughly a third of that total.

The bill arrived before the cash

Cloud itself earned real money. The segment booked $24.8 billion of revenue, and operating income tripled year over year to $8.8 billion, a 35.6% margin 2. Cloud now supplies more than 20% of Alphabet's revenue, up from 14% a year earlier 4.

The spend side moved faster. Management lifted 2026 capex guidance to $195 billion to $205 billion, the stock fell about 7%, and the quarter closed free-cash-flow negative 5. Quarterly capex doubled in a year to $44.9 billion, now 37 cents of every revenue dollar 6. Revenue beat estimates by $2.82 billion while EPS of $2.85 missed by six cents 7.

Alphabet's capex has climbed from 15% to 37% of revenue in eight quarters

  • Revenue
  • Capex
$0B$50B$100B$150BQ3 '24Q4 '24Q1 '25Q2 '25Q3 '25Q4 '25Q1 '26Q2 '26free-cash-flow negative quarter
Data
RevenueCapex
Q3 '24$88.27B$13.06B
Q4 '24$96.47B$14.28B
Q1 '25$90.23B$17.2B
Q2 '25$96.43B$22.45B
Q3 '25$102.35B$23.95B
Q4 '25$113.83B$27.85B
Q1 '26$109.9B$35.67B
Q2 '26$119.8B$44.92B
Quarterly capital expenditure and revenue in $ billions, calendar quarters Q3 2024 to Q2 2026, from Alphabet's SEC filings via Sharadar; share of revenue computed from the same series.6

Backlog is a promise, not revenue

Goldman counts $800 billion of hyperscaler capex landing in 2026 and $1.2 trillion penciled in for 2027, a share of GDP it calls larger than any technological investment cycle since the railroad build-out of the late 1800s 38. Against that bill, hyperscaler cloud revenue is annualizing only about $70 billion above the pre-AI trend 3.

Solid returns need far more than break-even. Ryan Hammond's math has AI users spending about $1 trillion a year on applications, close to two-thirds of the roughly $1.5 trillion the world spends on software in a year 3. Goldman phases the build-out at $633 billion for 2023 to 2025 and $1.73 trillion for 2026 to 2027, with payback expected between 2028 and 2030 9.

Break-even needs $300B a year; real returns need two-thirds of world software spend

  • Estimate
$0B$500B$1,000B$1,500BAI revenue to break even$300BAI user spend for solid returns$1,000BWorld software spending$1,500B
Data
Value
AI revenue to break even (estimate)$300B
AI user spend for solid returns (estimate)$1,000B
World software spending (estimate)$1,500B
Goldman Sachs estimates of required annual AI revenue and AI user spending, against the global software spend cited in the same note; $ billions per year, all figures analyst estimates.3,10

Demand still outruns supply

Pichai's camp has receipts too. Serving very large cloud customers is "highly ROI-positive" over the life of a deal, and only a "very small percentage" of enterprise workloads have shifted to cloud at all 51. Nearly 90% of the Fortune 100 already runs Gemini Enterprise 2. Goldman expects token consumption to rise 24-fold by 2030, with enterprise AI showing up in earnings "in coming quarters" and the skepticism already partly in the price 103.

Every Q4 print gets priced against a $230 billion gap

Every hyperscaler reporting into Q4 gets measured against one arithmetic: $70 billion of cloud revenue above trend, against a $300 billion break-even bar, a $230 billion gap 310. Alphabet's clock is shorter and more specific: management says half the backlog converts to revenue within two years 25. The backlog is a promise with a schedule. The capex is a bill with one.

How this brief was made

01Gathered & sourced242 channels · 1,439 articles▾

Agents swept 242 channels and ingested 1,439 articles, then de-duplicated and ranked them for signal.

02Verified & cross-validated10 claims · 26 data feeds▾
03Reviewed & edited1 human editor▾

One editor read the draft against the evidence, tuned the framing, and signed off before it shipped.

Become a contributor

Reporting on the business of AI and want it read? We take pitches from outside contributors who bring primary sources and a number worth arguing about.

Share

Deepdive

AI-generated from this story and its cited sources. Not investment advice.

Reader comments

0 comments

    Sign up

    Get your curated digest

    After email confirmation, you will receive a daily digest of the most relevant news that matter to your portfolio