TL;DR — Consensus framed Alphabet’s $175 billion to $185 billion 2026 capex plan as a defensive, speculative burden. The accounting says otherwise: a $242.8 billion revenue backlog, mostly Google Cloud, means much of that spending fulfills contracts already signed. The weak link is funding, not demand.
The prevailing read of Alphabet’s 2026 capex guidance of $175 billion to $185 billion is that it is a defensive burden, a bet on AI demand that may not arrive. That framing ignores a basic accounting bridge between what Alphabet spends and what customers have already committed to pay. In our view, the 2026 surge is less a gamble on future demand than the mechanical fulfillment of a contracted backlog.
The scale of the step-up is real. Alphabet spent $91.4 billion on property and equipment in 2025, so the midpoint of the 2026 range roughly doubles it. The question is what stands behind the spending.
A $242.8 Billion Backlog Turns Most of the 2026 Spend Into Contracted Demand
Alphabet’s 2025 Form 10-K reports $242.8 billion of remaining performance obligations at December 31, 2025, primarily related to Google Cloud. Management put the Cloud backlog alone at $240 billion, up 55% from the prior quarter. This is not a revenue projection. It is a disclosed figure for customer commitments not yet recognized, and it excludes cancellable contracts.
Timing matters as much as size. Alphabet expects to recognize just over 50% of the backlog within 24 months, which implies more than $120 billion of revenue coming due over 2026 and 2027. Building the capacity to serve that revenue is closer to building the factory after the orders are in the book.
Comparing a one-year capex figure with a two-year revenue tranche understates the match. The 10-K says Alphabet depreciates servers and network equipment over six years and data center buildings over seven to 40 years. A server bought in 2026 is still serving contracts in 2031, well after the first half of today’s backlog is recognized. The right comparison is the asset’s life against the contract’s life, and the backlog extends beyond 24 months for nearly half its value.
Three caveats keep this from being a blank check:
- Delivery risk: the 10-K ties recognition timing to Alphabet’s ability to deliver capacity under contract terms;
- Concentration: management attributed the backlog jump to enterprise AI demand from multiple customers, without naming how concentrated it is;
- Related investments: the 10-K flags commercial arrangements with companies Alphabet also invests in, where an impairment could reduce performance obligations.
Alphabet is not betting on AI demand. It is industrializing a backlog its customers have already signed.
Ironwood TPUs Let Alphabet Fill the Backlog Without a Silicon Markup
The margin-contraction thesis assumes Alphabet will keep paying a premium to outside hardware vendors indefinitely. It does not have to. Its seventh-generation TPU, Ironwood, reached general availability in November 2025, and Anthropic has said it plans to use up to one million TPUs. The risk runs the other way too: if chip generations turn over faster than the six-year schedule, depreciation would have to speed up.
Every contract served on internal silicon insources part of the cost of goods sold that would otherwise go to a chip supplier. That is how Cloud margins can expand even while capital intensity peaks. Google Cloud’s operating margin was already 30.1% in Q4 2025, on $5.3 billion of operating income and $17.7 billion of revenue. Our breakdown of how the $240 billion backlog converts to revenue covers the recognition schedule in detail.
Buybacks Were Supposed to Recycle the Backlog Into a Smaller Share Count
The final step in the original thesis was capital return. As backlog turns into revenue and cash, the cash funds repurchases that offset stock-based compensation dilution and shrink the float for Class A and Class C holders. Alphabet retired 240 million shares for $45.4 billion in 2025, and $69.5 billion remained under its repurchase authorization at year-end.
The cash flow was already tightening when this article first ran. Repurchases fell to $5.5 billion in Q4 2025 from $15.6 billion a year earlier, while 2025 free cash flow was $73.3 billion against a 2026 capex plan more than twice that size. Alphabet was also tapping debt markets, including a sterling bond due 2126 issued in February. Our RSU and SBC dilution report quantifies what buybacks need to offset.
Update: The Backlog More Than Doubled, and So Did the Funding Question
Update, October 2026. The demand side of the bridge strengthened. Remaining performance obligations reached $467.6 billion at March 31, 2026 and $519.5 billion at June 30, 2026, of which $513.9 billion related to Google Cloud. Part of the first-quarter jump reflects a reporting change that now includes contracts of one year or less, about $7.3 billion at the time.
The capital return leg did not hold. Alphabet raised capex guidance to $180 billion to $190 billion in April and to $195 billion to $205 billion in July. It bought back no stock in the first half of 2026 and in June raised $49.6 billion of equity, including mandatory convertible preferred shares. The backlog still justifies the spending, but for now it funds infrastructure, not float contraction.
What to watch next: each quarter’s backlog and the share expected within 24 months, Cloud operating margin as Ironwood capacity scales, capex against operating cash flow, and whether repurchases resume once free cash flow turns positive. A stall in backlog growth while capex keeps rising would break the bridge.
Sources: Alphabet’s 2025 Form 10-K and Q2 2026 Form 10-Q, plus quarterly releases, all on SEC EDGAR; call materials at Alphabet Investor Relations; capex guidance reaction via Bloomberg. Disclosure: the author has a long position in Alphabet (GOOGL/GOOG). This analysis is not investment advice. Read the full disclosure on our About page.
