Capital

Google Already Sold $514bn of Cloud Capacity

Alphabet's Q2 raised its 2026 capex to $205bn and the stock fell. The number that matters to you is the $514bn cloud backlog and customers who overshot their contracts by half.

CapitalBy Servola Tech Desk2026-07-234 min read

AI-assisted, edited by humans. | Editor-in-Chief: Leon Soliman | Editorial standards

Google Already Sold $514bn of Cloud Capacity

Key takeaways

Google Cloud is carrying $514 billion in signed-but-undelivered contracts, and its existing customers already ran more than 50 percent past their commitments, so the AI capacity you rent is being pre-sold years out, not held in reserve for you.
Alphabet raised its 2026 building budget to $195-205 billion (about EUR 180-190 billion), up from $180-190 billion, and spent roughly $45 billion in the quarter alone, close to double a year earlier.
Cloud revenue grew 82 percent while the company kept warning it is supply-constrained, which means a tight compute market is the plan, not a passing spike your bill will grow out of.
Gemini reached 950 million monthly users and nearly 90 percent of the Fortune 100 on Gemini Enterprise, so it is settling in as a default layer whether or not you chose it.
Investors sold the stock on the spending; an operator should read the same report as a capacity-and-lock-in warning and put capacity and pricing commitments in writing now.

The number under the falling share price

On its 22 July earnings call, Alphabet told investors it would raise its 2026 capital budget to as much as $205 billion, and the shares slipped even though revenue of $119.8 billion beat expectations and grew 24 percent. The market fixed on the spending line, read it as a margin risk, and sold. That reaction is the easy story, and it misreads what the report is actually about.

The figure that matters to anyone who rents computing is buried one level down. Google Cloud is now sitting on a backlog of $514 billion in contracted work it has not yet delivered, up from roughly $462 billion three months earlier. That is not a forecast. It is signed demand waiting in a queue, and the queue has a length.

A backlog is a queue, and you are standing in it

Alphabet also said existing cloud customers exceeded their prior commitments by more than 50 percent in the quarter. Read those two facts together and the picture is blunt: the companies already inside Google Cloud are consuming far more than they signed for, and half a trillion dollars of contracted capacity is spoken for before an uncommitted buyer walks in. Capacity is not sitting idle waiting for your workload.

For an operator this changes the negotiation. On-demand pricing and available headroom are what you lose first when a provider is this oversubscribed, because contracted customers are served ahead of spot buyers. If your plan assumes you can add GPUs next quarter at today's rate, the backlog is telling you to get that capacity reserved in writing, at a fixed price, before the next wave of committed demand prices it for you.

Why the spending is less reckless than it looked

The market treated $205 billion of capex as a bet. It is closer to a delivery schedule. A backlog that large is a promise Alphabet has already made to paying customers, and the buildout is what keeps those promises. Spending nearly $45 billion in a single quarter, roughly twice the year-earlier pace, is the cost of not defaulting on demand it has booked.

That distinction matters to you because it tells you the compute crunch is structural, not a spike that unwinds. When the largest cloud vendors are all raising capex into supply they have already sold, the shortage keeping your bill high is the industry's steady state for now. Planning for prices to soften as capacity catches up is planning against what the operators themselves are signalling.

Gemini is becoming a default you did not pick

Underneath the cloud numbers sits a quieter lock-in. The Gemini app reached 950 million monthly users, its daily users tripled year over year, and nearly 90 percent of the Fortune 100 now run Gemini Enterprise. Google's models are processing about 22 billion tokens a minute, up from 16 billion a quarter earlier, across more than nine million developers. That is the shape of a default forming.

Defaults are expensive to leave. As Gemini threads into the enterprise tools your suppliers and partners already use, the cost of standardising on a different model quietly rises, and the switching bill lands later, when you have less leverage. This is worth pricing now, while the choice is still yours to make on your terms rather than inherited from a vendor's roadmap.

What an operator should do before the next print

Treat the backlog as a supply warning aimed at you. Move the compute you actually depend on onto reserved, committed capacity with pricing fixed for the term, rather than assuming on-demand availability at a stable rate. Ask your provider, in writing, what happens to your headroom when contracted demand climbs, and get the answer before you need it.

Then price the Gemini question deliberately. Decide whether a default enterprise layer is a choice you are making or one being made for you, and cost the switch while it is still cheap. Investors read this quarter as Alphabet spending too much; an owner who rents compute should read it as the market it depends on tightening on purpose, and act while the terms are still negotiable.

Frequently asked questions

Does Google's $514 billion cloud backlog mean I cannot get capacity?

Not that you cannot, but that a great deal of it is contracted years out and served before uncommitted buyers. Existing customers already overshot their commitments by more than half, so the safe move is to negotiate reserved capacity in writing rather than rely on on-demand availability.

If earnings beat expectations, why did the stock fall?

Because Alphabet lifted 2026 capital spending to as much as $205 billion, and investors worried about near-term margins. For a customer, though, that spending funds a backlog it has already signed, which is why the capacity story matters more than the share-price reaction.

Should this change how I plan cloud and AI spending?

Yes. Assume the compute shortage is structural rather than temporary, lock reserved capacity and pricing for what you depend on, and put a number on the cost of leaving Gemini before it becomes a default you inherit rather than one you chose.

The headline was a spending figure and a falling share price. The story for anyone who rents compute is quieter and harder to unwind: the capacity has already been sold, and the queue is forming in contracts you have not signed yet.

AlphabetGoogle CloudCloud backlogAI capexGeminiCompute capacity

Google Delayed Gemini Because Coding Fell Short

Google has pushed Gemini 3.5 Pro months past its expected launch after the model missed internal coding and reasoning targets, and Alphabet shed about $225 billion in market value. Why the one axis it missed is the one that matters most for the software you build.

3 min read

Artificial Intelligence

Google's Flagship Model Is Late With No New Date

Google promised Gemini 3.5 Pro in June. It is months behind on coding, no new date exists, and Alphabet lost 200 billion dollars in a day. What that means for your roadmap.

4 min read

Law & Regulation

The EA Buyout Now Hinges on Brussels

The largest take-private in history has cleared EA shareholders and US antitrust. The last real gate is the EU - and not classic merger review, but the Foreign Subsidies Regulation, facing its first blockbuster test.