Google Cloud Operating Margin Hit 35.6% as Alphabet's Depreciation Grew 42%

Google Cloud earned $8.8 billion of operating income in the second quarter of 2026, a 35.6% operating margin. Two years earlier the same segment ran at 17.1%. Over the same stretch, Alphabet's quarterly depreciation rose 78%, to $7.1 billion. So far, Google Cloud operating margin is outrunning the depreciation bill the AI build creates. The open question is whether that holds as the record capex of 2026 starts to depreciate.

The short answer. Google Cloud operating margin was 35.6% in Q2 2026, up from 17.1% in Q3 2024. Alphabet's depreciation grew 42% over the same year to $7.1 billion a quarter. Cloud profit is growing about 3 times faster than depreciation, but much of the build has not reached the income statement yet, and the Q3 report is the first test of whether that gap holds.

This post is built on one source type. Every quarterly figure below comes from Alphabet's earnings releases filed with the SEC as Exhibit 99.1 to Form 8-K, read on 27 October 2026. Alphabet's third-quarter 2026 results are due this week. When they land, the tables here can be extended by one row and the argument checked against it.

Google Cloud operating margin by quarter, from the segment note

Alphabet reports three segments: Google Services, Google Cloud and Other Bets. A fourth line, Alphabet-level activities, holds costs it does not assign to any segment. The Google Cloud line is the cleanest public read on whether selling AI infrastructure and services makes money at Google.

The table below lays out 8 consecutive quarters of that line, next to Alphabet's company-wide depreciation for the same quarter. Margin is operating income divided by segment revenue. Nothing here is estimated. The last column is our own ratio, explained in the coverage section further down.

Source: Alphabet quarterly earnings releases, Q3 2025 to Q2 2026, each carrying the prior-year quarter. Depreciation is "depreciation of property and equipment" from the consolidated cash flow statement. Margins and ratios are Zan Digital calculations. Last verified 27 October 2026.

Revenue growth did most of the work

Cloud revenue grew 82% in Q2 2026, to $24.8 billion. Operating income grew 212%, to $8.8 billion. When income grows more than twice as fast as revenue, fixed and semi-fixed costs are being spread over a much larger base. That is operating leverage, and it is the honest headline here.

Alphabet's release attributes the revenue growth to Google Cloud Platform, "across enterprise AI Solutions and enterprise AI Infrastructure, as well as core GCP services." It does not attribute the margin change to any cost line. That silence matters, because the release never says whether depreciation inside Cloud rose faster or slower than revenue.

Incremental margin is the cleaner number

Incremental margin is the extra operating income earned on each extra dollar of revenue, measured year on year. It strips out the legacy base and shows what new business earns. Between Q2 2025 and Q2 2026, Cloud added $11.1 billion of quarterly revenue and $6.0 billion of operating income. That is an incremental margin of about 54%.

The same calculation gives about 43% for Q3 2025, 56% for Q4 2025 and 57% for Q1 2026. I read that run as the strongest single piece of evidence in this post. New Cloud revenue is arriving at margins well above the segment average, while the AI build is at its most intense.

Alphabet's depreciation is growing at 40% to 45% a year

Depreciation is the accounting charge that spreads the cost of a long-lived asset across its useful life. A server bought today does not hit profit today. It reaches the income statement in slices, quarter by quarter, once it is placed in service. Alphabet's 10-Q says cost of revenues includes "depreciation expense, primarily related to our technical infrastructure."

Alphabet's quarterly depreciation of property and equipment went from $3.99 billion in Q3 2024 to $7.10 billion in Q2 2026. Year on year, the growth rate was 40.8% in Q3 2025, 43.6% in Q4 2025, 44.5% in Q1 2026 and 42.1% in Q2 2026. That is a steady band, not an acceleration.

Why depreciation lags capex

Capex is the cash spent on the asset. Depreciation follows it with a delay, for two reasons. Assets under construction, such as a data centre shell, are not depreciated until they are finished and in use. Once in service, each asset is spread across years rather than quarters.

Alphabet completed a review in January 2023 that set the estimated useful life of its servers and certain network equipment at six years, per its Q1 2023 filing. Every $10 billion of servers on that schedule adds about $417 million of depreciation a quarter, before any salvage value. Whether six years is the right life for AI accelerators is argued in the analysis of what server useful lives do to hyperscaler profit, and this post takes the schedule as given.

The capex-to-depreciation ratio

One way to see the queue is to divide capex by depreciation. A ratio of 1 means the company is spending only what it writes off, so the asset base is flat. A high ratio means the asset base is growing fast, and future depreciation is being locked in.

In Q3 2024, Alphabet spent $13.1 billion on property and equipment against $4.0 billion of depreciation, a ratio of 3.3. In Q2 2026 it spent $44.9 billion against $7.1 billion, a ratio of 6.3. On a trailing 12-month basis, capex was $132.4 billion against $25.2 billion of depreciation, or 5.2 times. A year earlier that ratio was 3.8.

Cloud profit now covers Alphabet's whole depreciation bill

The coverage ratio is our original figure for this post. It divides Google Cloud segment operating income by Alphabet's total depreciation in the same quarter. It answers a blunt question: is the business that sells compute earning enough to pay for the wear on all the compute Alphabet owns?

For most of 2024 and early 2025 the answer was no, by a wide margin. Cloud earned about half of the company's depreciation charge. The ratio crossed 1.0 in Q1 2026, at 1.02, and reached 1.24 in Q2 2026. In 7 quarters, Cloud operating income grew 4.5 times while company depreciation grew 1.8 times.

What the ratio does not say

The ratio mixes a segment number with a company number. Most of Alphabet's depreciation supports Search, YouTube and its own model training, not Cloud customers. So a ratio above 1 does not mean Cloud has paid for its own infrastructure. It means Cloud's profit is now large enough to absorb a charge the size of the whole company's.

That is still a useful threshold. In 2024, a sceptic could argue that the AI build was being paid for entirely by advertising. On these numbers, that argument has weakened. The cash side of the same story, including why Alphabet still needed outside capital, is covered in the breakdown of Alphabet's equity raise and its cash flow.

Where Google Cloud's operating margin may be flattered

A segment margin is only as clean as the cost allocation behind it. Alphabet keeps a large cost pool outside all three segments. Its Q2 2026 release says Alphabet-level activities "primarily reflect expenses related to our shared AI research and development."

The segment description goes further. It includes "technical infrastructure usage costs associated with the development of our general AI models." In plain terms, part of the compute used to build Gemini is charged to the parent rather than to Cloud, even though Cloud sells access to those models.

The shared AI research line is growing faster than depreciation

Alphabet-level activities cost $5.8 billion in Q2 2026, up 72% from $3.4 billion a year earlier. That is faster than the 42% growth in depreciation. A rising share of the AI build's running cost sits in a line no segment carries.

This is not an accounting trick. Model research serves Search, YouTube and Cloud at once, and any split would be a judgement. It does mean you cannot read 35.6% as the margin of selling AI to enterprises. It is the margin of selling AI to enterprises after someone else has paid for the frontier model.

A stress test with all of it charged to Cloud

Here is the harshest version. Suppose every dollar of Alphabet-level cost were assigned to Cloud. That is unfair to Cloud, because much of that spend serves the other segments. It sets an outer bound, not an estimate.

On that basis, Q2 2026 Cloud income falls from $8.8 billion to $3.0 billion, a margin of about 12%. The same treatment in Q2 2025 gives a loss of about $0.5 billion, a margin near minus 4%. Even under the harshest allocation, the year-on-year direction is the same. That is why I treat the improvement as real, while treating the level as flattered.

The depreciation still in the Google Cloud pipeline

The 8 quarters above mostly reflect assets bought in 2024 and 2025. The 2026 spend is much larger, and most of it has not yet reached the income statement. Alphabet's CEO said in the Q4 2025 release that 2026 capex was "anticipated to be in the range of $175 to $185 billion."

The balance sheet shows the queue forming. Property and equipment, net, rose from $246.6 billion at 31 December 2025 to $321.2 billion at 30 June 2026, per the Q2 2026 10-Q. That is $74.6 billion, or 30%, in 6 months. Depreciation lags that number, so the next 4 quarters inherit it.

Backlog is the counterweight

The case for patience sits in remaining performance obligations, the contracted revenue not yet recognised. Alphabet's 10-Q reports $513.9 billion of it related to Google Cloud at 30 June 2026. That backlog does not guarantee margin, but it means much of the capacity being built already has a named buyer.

How that backlog compares with Microsoft, Amazon and Oracle, and how much of it is concentrated in a few AI labs, is the subject of the comparison of cloud backlog and RPO across providers. For this post, the point is narrower. Contracted demand is what allows depreciation to rise without margin falling.

Where this argument is weakest

Three problems with the analysis above deserve stating plainly, because each could flip the conclusion.

First, Alphabet does not disclose depreciation by segment. Every comparison in this post sets a segment profit against a company-wide cost. If Cloud's share of depreciation is rising fast, its own charge could be growing well above 42%, and the margin expansion would then look even more impressive. If Cloud's share is small, the comparison flatters the risk. Neither direction can be proved from public filings.

Second, 8 quarters is a short series, and it covers an unusual window. Demand for AI compute has been supply-constrained, which lets a provider price capacity well. A margin earned in a shortage is not evidence of a margin that survives a glut. I would not extrapolate 35.6% into 2027 on this data alone.

Third, the counter-view may simply be right. The fastest margin step came in Q4 2025 and Q1 2026, before the doubled 2026 capex had time to depreciate. A sceptic can argue the margin peaked on old assets, and Q3 2026 is the first quarter that tests it. That is a fair reading, and the data in this post cannot rule it out.

What Alphabet's Q3 report has to show on Google Cloud operating margin

Alphabet's Q3 2026 figures are not out at the time of writing. Rather than guess them, here are three tests that the release will answer, with the threshold I would use for each. Each one can be checked in under 10 minutes from the 8-K.

Thresholds are the author's judgement, not company guidance. Q3 2025 base figures from Alphabet's Q3 2025 earnings release, 29 October 2025.

The depreciation number to anchor on

Q3 2025 depreciation was $5.6 billion. If the 42% growth rate of Q2 2026 simply holds, Q3 2026 depreciation would be about $8.0 billion. That is arithmetic on a trend, not a forecast. A figure well above it would mean the 2026 build is reaching the income statement faster than the last four quarters suggested.

I would treat the first test as decisive. If Cloud's operating income stays above Alphabet's entire depreciation charge in the quarter when that charge is growing fastest, the margin story has survived its first real test. The full picture of how the four hyperscalers fund this spend is set out in the comparison of hyperscaler capex and operating cash flow.

What this means if you buy Google Cloud capacity

If you are a founder or finance lead buying compute, a 35.6% segment margin is a fact about your supplier, and it shapes your negotiation. Your provider is earning well on new business. The incremental margin near 54% says the marginal dollar you add is profitable for them.

That gives you room on committed-use discounts, especially on multi-year terms that help fill capacity Google has already paid for. It also argues against accepting list-price increases framed as "cost recovery." Depreciation is rising, but on these numbers it is not outrunning Cloud revenue. The broader question of how much of the build any customer will pay back is laid out in the payback math behind 2026 AI capex.

The risk runs the other way if Q3 shows the margin bending. A supplier with a falling segment margin and a rising depreciation bill has every reason to tighten discounts at your next renewal. Watch the release before you sign anything long.

Frequently asked questions

Where to start

Open the Q3 2026 release the day it lands and compute one number: Google Cloud operating income divided by depreciation of property and equipment. It takes two lines from two tables. If the result is above 1.0, Cloud is still outrunning the build. If it falls below, the margin story has its first crack.

Then, if you buy Google Cloud, put the incremental margin in front of your account team at renewal. A supplier earning about 54% on new revenue can fund a better committed-use rate than the one on your current order form.

Related reading