Morning Edition · Sunday, June 28, 2026

AI's First Revenue Census Puts Generative-AI Sales at $110 Billion, Above Infrastructure Depreciation

Exponential View's first report measures real end-customer spending, and it finds that quarterly sales outside China now exceed the quarterly cost of data centers and chips wearing out, the first time that has happened.

Save

AI's First Revenue Census Puts Generative-AI Sales at $110 Billion, Above Infrastructure Depreciation

The generative artificial intelligence (AI) sector booked more than 110 billion US dollars in real end-customer revenue over the trailing twelve months, according to the first State of the AI Economy report from Azeem Azhar's Exponential View. The figure deliberately removes supply-chain double counting, so it reflects what end customers actually paid rather than the marked-up total of every reseller and cloud fee in between.

The comparison that matters most to anyone modeling the buildout is revenue against depreciation. The report finds that first-quarter 2026 AI sales outside China reached roughly 25 billion dollars, ahead of the estimated 21 billion in quarterly depreciation on data-center and chip infrastructure, as summarized in early coverage. The annualized run rate now exceeds 175 billion dollars, and the report puts revenue growth at roughly three times the pace of the early mobile and internet eras, drawn from spending data across more than 1,000 companies.

The framing deserves skepticism. Revenue exceeding depreciation in a single non-China quarter is not the same as the industry covering its full cost of capital, which includes the financing and energy that do not appear in a depreciation line, and the comparison excludes China's own large and partly state-funded spending. The author benefits from an optimistic reading, and the figure is a vendor-adjacent estimate rather than audited accounts. What is verifiable is that the methodology measures realized cash, not a projection of total addressable market, which makes it a more reliable baseline than most figures now circulating.

Veracity: Plausible

71/100

If true, who benefits: Model vendors, hyperscalers, and infrastructure investors who need realized revenue to justify the data-center capital spending.
The nuance: The report itself calls this the second straight quarter revenue beat depreciation, not the "first time," and depreciation still consumes about two-thirds of revenue before power and financing.

An open-source-intelligence read of how likely this story is true with its real nuance, not a judgment of any outlet. It assesses the claim, weighing independent and adversarial reporting. How we label confidence.

What this means

If realized AI revenue is genuinely outpacing hardware depreciation, the argument that the buildout is pure capital spending with no return weakens, which supports the valuations of model vendors and the cloud providers reselling their capacity. It also sets a higher expectation for what comes next. Investors will now expect the gap between revenue and the full, financed cost of compute to keep narrowing, not just the accounting-depreciation portion.

What to watch

Whether independent analysts reproduce the 110 billion figure using comparable double-counting adjustments, which would show whether the number is a durable baseline or an outlier estimate.
Hyperscaler guidance on AI revenue versus capital spending in the next earnings cycle, since management commentary will either confirm or contradict the report's claim that income is overtaking infrastructure cost.
How much of the growth concentrates in a few frontier vendors versus broadening across the ecosystem, which signals whether the economics support many businesses or only a few.

Observations to monitor, not financial advice.

3 sources

Synthesized from: Polylog editors · Crypto Briefing · Exponential View

Part of a tracked trend

AI Revenue Versus Buildout Economics

As AI capital spending compounds, deduplicated end-customer revenue and its margin over depreciation become the decisive gauge of whether the compute buildout is self-funding or speculative, and each new data point moves capital toward or away from the trade.