AI Demand Is Holding. So Why Is BoomRisk Elevated?

NVIDIA’s latest results provide strong evidence against one of the more bearish interpretations of the AI investment cycle: a near-term collapse in demand for AI compute.

The company reported quarterly revenue of $96.2 billion, up 106% from a year earlier. Data Center revenue reached $89.0 billion, up 117%, while NVIDIA guided to approximately $108 billion of revenue for the following quarter.

BoomRisk’s semiconductor-demand proxy therefore remains its clearest counterweight to the pressures visible elsewhere in the dashboard.

Yet the overall BoomRisk reading remains 3.38/5 — Elevated Risk, with five of eight pressure tests currently at stress level.

Those observations are not contradictory.

They are the reason BoomRisk separates demand from the economics surrounding the investment cycle.

The demand proxy is not signalling a collapse

BoomRisk’s Semiconductor Demand pressure test asks whether actual compute demand continues to support the scale of the infrastructure buildout.

The latest NVIDIA results strengthen that evidence considerably.

They do not prove that every part of the AI economy is growing sustainably. NVIDIA is one company, and semiconductor revenue is a proxy rather than a complete measure of end-user AI demand.

But the available evidence does not currently support the proposition that demand for leading AI compute has broken.

That is why Semiconductor Demand remains BoomRisk’s lowest-risk pressure test at 1/5.

This matters because BoomRisk is not designed to produce a predetermined bearish conclusion. Evidence that argues against financial stress must affect the framework just as evidence of deterioration does.

So where is the pressure?

BoomRisk is not primarily trying to determine whether AI is useful or whether demand for AI exists.

It asks a different question:

Are the financial returns supporting the scale of investment keeping pace with the capital being committed?

At present, several indicators suggest increasing tension between those two sides of the cycle.

Capital expenditure is outpacing cash returns

Across Microsoft, Alphabet, Amazon and Meta, BoomRisk currently measures capital-expenditure growth of approximately 72%, while aggregate free-cash-flow growth is approximately -16%.

That puts the Capex & Cash Returns pressure test at 4/5.

This is not evidence that the investment is failing.

Infrastructure investment is inherently forward-looking. Capital is spent before the assets it finances have had time to generate their full economic return.

But the larger the gap becomes — and the longer it persists — the greater the future cash generation required to justify the investment already being made.

That is the imbalance BoomRisk is monitoring.

The monetisation proxy is also lagging investment

BoomRisk’s AI Monetisation pressure test tells a similar story, with an important limitation.

The current proxy compares AWS and Google Cloud revenue growth with matched capital-expenditure growth. It is not a direct measure of AI revenue, because sufficiently comprehensive and comparable AI-specific revenue data is not publicly available.

The current reading shows approximately 24% cloud-revenue growth against roughly 65% matched capex growth.

The resulting gap leaves the pressure test at 4/5.

Again, this does not indicate an absence of commercial demand.

It indicates that the observable revenue proxy is currently expanding substantially more slowly than the investment measure against which BoomRisk compares it.

Strong revenue growth and financial pressure can exist at the same time.

Financing is becoming a larger part of the story

The infrastructure buildout is also beginning to involve financing structures beyond direct hyperscaler capital expenditure.

In August, NVIDIA announced proposed partnerships with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to establish independent compute-financing platforms designed to mobilise more than $500 billion of third-party capital over time.

The proposed platforms are evidence of significant institutional appetite to finance AI infrastructure.

They also make financing conditions increasingly relevant to the investment cycle.

The more infrastructure is supported by external capital, the more eventual utilisation and cash generation matter to the economics of the assets being financed.

BoomRisk’s broad Credit Stress proxy currently remains at 3/5.

In other words, the dashboard is not signalling a credit crisis.

It is signalling that financing deserves continued observation as the scale and structure of the buildout evolve.

Market concentration adds another form of exposure

BoomRisk’s Market Concentration pressure test currently sits at 4/5.

The ten largest companies represent approximately 37% of the S&P 500 under the dashboard’s current measure.

Concentration does not cause a market decline.

Nor does it tell us whether the largest companies are correctly valued.

What it does indicate is that changes in expectations around a relatively small number of companies can have an unusually large influence on the broader index.

That becomes particularly relevant when several of those companies are also among the largest participants in the AI investment cycle.

Why BoomRisk remains elevated rather than high

The current BoomRisk reading is not saying:

AI demand is collapsing.

The semiconductor-demand proxy currently indicates the opposite.

Nor is it saying:

The AI investment boom is about to crash.

BoomRisk does not make that prediction.

The current evidence is more nuanced:

Demand for leading AI compute remains exceptionally strong while several measures of the economics surrounding the investment cycle remain under pressure.

Five of eight pressure tests currently sit at 4/5 stress.

But Semiconductor Demand remains at 1/5, while Credit Stress and Valuation Expectations remain at 3/5.

Together, the eight readings produce the current composite of 3.38/5 — Elevated Risk, below BoomRisk’s 4.00 threshold for High Risk.

What would make the risk fall?

BoomRisk is not structurally designed to become more bearish over time.

The reading could fall if the evidence improves.

Free cash flow could recover relative to the capital being invested.

Observable monetisation could begin closing the gap with infrastructure spending.

Strong semiconductor demand could translate into sustained utilisation and profitable capacity throughout the wider ecosystem.

Market concentration could decline.

Credit conditions could remain benign even as the buildout expands.

If those developments occur, they should flow through the individual pressure tests and ultimately the composite score.

What would make the risk rise?

The opposite developments matter too.

Semiconductor demand could weaken after major infrastructure commitments have already been made.

The gap between investment and observable monetisation could remain wide or increase.

Free-cash-flow pressure could deepen.

Credit conditions could deteriorate.

Or increasingly capital-intensive financing structures could become more dependent on future utilisation assumptions that ultimately prove too optimistic.

Those are observable developments.

BoomRisk is designed to monitor them rather than predict when — or whether — they will occur.

The distinction that matters

NVIDIA’s latest results strengthen the evidence that demand for leading AI compute remains exceptionally strong.

They do not answer a different question:

Will every layer of investment being built around that demand ultimately earn an adequate financial return?

Technological success, demand growth and investment returns are related.

They are not the same thing.

Right now, BoomRisk’s evidence supports two conclusions at once:

AI demand is holding.

The economics around it remain under pressure.

BoomRisk is a financial-risk monitoring framework, not an investment recommendation or market forecast.