The AI demand signal was $110 billion. Supply was all you could see.
For two years, the AI economy ran without a demand-side number. That changes today.
Half of CEOs told BCG their jobs depend on getting AI right. They made that bet without a demand-side number. Now the number exists.
"The eye sees only what the mind is prepared to comprehend." — Robertson Davies
THE SIGNAL
Exponential View published the first bottom-up, deduplicated measure of AI demand on June 25, 2026.
When the same dollar moves through multiple companies in a supply chain, it can get counted at each step and inflate the total. Deduplicated means they stripped that out and counted each dollar only once, at the point where the end customer actually spent it.
The number: $110 billion in revenue generated by the AI economy over the trailing 12 months.
Annualized run rate: $175 billion.
Growth rate roughly three times faster than the mobile or internet waves at comparable stages.
The supply side has been visible for years.
GPU shipments.
Hyperscaler CapEx.
Chip orders.
What did not exist until this report was the demand side. Actual dollars flowing from end customers to AI providers, stripped of double-counting across the supply chain.
Executives making AI investment decisions have been doing so without that number. Not because it was hidden. Because the instrument to surface it did not exist.
THE FAILURE POINT
The failure is not in the data. It is in what organizations did in its absence.
Half of CEOs surveyed by BCG reported they believe their jobs depend on getting AI right. That is a threat-response signal.
Leaders under existential pressure do not wait for confirmation. They act on what is visible:
Competitor moves.
Analyst projections.
Board expectations.
The supply side was visible. So the supply side became the proxy for demand. Billions in capital commitments were made on the wrong signal.
Not a technology problem. A decision architecture problem. When the real signal cannot travel, leaders substitute what can. They make the best decision available on distorted input.
SIGNAL WITHIN THE SIGNAL
Signal Compression, applied at economy scale.
The demand-side reality existed. $110 billion in actual customer spending.
But it existed distributed across hundreds of private companies with no disclosure obligation.
It could not aggregate.
It could not travel.
What reached the decision level was the supply side, CapEx filings, chip revenue, infrastructure buildout, which was measurable but was not the signal that mattered.
Every organizational failure this framework has documented has this as a precondition: leadership makes a decision on a version of reality that is not false, just incomplete beyond utility.
The AI investment cycle of 2024 and 2025 is the largest instance of this pattern ever recorded.
BEHAVIOR UNDER PRESSURE
Leaders narrowed to what they could see and treated it as complete. The supply-side signal was loud, specific, and institutional. Nvidia’s revenue. Amazon’s CapEx guidance. Microsoft’s infrastructure commitments. Real numbers from public companies. They created the perception of a confirmed market.
The pattern is predictable. Under pressure to position correctly, with an incomplete picture, leaders default to the frame that is most visible and most socially reinforced.
In boardrooms across 2024 and 2025, that frame was: look at what is being built. The question that did not get asked with equal force: look at what is being bought, by whom, at what rate, covering what portion of the build cost.
That question now has an answer. The window for acting on the old frame is closing.
SYSTEM DRIVER - MOS
AI investment review processes must now incorporate demand-side metrics alongside supply-side indicators.
The Exponential View methodology, a deduplicated, bottom-up model tracking end-customer spend across public statements, reported figures with confidence weighting, and internal consistency checks, is the template.
One structural fix: any capital allocation proposal for AI infrastructure, tooling, or capability that does not include a demand-side signal reference does not clear the review gate.
Not a supply-side proxy. Actual revenue data from the AI economy. That discipline did not exist because the data did not exist. The data now exists.
LEADER DRIVER - INTERNAL OPERATING SYSTEM (IOS) - REGULATE
The leader who committed to AI because the board expected it, because competitors were moving, because the supply-side build made it feel inevitable, was operating under Internal Signal Compression.
The threat response collapsed the decision into binary: move or fall behind. The calibration question did not fit in that space.
The regulated version of that same decision asks:
What is the demand evidence?
What is the revenue coverage ratio?
What am I committing to, and on what signal?
Those questions were harder to answer in 2024. They are answerable today.
IF YOU DO ONE THING TODAY
Pull your organization’s AI investment plan.
Identify which assumptions about market size, adoption rate, or competitive necessity were made without demand-side revenue data.
Before the next capital commitment, validate those assumptions against the $110 billion baseline and the $175 billion annualized run rate now on record.
If the plan was built on supply-side signals alone, it needs a second pass.
That pass is now possible. Run it.
PRESSURE / REGULATE
Pressure: 8.0 / Regulation: 5.5
Gap Index: 1.45 (Instability Forming, direction closing now that demand signal exists)
Pressure source: Companies poured money into AI infrastructure before anyone had confirmed customers were actually spending at that scale. The gap starts closing now that the demand number exists and investment decisions can be built around it.
FINAL SIGNAL
The market was always real. The signal just could not travel until someone built the instrument to carry it.
CTA
Subscribe to The Tempered Signal.
Send this to the leader who approved AI spend without a demand-side number to stand on.
SOURCES
Azhar, A., Gildea, W., Warren, N. et al. “The State of the AI Economy.” Exponential View, June 25, 2026. exponentialview.co. BCG CEO AI Survey, 2026. McKinsey State of AI Report, 2026.
The IOS framework this edition draws on is developed fully in Regulate.
WHAT THE TEMPERED SIGNAL REVIEWS
The Tempered Signal doesn't cover the news. It finds where the news is hiding the decision.