AI Is Already Showing Up in the Hard Data
Is U.S. productivity entering its 1996 moment?
Most discussion around AI is still showing up in surveys, sentiment, and earnings calls. That’s not where technological shifts show up first.
The more relevant question is whether this is already observable in the data, rather than waiting for it to appear in lagging indicators.
The 1996 Parallel
The late-1990s productivity boom didn’t look obvious at the start. Computers had been around for decades. The internet wasn’t new. But around 1995–1996, productivity growth began to accelerate meaningfully as firms learned how to reorganize around new technologies. It wasn’t the invention that mattered most. It was the deployment.
In a recent discussion with Cliff Waldman and Dan Bachman, we revisited this period and how long it took markets and policymakers to recognize the scale of the shift.
That distinction matters today, because AI appears to be following a similar adoption curve.
The Productivity Data Is Turning
The chart below tells the story.
After more than a decade of post-Global Financial Crisis stagnation, U.S. productivity growth is beginning to re-accelerate. The 10-year moving average has turned higher for the first time since the early 2000s recovery.
We’re still well below the peaks of the late 1990s, but the trend has clearly shifted and the data is beginning to reflect it.
Why AI’s Impact Appears First in Physical Data
When economists try to measure technological change, they often look in the wrong places:
Business surveys
Hiring intentions
Capex guidance
Earnings commentary
But technological revolutions show up first in deployment infrastructure, not opinions.
This is where observational data becomes powerful.
From space, through geospatial measurement, we can already see AI’s footprint expanding in real time.
Deployment > Invention
This brings us back to the 1996 analogy.
The late-90s productivity surge didn’t happen when computers were invented.
It happened when:
Firms reorganized workflows
Supply chains digitized
Communication costs collapsed
Information became real time
AI is entering that same deployment phase today.
We’re moving from experimentation to integration.
Why This Matters for GDP
As I stated in our interview with Dan and Cliff, productivity (particularly Total Factor Productivity or “TFP”) is the single most important long-run driver of economic growth.
If AI is lifting output per worker, several macro implications follow:
Higher potential GDP
Stronger corporate margins
Lower unit labor costs
Delayed disinflation in services
Higher neutral rates (r*)
The risk is not missing the level, but misreading the timing, as productivity gains tend to appear with a lag in official data.
The Observational Economy vs The Survey Economy
Traditional macro measurement is lagged by design. Productivity data itself is revised for years. By contrast, the physical economy updates daily.
By observing:
Construction activity
Energy utilization
Industrial throughput
Logistics flows
We can identify early signs of technological deployment in real time.
Including increases in power demand and data center construction are already visible in key regions, which historically precede shifts in reported output and productivity.
More on measuring construction in near real-time from outer space here.
Are We in 1996?
It’s too early to declare a full productivity renaissance.
But the ingredients are familiar:
Breakthrough general-purpose technology
Massive infrastructure buildout
Corporate workflow reorganization
Capital deepening cycles
The late-90s boom didn’t look obvious at the start either.
It looked like a slow turn higher in the data.
Exactly like today.
Closing Thought
AI’s economic impact won’t arrive all at once.
It will compound quietly through:
Compute infrastructure
Energy demand
Industrial production
Logistics optimization
Long before it shows up cleanly in GDP releases.
The productivity story of the next decade may already be underway.
By the time this shows up clearly in GDP data, the adjustment will already be underway.
Interested?
If you’re interested in how satellite and geospatial data can track AI infrastructure, industrial activity, and productivity shifts in real time, feel free to reach out or subscribe below.