The global economy is currently witnessing a strange mathematical tug-of-war. On one hand, we see staggering investment figures; on the other, the official GDP charts seem almost indifferent. Understanding this requires reconciling two seemingly contradictory perspectives from the world’s leading economists.
At first glance, the numbers seem contradictory. American companies are pouring hundreds of billions of dollars into artificial intelligence infrastructure (chips, data centers, software platforms) at a pace rarely seen in economic history.
Five major corporations (Amazon, Microsoft, Google, Meta, and Apple) poured roughly $600 billion into AI infrastructure in 2026 (Investing.com, Goldman Sachs).
Yet official GDP figures remain stubbornly unimpressed. Growth is positive, but modest. The productivity revolution that was supposed to follow the AI wave has not yet materialized in any measurable way.
This disconnect has puzzled analysts, policymakers, and investors alike. Is AI investment simply failing to deliver? Or are the tools we use to measure economic output poorly equipped to capture what is actually happening?
Two of the most respected economists in the United States have each offered a piece of the puzzle. We will see how their analyses paint a picture that is both reassuring and alarming.
Jan Hatzius’ Zero Impact
In the Atlantic Council video (January 2026), Goldman Sachs Chief Economist Jan Hatzius explains that the measured impact of AI on US GDP in 2025 was basically zero due to import accounting.
Import Leakage
A large portion of AI investment (approximately US$450-700 billion in capex) goes towards purchasing hardware (semiconductors from NVIDIA/TSMC, for example). In GDP calculations, this investment is counted positively, but since the equipment is mostly produced in Taiwan or South Korea, it is recorded as a negative import.
Net Result
The net exports’ negative effect offsets the investment’s positive effect. According to Hatzius, this spending “helps the GDP of Taiwan and Korea, but not much that of the US.”
Technical Classification
He also mentions that part of the spending on semiconductors is classified as “intermediate inputs” and not as “final investment” in national accounts, which means it doesn’t directly appear in GDP growth.
As a result of this classification, this specific expenditure ends up not being accounted for in the United States’ GDP, even though it is a significant expense for the sector.
Jason Furman’s 92% of Growth
Jason Furman, the Aetna Professor of the Practice of Economic Policy at the Harvard Kennedy School (HKS), contends that economic growth concentrates in a narrow set of technological sectors.
Information Processing Equipment and Software
Furman points out that, in the first half of 2025, investment in these categories accounted for almost all marginal GDP growth (4%). However, he argues, if you remove these components, the rest of the American economy grew by only 0.1%.
In other words, without the technology spending “boom” (driven by AI), the US would have stagnated or entered a recession.
Focus on Gross Investment
While Hatzius looks at net value added (excluding imports), Furman analyzes spending dynamics. For him, the fact that companies are spending aggressively on AI makes it carry the growth on its back.
How to Understand and Conclude Both
Hatzius says AI is not yet generating net domestic growth, while Furman says AI is the only reason why total growth is not negative. Both agree that the real impact on productivity (producing more with less) has not yet arrived. Goldman Sachs predicts that this productivity leap will only be measurable in GDP from 2027.
To better visualize the dynamics described by Furman and Hatzius, imagine the economy as a car engine with multiple cylinders.
Traditional cylinders (Consumption, Real Estate, Manufacturing) are failing or idling due to high interest rates and cooling consumer demand. On the other hand, the AI/Tech cylinder is operating at maximum RPM, which compensates for the others.
Short Term (Hatzius)
On paper, GDP isn’t rising much right now because we’re just exchanging money for imported machines. The real impact on American production is still negligible.
If an American company spends $10 billion on NVIDIA chips produced in Taiwan, that money is removed from the US GDP calculation as an import. The investment happens, but the accounting benefit leaks out of the country.
What we are seeing now is the CapEx (capital expenditure) phase. The American GDP will grow when this investment translates into productivity.
Moreover, the capital that companies are injecting into GPUs and AI infrastructure is often being withdrawn from other areas (such as marketing, hiring, or physical expansion).
Economic Dynamics (Furman)
AI is the only engine that’s running strong. Without this flow of capital into technology, the economy would be at a standstill, regardless of where the machines are manufactured.
Jason Furman’s point is the most alarming: if you subtract the AI frenzy, the indicators suggest that the rest of the American economy would stagnate or enter a recession.
The AI Tug-of-War: A Summary
The contrast between Hatzius’ “accounting zero” and Furman’s “solo engine” is the perfect lens through which to view 2026. Here is a brief synthesis of the friction between national accounting (GDP) and market dynamics (CapEx).
| Perspective | Economist | Core Argument | Metric | Economic Sentiment |
| The Accounting Reality | Jan Hatzius | Import Leakage: US imports the physical guts of AI (chips); the net GDP impact is neutral. | Impact Zero on 2025 GDP | Cautious / Patient |
| The Structural Reality | Jason Furman | Solo Engine: Remove tech/software investment, and the rest of the economy is essentially flatlining | 92% of the 2025 GDP | Alarmist / Realistic |
Conclusion
AI is not creating an economic miracle of +5% growth per year, but it creates a safety net. It’s stopping the American economy from decelerating, keeping growth in around 2%-2.5%, while the tech market tries to figure out how to transform $600 billion in hardware into real profit and operational efficiency.
The big test will be in 2027: if CapEx decreases and productivity hasn’t yet increased, the gap left by AI in GDP will be brutal.
As companies divert massive amounts of capital toward AI infrastructure, having AI professionals is a survival requirement. After all, this is where the bridge between technology and human capital becomes critical.
To achieve this goal, many firms turn to DistantJob to source high-tier global AI developers. By leveraging remote expertise, companies can keep their momentum in the AI Era while saving costs.
DistantJob fits perfectly with maintaining growth while optimizing headcount costs. Schedule a meeting with us today, at no cost, to discuss your ideal AI expert.
