We are now 4 years into the AI boom
A productivity is boom typically this:
Phase 1: The Investment Phase (The Deficit)
When a country starts its productivity boom, it cannot instantly produce everything it needs. It must import the "building blocks" of productivity first.
- In Civil War US Example: In the late 1860s and 1870s, the US ran trade deficits because it was importing massive amounts of British steel, locomotives, and capital to build the transcontinental railroads.
- In 1980s China Example: China heavily imported advanced manufacturing machinery from Japan and Germany because its domestic factories couldn't build those machines yet.
Phase 2: The Output Phase (The Surplus)
This is the stage where the imports pay off and the trade balance flips.
- The US Result: Once the railroads were built, shipping costs plummeted. The US began exporting massive amounts of wheat, cotton, and eventually, its own manufactured machinery, turning into a trade surplus powerhouse.
- The China Result: Once those German and Japanese machines were bolted to the factory floors, China’s productivity skyrocketed. They flooded the world with shoes, textiles, and electronics, creating the massive trade surplus we see today.
In my (classical economics) book, if a country’s total production (Goods + Services) does not exceed its total consumption, or worse goes negative because of that boom, calling it a "productivity boom" is a misnomer. I waited a bit before coming to that conclusion because it is normal to have a lag, but after 4 years?
A genuine productivity boom must manifest as a physical or structural surplus of output. If the math shows you are importing more than you export, you are simply on a credit-fueled consumption binge.
The Accounting Truth: GDP vs. Spending
In national accounting, the math is absolute:
If the trade balance is negative, Spending > Production.
By definition, if a country is spending more than it produces, it is making up the difference by selling off its assets (stocks, real estate, government bonds) to foreigners.
If a nation relies on foreigners to supply its material goods while it hands over financial deeds and IOUs, it is not solving its structural imbalances, it is making them worse with that “boom”
A country that only boomed in services and asset prices faces a massive structural correction.
I was cautious in calling this AI boom, but after 4 years I think we have yet another consumption boom with the only result NOT an increase in sales of goods and services abroad BUT exporting financial assets to finance that consumption boom.
That consumption boom can be from both corporations or consumers, what I mean by “consumption” boom is that there are tons of expenditures but not the boom in output that flips the trade imbalance. A boom in “Inputs” only.
After 4 years I am worried that a large use of chatGPT is to advise the top 10% shoppers (who surf on that financial asset boom to fuel their consumption) on what gravy goes best with roast duck and what wine to buy.
The Trade deficit induced by AI: or said otherwise the reduced production versus consumption induced by this “PRODUCTIVITY” BOOM (lol)
Between 2023 and 2025, the U.S. overall goods trade deficit climbed to a record $1.2 trillion, and economic data directly blames the insatiable demand for AI hardware. In fact, research by the Federal Reserve Bank of Minneapolis notes that without the post-2023 surge in AI imports, the U.S. trade deficit would have been roughly 16% smaller ($194 billion lower).
The specific detail of how this "AI deficit" broke down across hardware, chips, and infrastructure shows a highly import-intensive dynamic:
1. The Numbers: AI vs. Non-AI Imports
While the rest of the economy remained relatively flat, AI-related trade exploded:
- The AI Surge: Total U.S. dollar imports of AI-relevant products hit $379 billion in 2025—a massive 72.6% increase from 2023.
- The Non-AI Baseline: Over the exact same period, imports of products with low AI relevance grew by a meager 2.5%.
- The Trade Deficit Gap: In 2025 alone, the U.S. imported $265 billion in AI-specific goods but only exported $71 billion, creating a massive net drag on the balance of trade.
2. Breakdown of What the U.S. is Importing
Data center buildouts require heavy physical components, which are split across two core supply chains: [1]
- The Computing Stack ($580 billion total by 2025): This includes the servers, high-performance GPU clusters, advanced networking switches, optical cables, and liquid cooling systems needed to keep AI models running. Imports of "automatic data processing machines" (like servers) alone drove roughly three-quarters of all U.S. merchandise import growth.
- The Power Stack ($70+ billion total by 2025): AI data centers consume vast amounts of electricity. The U.S. has been forced to massively increase imports of electrical infrastructure, including high-capacity power transformers ($30 billion), switchgear ($22 billion), and lithium-ion battery storage ($20 billion).
- Advanced Semiconductors: While U.S. tech giants design the chips, the physical manufacturing happens abroad. The U.S. semiconductor trade deficit remains heavily tied to packaging and fabrication hubs in Taiwan, Malaysia, and South Korea. U.S. imports of computer hardware and semiconductors surged 60% in a single 12-month period leading into early 2026.
3. The Tariff Loophole
What makes this widening deficit unique is that it bypassed aggressive national trade barriers. While the effective tariff rate on standard non-AI imports climbed to 12.1%, the effective tariff rate on AI-relevant goods sat at just 4.5%. Tech companies successfully used product-level exemptions for consumer electronics and advanced components, allowing 69% of AI-related imports to slip past tariff walls completely.
Now we are in 2026 and there is no sign that this trade imbalance is really compensated by any meaningful export of API ... And after 4 years this productivity boom story becomes really suspicious.
Four Years into the boom, where do we stand in output aka productivity?
By mid-2026, the data heavily supports my skepticism. The promised transition from the Investment Phase to the Output Phase has stalled, and the economic math is exposing this "productivity boom" as a structural consumption and asset-export engine.
This is the current 2026 economic landscape:
1. The API Export Myth
The core defense of the AI productivity boom was that the U.S. would export high-margin, intangible assets—like AI software, enterprise APIs, and cloud computing models—to balance out the physical hardware imports.
In reality, the net export of digital services has barely budged.
- The Domestic Bottleneck: Major AI models are primarily being consumed domestically by U.S. corporations trying to automate internal workflows, or by American consumers using AI apps.
- Geopolitical & Regulatory Walls: Europe's strict AI regulations and China's total ban on Western AI models have severely limited the global market for American AI APIs. The U.S. is importing physical, hardware-heavy capital from Asia and Mexico, but it cannot export the digital software back to them at anywhere near the same scale.
2. The CHIPS Act Delays (The Missing Domestic Supply)
The plan to correct the hardware deficit was to build domestic advanced chip foundries (fabs) via the CHIPS Act. However, by 2026, these multi-billion-dollar projects have faced severe delays:
- The Production Gap: Major advanced fabs in Arizona and Ohio have pushed their full commercial production timelines further out due to labor shortages, environmental permits, and supply chain bottlenecks for specialized equipment.
- Continued Reliance: As a result, the U.S. remains just as dependent on importing advanced GPUs and server infrastructure from Taiwan, South Korea, and Southeast Asian packaging hubs as it was in 2023.
3. The Asset-Export Reality Check
Because the U.S. is not producing enough physical goods or global digital services to cover its massive AI infrastructure bills, it is balancing the books exactly the way was described earlier: by exporting financial assets.
To fund the insatiable demand for data centers and energy grid upgrades, U.S. tech giants and utilities are issuing massive amounts of corporate debt and equity. Foreign capital is rushing in to buy these assets, attracted by high U.S. interest rates and tech stock valuations.
- The Result: The U.S. is trading ownership of its financial future (stocks, bonds, and real estate) to pay for depreciating physical hardware (servers that become obsolete every 3 to 4 years).
Conclusion: A Consumption Boom in Disguise
After nearly four years, the distinction holds up: this looks much more like an infrastructure consumption boom than a national productivity boom.
Individual workers might be writing code or emails faster using AI, but at a national accounting level, the country is spending EVEN more than it produces. The trade deficit has not been resolved as it should have been with a productivity boom.
Remember, a productivity boom means that all of the sudden you produce a lot more than you consume, and that should result in a shrinking trade imbalance.
AI does not make imbalances better, it makes them worse.:
Now you wonder why the US needs to steal the share of production from the GCC countries and sell that $Oil at a higher price. Lol
Now My Erudite Conclusion
This s*** is going to blow up.