The economy is running on GPUs now, and the bill is coming due in ways most people haven't connected yet.
The Summary
- AI hardware drove August import price increases, continuing a year-long trend that's reshaping trade flows and input costs across industries
- Two senior economists independently cite AI investment as "critical" and a "huge factor" in current US economic performance, rare consensus from different political camps
- The data center construction boom is real economic activity, but it's colliding with commodity volatility from geopolitical conflict, squeezing margins on both sides
The Signal
We're watching AI capex become a legitimate pillar of the US economy in real time. When Torsten Slok at Apollo and Stephen Moore, former Trump economic adviser, both use superlatives about the same trend within 48 hours, that's signal. These aren't conference panel takes. These are the people briefing institutional allocators and policymakers.
The mechanism is straightforward: AI hardware imports are driving measurable increases in trade volume. Data center construction is booming. That means jobs, steel, concrete, and power infrastructure, all showing up in GDP. But it's not just gross numbers that matter.
"AI hardware continued to drive August's increase in import prices, as it has throughout the year."
The cost structure is shifting underneath everything. Companies face higher input costs tied to the AI buildout, which means margins compress or prices rise. At the same time, commodity-price volatility from the Iran war is hitting industrial materials. Energy prices are climbing, and as Moore notes, when energy goes up, everything else follows.
Here's what that means in practice:
- Nvidia chips cost more to import, raising baseline AI infrastructure costs
- Building a data center costs more because steel and diesel are up
- Running that data center costs more because power is expensive
This is the hidden tax on the agent economy. Every company racing to deploy AI is paying a premium that didn't exist 18 months ago. And it's not temporary. Slok was discussing the Fed's rate decision in the context of AI investment, which tells you central bankers are now factoring AI capex into monetary policy. That's a new variable in the economic model.
The irony: AI spending is propping up growth numbers while simultaneously making AI more expensive to deploy. Winners are the companies already scaled. Losers are anyone trying to catch up on a budget. The gap between haves and have-nots in AI capability is widening because the cost floor is rising.
The Implication
Watch for a bifurcation in how companies talk about AI investment in Q3 and Q4 earnings. The hyperscalers will keep spending because they have to. Everyone else will start showing strain. If you're building with AI, your 2027 budget just got harder to justify. If you're investing, the returns to scale in this market are steeper than most models assume.
Moore says he's bullish despite high energy prices. That's only true if you believe AI productivity gains materialize fast enough to offset rising input costs. The clock is ticking on that bet.