The Numbers Are Real. So Is the Vulnerability.

AI is no longer a side bet. It's the main event.

According to Breitbart's business analysis, investment in information-processing equipment and software averaged 0.3 percentage points of annual GDP contribution for over a decade before 2025. In the first half of 2025, that same category accounted for nearly all of the economy's growth. Strip out AI-related investment and the U.S. economy was barely treading water.

This isn't about stock prices or Silicon Valley hype cycles. AI spending is now showing up in construction permits, factory orders, power equipment procurement, cooling systems, networking infrastructure, and professional services hiring. Real buildings. Real workers. Real dollars.

What a Real Investment Boom Looks Like

Breitbart's analysis makes a historically grounded bull case. The U.S. went through this before — railroads, electrification, the automobile, computing. Each one looked like a speculative frenzy in the early stages. Each one required massive upfront infrastructure investment before productivity gains arrived, sometimes a decade later. And each one ultimately made America richer, not poorer.

The argument is that AI is the same kind of capital-deepening cycle. Businesses are already reporting measurable productivity gains in coding, research, customer service, and workflow automation. These aren't projections anymore — they're showing up in firm-level data. If model capabilities keep compounding and utilization rates climb, today's infrastructure buildout could theoretically support 2 to 3 percent structural GDP growth for years.

The financing loop, as described by Breitbart's digest, runs like this: capital flows to hyperscalers, hyperscalers build data centers, data centers generate cloud revenue, cloud margins justify the capital expenditure, and the cycle sustains itself as long as marginal returns beat the cost of capital. The operating margins of major cloud providers — Microsoft, Amazon, Google — suggest that condition still holds. For now.

The Risk Nobody Wants to Say Out Loud

The problem is unavoidable: the same facts that make the bull case compelling also make the downside catastrophic.

When one category of investment is doing the heavy lifting for an entire national economy, you don't have diversified growth. You have a single point of failure.

Bloomberg's paywalled coverage flagged energy costs as an emerging profitability threat for AI infrastructure. Data centers are power-hungry on a scale that makes previous tech infrastructure look quaint. Building out gigawatt-scale compute capacity requires electricity that the U.S. grid is not currently engineered to deliver at the speed the industry demands. That means delays, cost overruns, or both.

Energy constraints could compress hyperscaler margins. Compressed margins could slow capital expenditure. Slower capex could hit the construction, manufacturing, and services sectors that have been riding the AI wave. The chain reaction runs in both directions.

Coverage Gap

Left-leaning outlets are largely covering this story through a redistribution lens — who benefits, who gets left out, whether AI is "colonial" in some framing. Right-leaning coverage tends toward unqualified boosterism — AI as American triumph, full stop.

The understated angle is concentration risk at the macroeconomic level. This isn't about any one company or one sector. It's about whether a single investment theme can sustain a $28 trillion economy through a period of otherwise sluggish demand. History says those bets eventually resolve — but the resolution isn't always gentle.

The Energy Problem Is Structural

Bloomberg's reporting on energy costs points to something the AI cheerleaders consistently wave away. Training and running large language models at commercial scale consumes enormous amounts of power. Microsoft, Google, and Amazon have all announced aggressive data center expansion plans running into the hundreds of billions of dollars over the next several years. Every one of those facilities needs reliable, cheap electricity.

The U.S. power grid wasn't built for this. Permitting alone for new generation capacity can take years. Nuclear, which AI companies have been quietly courting as a solution, takes even longer. The productivity gains from AI are supposed to arrive before the energy infrastructure catches up. If they don't, the math stops working.

What This Means for Regular People

If you work in construction, manufacturing, electrical contracting, or professional services, the AI boom has been genuinely good for employment and wages. That's real.

But if this investment cycle hits a wall — energy constraints, margin compression, a plateau in model capabilities, or simply a shift in capital allocation — the sectors that rode the wave up will feel the correction. The people least insulated from that are the workers who built the infrastructure, not the executives who ordered it.

The question isn't whether AI is a transformative technology. It clearly is. The question is whether the investment is arriving faster than the productivity payoff, and whether the U.S. economy has enough other growth engines to absorb the shock if the timing is off.

Right now, the honest answer is: not obviously, no.