The $1 Trillion Invoice: Who's Paying for the AI Build-Out?

June 2, 2026 · 6 min read

That number isn't hypothetical. It comes from Josh Bersin, the enterprise industry's most respected analyst. He ran the math on hyperscaler spending — Amazon, Google, Microsoft, Meta — plus Nvidia, TSMC, and the rest of the semiconductor chain.

"No matter how you look at it, someone (that's us) is going to pay twice as much for enterprise software or twice as much for ads for this investment to pay off."
— Josh Bersin

The bill for the biggest infrastructure build in human history is arriving. And it's not addressed to Google or Microsoft. It's addressed to you.

The Math

Here's what Bersin documented:

• $410 billion — Big 4 hyperscaler AI spending in 2025 (Bridgewater/Reuters)
• $650 billion — Projected 2026 run-rate
• $1 trillion — Add Nvidia, TSMC, and the supply chain
• $6.3 trillion — Gartner's 2030 projection

For investors to earn a 15% return on this spending (assuming generous 5-year depreciation), the industry needs to generate at least $1 trillion per year in new revenue. Most likely more.

The Evidence It's Already Happening

Uber burned through their complete 2026 AI budget by April. Zero noticeable productivity increase. They paid for the whole year. They got four months.

Source: Forbes, Mizzima, Josh Bersin (June 2026)

The Price-Performance Gap

Here's the part that proves it's not about infrastructure costs:

"Gemini 3.5 Flash, just announced this week, is supposedly 10-times less expensive than Opus 4.7, so the battle for price-performance now officially begins."
— Josh Bersin (UPDATE)

If AI pricing reflected actual costs, two comparable models couldn't have a 10x price difference. The difference isn't compute. The difference is Rent.

One vendor chooses to extract. Another competes on price. Same technology. Same infrastructure. Ten times the cost.

What This Means for You

The $1 trillion invoice isn't being split equally. It's being distributed based on how locked in you are.

• Not locked in? You can shop around. Switch models. Use the cheaper option.
• Locked in? You pay whatever they charge. Token by token. Month by month. Surprise bill by surprise bill.

The enterprises being sued for $100M? Locked in. The CIOs outsourcing to India? Locked in. Uber's burned budget? Locked in.

The Alternative

You can own your AI tools the same way you own your phone, your laptop, your car.

• One payment, not a meter running in the background
• Your data, not a vendor's training dataset
• Your workflow, not a subscription they can change tomorrow

The Bottom Line

$1 trillion has to come from somewhere. Bersin did the math. The vendors are doing the extracting. The only question is whether you're paying by choice — or by trap.

13 days until Anthropic's June 15 credit split. If you're on a metered plan, the next invoice is already being calculated.

— Frontier Flat Rate