Once a decade, a single number comes along that quietly reshapes the wealth landscape. In 2026, that number is $725 billion.

That’s the combined 2026 capital expenditure plan from Microsoft, Amazon, Alphabet, and Meta — four companies that, less than five years ago, were spending a fraction of that. Their 2026 budgets represent a roughly 77% increase over their record 2025 spending. None of this money is going into stock buybacks, dividends, or marketing. It’s almost all going into AI: GPUs, data centers, networking, power, cooling, the actual physical infrastructure that will run the next generation of models.

And next Wednesday, May 20, NVIDIA will report earnings that will turn that $725B from an abstract plan into a concrete revenue number on a P&L.

If you understand what this wave actually is — and what it isn’t — it changes how you should think about your portfolio, your career, and your own wealth trajectory for the rest of this decade.

Capex isn’t bubble talk. It’s the strongest signal in markets.

Most of the debate around AI right now is qualitative. Is it a bubble? Are valuations stretched? Will AI deliver on its promises?

Capex cuts through that noise. Companies of Microsoft’s and Amazon’s scale do not commit hundreds of billions of dollars to physical infrastructure on vibes. They commit it when their internal demand signals — backlog, contract pipeline, customer commitments, internal usage growth — make the spending mathematically defensible.

Microsoft’s AI revenue run-rate is now around $37 billion, growing 123% year over year. That isn’t a science project. That’s a business large enough to make the entire AI-skeptic case look quaint. Amazon’s AWS Bedrock business is in a similar place. Google’s TPU build-out is being driven by both first-party Gemini demand and Cloud customers. Meta is spending because it sees AI-driven engagement and ad targeting moving the core P&L.

When four hyperscalers, all with different motivations, independently arrive at “spend more than the entire GDP of Switzerland on AI in a single year” — that’s not a marketing cycle. That’s an industrial buildout. It’s worth remembering that even our analysis of sticky inflation identified this AI capex as a stealth stimulus that the Fed cannot easily offset.

Why NVIDIA’s May 20 number matters more than any prior quarter

NVIDIA is the chokepoint. Roughly 90% of the AI chips going into those $725B worth of data centers will be NVIDIA’s. Which means NVIDIA’s quarterly revenue is, in effect, the most direct measurement of how that capex is converting into real spending.

Wall Street is looking for $78.8 billion in revenue and $1.77 in EPS for the quarter. Management’s own prior guidance already implied roughly that, excluding China. The print itself isn’t the interesting part. What matters is three other things:

Forward guidance. If NVIDIA guides Q2 revenue above current consensus ($82B-ish), that confirms the $725B capex number is being converted into real orders, not vaporware roadmaps. If they guide below, the $725B figure starts looking aspirational.
Gross margin trajectory. NVIDIA’s data center gross margin has been in the mid-70s. The Blackwell ramp could see that compress slightly. Anything above 73% is a green light for the bull thesis.
Sovereign AI commentary. A growing share of NVIDIA’s order book is now non-hyperscaler — UAE, Saudi Arabia, Japan, France, South Korea, and an emerging set of country-level AI buildouts. If this segment is growing faster than the hyperscaler segment, the runway extends well beyond the current four-company picture.

Jensen Huang has already publicly framed this: he expects at least $1 trillion of opportunity tied to Blackwell and Rubin systems through 2027. The earnings on May 20 will either narrow or widen the band around that claim.

The wealth implications nobody is making explicit

Here is where most coverage stops. “NVIDIA earnings important. AI capex big. Maybe buy NVIDIA.”

That’s the surface read. The deeper read — the one that actually matters for your wealth — is that the $725B AI capex wave is creating second-order and third-order wealth effects that most retail investors are still completely ignoring:

Power and energy infrastructure. Each gigawatt of AI data center demand is a multi-billion-dollar capex event for the local utility. Independent power producers, nuclear, natural gas pipelines, and grid equipment are all riding the same wave. Many of these names have moved a lot, but several have not.

Cooling and industrial real estate. Hyperscaler data centers consume enormous amounts of water and need specialized industrial space. The publicly traded cooling and data-center REITs are direct beneficiaries.

Labor and skills repricing. This is the wealth angle most retail conversations miss entirely. A senior ML engineer salary has gone from $400K total comp to $1M-plus at top labs in roughly 36 months. AI-adjacent product, GTM, and sales roles are seeing 30-60% premiums versus equivalent non-AI roles. If you have any technical background at all, the capex wave is also a personal-income wave.

The 99% of small companies that will use this infrastructure. OpenAI, Anthropic, and the hyperscalers get the headlines. But the real wealth creation cycle from prior tech buildouts — railroads, electricity, the internet — happened on top of the infrastructure, not in it. The 2026 equivalents of “the businesses built on top of cheap electricity” will be the meaningful wealth stories of the next 10 years. Many of them will be tiny businesses run by individuals using AI tools that didn’t exist 24 months ago.

What this means for the average reader

If you take the $725B capex number seriously, three things should change about how you operate:

1. Have explicit AI exposure in your portfolio. Not “I own an S&P 500 fund and it has Microsoft in it.” Real exposure: AI infrastructure names, power/utility plays that are clearly tied to data center demand, and a position in the chokepoint chip names. None of this needs to be a moonshot bet. A combined 10-15% AI-tilt sleeve on top of a diversified core portfolio is enough to meaningfully participate.

2. Treat your career as the highest-leverage AI play you own. If you can move into an AI-fluent role at your current employer — or start using AI tools to make yourself measurably more productive — the personal-income upside dwarfs anything you’d realistically generate in your brokerage account. The capex wave is paying engineers, salespeople, and operators who can leverage these tools today.

3. Watch May 20 for the second derivative. The earnings number itself will be in line. What matters is whether the forward picture is widening or narrowing. If guidance is strong and sovereign commentary is heavy — the runway extends and AI exposure stays a priority. If guidance disappoints meaningfully — the regime hasn’t broken, but the pace of the wave has, and that’s worth knowing.

The bigger picture

It’s easy to read this as a “buy NVIDIA” essay. It isn’t.

The point is bigger: $725 billion of corporate spending in a single industry, in a single year, with credible 70%+ growth, is one of the rarest macroeconomic events of our lifetimes. Comparable historical analogs — the railroad build-out of the 1880s, the post-WWII auto and highway buildout, the late 1990s internet capex cycle — all created enormous wealth, but the wealth was very unevenly distributed. The people who paid attention and positioned early did well. The people who waited for confirmation did not.

You don’t have to bet the farm. You just have to not be on the sidelines.

Want to actually track whether your portfolio is keeping up with this kind of wealth-creation cycle? The Wealtharian Wealth Tracker lets you log your net worth, AI exposure, and FU money progress month over month, so you can see in concrete numbers whether you’re participating in the wave or watching it from the shore. Try it free →