Two Trillion Dollars Evaporated Because Wall Street Finally Did the Math

Wall Street spent eighteen months telling everyone that AI would make every tech company more valuable. Then in early February, someone did the math and realized that if AI automates legal research, IT support, consulting, and logistics, the companies currently charging premium prices for those services are not “AI beneficiaries.” They are “AI casualties.” Two trillion dollars evaporated in 48 hours.

Oracle got hit the hardest. Larry Ellison watched $49 billion of his net worth disappear in two days, which is roughly 20 percent of everything he has. To put that in perspective, that is more money than the GDP of most countries vanishing because a spreadsheet nerd in midtown had a revelation about database wrappers.

The software sector dropped across the board. Not because anything went wrong. Not because earnings missed. Because investors suddenly processed what people in this industry have been saying for over a year: the companies with the fattest margins are the ones most exposed. When your entire product is moving data from one screen to another and charging enterprise clients a fortune for the privilege, and a language model can do the same thing for pennies, your moat is not “deep relationships with IT procurement.” Your moat is the time it takes your clients to figure out you are replaceable.

The DRAM Problem Nobody Wants to Talk About

While the stock market was having its existential crisis, a quieter disaster was unfolding in the supply chain.

DRAM prices are up 75 percent since December. Not because of tariffs. Not because of a factory fire. Because every AI company on earth is buying memory chips faster than Samsung and SK Hynix can manufacture them. Every Nvidia H100 cluster needs enormous amounts of high-bandwidth memory. Every new data center Google and Meta announce eats another chunk of global DRAM supply.

The result is that a low-end smartphone that used to spend 10 percent of its bill of materials on memory now spends closer to 30 percent. Tesla and Apple are both signaling production constraints. If you wanted to build a new consumer electronics product in 2026, the memory alone would blow your budget.

So here is the situation: the companies building AI infrastructure are consuming physical resources so aggressively that they are creating shortages in the components everyone else needs. AI is not just disrupting business models. It is physically consuming the supply chain. The compute buildout is so massive that it is crowding out the rest of the electronics industry.

The DeepSeek Crash Was a Warning Shot

January’s DeepSeek crash was about a specific threat — a Chinese lab proved you could build competitive models for a fraction of the cost, which meant the hundreds of billions being spent on compute might be overkill. That hit the chip stocks directly.

February’s crash is different. February is about the end users. It is the market finally internalizing that when AI disrupts everything, “everything” includes the companies in your portfolio. The January crash asked “are we spending too much building this?” The February crash asks “what happens to the economy we are building it for?”

These are not the same question, and the second one is much harder to answer.

Who Actually Benefits

Here is the uncomfortable truth that nobody in venture capital wants to hear: the primary beneficiaries of AI disruption are not companies. They are individuals.

A solo developer who used to need a team of five can now ship a product alone. A freelance lawyer who used to spend hours on document review can now do it in minutes. A small business owner who could not afford a marketing department can now generate everything they need for the cost of an API call.

The winners are not the ones selling AI. The winners are the ones using it to not need the companies that just lost two trillion dollars in market cap.

This is why the stock market correction is not irrational. It is the market correctly pricing the fact that AI makes large organizations less necessary. Every enterprise SaaS company that charges per seat is watching the number of seats shrink. Every consulting firm that bills by the hour is watching the hours disappear. The market is not wrong. The market was wrong before, when it assumed these companies would somehow benefit from the technology that replaces them.

The Memory Tax

Meanwhile DRAM prices keep climbing. The chip shortage is not temporary. Samsung, SK Hynix, and Micron are investing in new capacity, but fabrication plants take years to build. The demand curve for AI memory is not flattening. Every new model is bigger. Every new cluster needs more bandwidth. Every company that just raised a billion dollars for AI infrastructure is about to spend a significant portion of it on memory chips at prices that would have been considered absurd eighteen months ago.

This is the AI tax. Not a government tax. A physics tax. The compute required to run frontier models at scale requires physical resources, and physical resources have limits. You cannot download more DRAM. You cannot 3D-print high-bandwidth memory. You have to mine the silicon, fabricate the wafers, package the chips, and ship them across an ocean. And right now, AI is consuming all of it.

The trillion-dollar question — literally — is whether the value AI creates exceeds the cost of the resources it consumes. Two trillion dollars of evaporated market cap suggests the market is no longer sure.

What This Means for You

If you are an individual with skills and access to AI tools, this is the best environment you have ever worked in. The cost of capability is collapsing. What used to require a team now requires a subscription. What used to require capital now requires competence.

If you are a large company charging premium prices for services that AI can approximate, start running.

If you are an investor who bet that every tech company would be an AI winner, congratulations on learning an expensive lesson about the difference between disruption and value creation. They are not the same thing. They never were.

The two trillion dollars did not disappear. It was redistributed — from the assumption that incumbents would capture AI value to the realization that AI might be the thing that finally kills them. The money is still out there. It is just no longer pretending that Oracle is the future.