Why IBM’s $67 billion crash is really about AI

IBM just had the worst single-day stock drop in its 115-year history, losing around a quarter of its value and wiping out $67 billion in market cap. On paper, the company’s results weren’t great—but they also weren’t the kind of disaster you’d normally expect to trigger a historic crash.

So what actually happened? Underneath the earnings miss is a much bigger story about how the AI boom is scrambling corporate budgets, distorting hardware markets, and potentially setting up the next big economic shock.

What really happened to IBM’s stock

IBM reported second-quarter revenue of $17.2 billion. That was up about 1% year over year, but it still fell roughly $660 million short of Wall Street expectations. Some parts of the business grew, others shrank:

• Software revenue grew around 5%, but investors were hoping for double-digit growth.
• Infrastructure revenue (including mainframes) fell about 7%.
• Consulting revenue was flat, when it was expected to rise.

Those numbers are disappointing, but not usually “lose a quarter of your company’s value in a day” bad. The real shock came from IBM’s explanation: a sudden, massive shift in how its customers are spending money—driven by AI.

How AI is wrecking old software business models

AI is now at the center of almost every big tech and economic story, and IBM is no exception. The first big shift is on the software side. Many of the things companies used to buy traditional software for can now be done faster and more flexibly with AI tools.

For a company like IBM, which sells a lot of enterprise software, that’s a problem. When AI systems can automate updates, testing, documentation, and even parts of development, customers start to question how much they should keep paying for older, more rigid software products and licenses.

One example: a new AI model from Anthropic was promoted as being able to streamline updates for one of IBM’s long-standing software products. That kind of announcement doesn’t just threaten a single product line—it makes customers rethink their whole roadmap. If AI can replace or radically reduce the need for certain tools, why lock in big, multi-year contracts today?

The AI hardware gold rush: why memory chips matter

The second big shift is on the hardware side, and it’s even more dramatic. Training and running large AI models requires enormous amounts of specialized hardware—especially high-bandwidth memory (HBM) chips used in AI servers and accelerators.

There are essentially two broad types of memory chips in this story:

• Standard memory chips used in everyday devices and servers.
• High-bandwidth memory chips designed for AI workloads, which can command 3–5x more revenue per chip.

Because HBM is so much more profitable, the major memory manufacturers—Samsung, Micron, and SK Hynix, who together control over 95% of global production—have been aggressively shifting capacity away from standard memory and into HBM.

That’s created a severe supply crunch. SK Hynix has already committed its entire HBM production capacity for 2026 under long-term contracts. Micron has warned that new factories won’t meaningfully ease supply until around 2028. In the meantime, prices are exploding: by the end of this year, HBM prices are expected to be up as much as 355%.

Why IBM’s customers suddenly slammed the brakes

Now imagine you’re a large enterprise trying to build or scale AI systems. You know HBM prices are soaring, and supply is tight for years. The rational move is to buy as much hardware as you can, as early as you can, before it gets even more expensive or harder to find.

That’s exactly what many companies did. They pulled forward their hardware spending—especially on AI servers and memory chips—into the present. But budgets aren’t infinite. If you’re suddenly spending far more than planned on AI hardware, you have to cut or delay something else.

For IBM, that “something else” was core revenue:

• Mainframe upgrades got postponed.
• Software license renewals were delayed.
• Consulting projects were pushed back.

IBM admitted it expected some shift in client demand, but badly underestimated how big and fast it would be. The company said that “numerous large deals failed to close on the timelines we expected,” and that it didn’t adapt quickly enough to the new reality.

Anthropic’s AI and the cybersecurity shock

Making things worse, Anthropic’s new AI model—pitched as being able to find vulnerabilities in company software before the company itself does—spooked a lot of IBM’s customers.

If an AI can rapidly uncover security holes, then every organization suddenly has a new, urgent problem: what if someone else’s AI finds your vulnerabilities first? That fear is pushing companies to redirect even more budget into cybersecurity tools, audits, and defenses.

Again, that money has to come from somewhere. For many IBM clients, it meant yet another reason to delay spending on IBM’s traditional software, infrastructure, and consulting services.

Is this a one-off quarter or a warning for everyone?

The big question now is whether IBM’s bad quarter was a temporary timing issue—or a sign of a deeper, longer-term shift.

IBM’s own story is optimistic: customers delayed some IBM spending so they could load up on scarce AI hardware. Now that they’ve secured the chips, IBM expects those delayed deals to come back in future quarters.

If that’s true, IBM could rebound as soon as the next earnings report. But if it’s not, then we’re looking at something more structural: a multi-quarter, possibly multi-year reallocation of budgets away from legacy software and services, and toward AI infrastructure and AI-native tools.

Analysts are already warning that the same pattern is likely playing out at other companies that haven’t reported yet. In that sense, IBM might just be the first major casualty of a broader realignment of the global tech economy.

If you’re interested in how these shifts could reshape winners and losers over the next few years, it’s worth looking at broader AI trends and the companies positioned to benefit, like in this breakdown of three AI mega trends and the stocks tied to them.

The new AI elite: massive revenue, even bigger burn

While legacy players like IBM are struggling with AI-driven disruption, a new group of AI-first companies is attracting staggering amounts of capital. Think of firms building foundation models, AI infrastructure, or AI-native platforms.

Some of these companies are already generating impressive revenue, but many are also burning cash at an even faster rate. Their valuations are being pushed sky-high by:

• Huge private funding rounds from venture capital and sovereign wealth funds.
• Strategic investments from big tech companies that need access to cutting-edge models.
• Expectations that AI will underpin the next decade of economic growth.

In some cases, founders and early executives are becoming billionaires overnight after major funding rounds—long before their companies have proven they can be sustainably profitable. That kind of wealth creation, detached from long-term performance, is exactly what makes people worry about bubbles.

Is the AI boom a bubble—or something different?

There’s a growing debate about whether AI is in bubble territory. The two main camps see it very differently:

The bulls’ view:

• Today’s AI leaders generate real revenue and cash flow, unlike many dot-com companies in the 1990s.
• On the S&P 500, AI-related valuations are around 22x earnings—below the 25x level often associated with classic valuation bubbles.
• AI is seen as a foundational technology, more like electricity or the internet, justifying heavy long-term investment.

The bears’ view:

• The bubble isn’t just in stock prices—it’s in the earnings themselves.
• A lot of current revenue comes from circular spending (AI companies buying from each other), aggressive private funding, and short-lived hardware upgrade cycles.
• If capital dries up or hardware demand normalizes, earnings could fall sharply, exposing how fragile the current boom really is.

In that scenario, when the music stops, it won’t be the executives with early equity and golden parachutes who suffer most. It will be ordinary investors with retirement accounts, employees whose companies bet too hard on the wrong side of the AI wave, and smaller retail investors chasing hype instead of fundamentals.

For a deeper look at how specific AI capabilities—not just hype—can drive real value, it’s helpful to study concrete breakthroughs, like new visual reasoning techniques that meaningfully expand what AI can do.

What this means for you

You don’t need to be a trader to be affected by what’s happening around IBM and the AI boom. If you have a 401(k), pension, or index fund, you’re already exposed to these shifts. If you work in tech, software, consulting, or any industry that relies heavily on IT, your company’s budget decisions are likely being reshaped by the same AI forces that hit IBM.

The key takeaway isn’t to panic or try to time the market. It’s to understand that AI isn’t just a new feature—it’s reshaping where money flows, which business models survive, and who gets squeezed in the process. IBM’s $67 billion crash is a dramatic reminder that even the most established players aren’t safe from that realignment.