The boom that powered semiconductor stocks to record-setting heights just last month is now sputtering.

Why it matters: Chipmakers are the picks and shovels of the AI gold rush, and they're looking shaky.

• A recent selloff, combined with new stress over competition from Chinese open-source AI models, is raising concerns over the durability of the AI trade more broadly.

Catch up quick: Wall Street learned to love the stocks of semiconductor firms like Intel, AMD and Micron Technology this year.

• These companies sell the chips underpinning the massive AI buildout by the hyperscalers, and their stocks have soared by double- and triple-digit percentages this year.

The latest: This month, chips got crushed.

• The stocks "got to extremes," John Roque, head of technical analysis at 22V Research, a Wall Street research firm, tells Axios.
• "I think what we're seeing now is a correction. It's as rash now on the downside as it was rash on the upside."

By the numbers: Since hitting a record high on June 22, the closely watched iShares Semiconductor ETF, a fund that trades under the ticker SOXX and tracks 30 chipmaker industry stocks, has fallen 20%.

• The drubbing was even more brutal for Intel, down about 33% and Micron, down nearly 30%.
• Nvidia, which was a high-flying chip stock before those other guys ever got out of bed this year, held up a bit better. It's now down close to 3% over the same period.

"Moves of the type we're seeing in semiconductors are incredibly rare," says Steve Sosnick, chief strategist at Interactive Brokers, describing them as "parabolic."

• "They tend to end unpredictably and in an ugly manner. And I think, to some extent, that's what we're seeing."

Yes, but: For the year to date, these stocks are still absolutely slaying: SOXX is up 66%, Micron 169%(!) and Intel 141%.

• And so far the stock market overall is holding up OK, as investors rotate into other equities.

State of play: Still, investor faith in the AI trade looks shakier. Investors are uncertain which of the hyperscalers will actually profit from their record spending on AI, Christopher Wood, global head of equity strategy at Jefferies, wrote earlier this month.

• "The new quarter has begun with much talk of 'AI fatigue' as investors look out for a peaking out of momentum."
• Still, Wood remains optimistic about AI picks and shovels more generally.

Between the lines: The recent moves down shouldn't be totally surprising. "There were some tells along the way," Sosnick says.

• He points out that Nvidia reported strong financial results in May, but its stock fell afterward. Something similar happened with Samsung.

The intrigue: South Korea's benchmark stock index, the KOSPI, is proving to be a bit of a canary in the coal mine.

• The KOSPI shot up by triple digits at its peak, driven largely by the stock prices of just two companies. It's now in bear market territory.
• Samsung and SK Hynix alone recently accounted for nearly half the KOSPI's total market value, per Roque's calculations.

Reality check: The U.S. market, however, is much less concentrated.

What to watch: Financial reports from Alphabet (Wednesday) and Intel (Thursday) are coming out this week — the AI trade's next stress test.

What's next: No one really knows what comes next, as Warren Buffett ruminated in Berkshire Hathaway's 2000 shareholder letter just as the dot-com bubble was bursting.

• "Nothing sedates rationality like large doses of effortless money," he wrote, comparing the behavior of investors to "that of Cinderella at the ball."
• "The giddy participants all plan to leave just seconds before midnight. There's a problem, though: They are dancing in a room in which the clocks have no hands."

No, we're not talking about Social Security. "Generational" is how some analysts describe the current transfer of money from hyperscalers to chipmakers.

Why it matters: The chip stock boom that preceded the current malaise was based on that simple reality, which is reflected in the above chart.

• In short, the massive investments that hyperscalers are making in "compute" are pumping cash into the coffers of the companies that make the products they need — computer chips foremost among them.

What they're saying: "A generational transfer in free cash flow is taking place," wrote analysts at Bank of America in a research report published earlier this month.

The big picture: As any Keynesian could tell you, this is, basically, how the economy works.

• One person or company's spending is, mechanically, another company's revenue or person's paycheck.

The intrigue: Will hyperscalers, who have been writing the lion's share of the checks to cover the building boom, continue to do so now that their own cash generating powers have withered? If so, they'll be increasingly reliant on the kindness of bond and stock market investors to finance their plans.

Reality check: We should stress that just because a company's free cash flow is negative, it doesn't mean it's broke.

Case in point: Amazon is expected to end the year with negative free cash flow of more than $12 billion.

• But Jeff Bezos' erstwhile online book emporium is still expected to have over $100 billion in cash and cash equivalents in the bank by the time the dust clears on 2026.

Between the lines: Still, if hyperscalers do try to cut back on investment, the market could face a corporate variation on another classic Keynesian concept: the paradox of thrift.

• That's the idea that it often makes sense for an individual household to tighten its belt, cut spending and sock away cash in the face of uncertainty. But when everyone does it simultaneously, it transforms mere worries about the future into an actual economic downturn that can leave even the most frugal families worse off.
• In the current context, it might make sense for any one hyperscaler to trim its sails on AI and sock away a bit more cash going forward — especially in light of recent developments in cheaper Chinese AI models.
• But if all of the big AI spenders close their wallets at once, the market — and the economy — could hit a jarring air pocket, or perhaps something a bit worse.