On Friday evening, July 17, 2026, a correction turned into a milestone. The Philadelphia Semiconductor Index, or SOX, that thirty-stock barometer of the chip industry that has stood for the entire 2026 stock market like no other gauge, closed more than 20 percent below its late-June record high. That makes official what traders had only sensed all week: semiconductors, the beating heart of the artificial-intelligence trade, are now in a bear market. Bloomberg captured it in a headline worth remembering: “Chips Stocks Sink Into Bear Market as 105% AI Rally Fizzles.”
But the story that matters is not the crash itself. It is what the crash claims — and who takes the witness stand next week to either confirm or refute that claim. Because over the past few days the market has placed a very precise bet: it punished the sellers of AI infrastructure. On Wednesday, the buyers testify.
What happened Friday — the anatomy of a bear market
The raw numbers describe a reversal that unfolded at breathtaking speed. From its March low to its late-June peak, the SOX had climbed a staggering 105 percent — a doubling in barely three months, powered by the conviction that demand for compute chips was limitless. This past week the same index shed roughly 11 percent, its worst weekly performance since March 2025. On Friday alone it fell as much as 5.7 percent intraday. That pushed its drawdown from the June high past the 20 percent mark, the technical threshold analysts use to declare a bear market.
The broad market felt the weakness without collapsing. The tech-heavy Nasdaq Composite lost 1.5 percent Friday to close at 25,881.95, while the broad S&P 500 gave up a more modest 0.5 percent to 7,533.77. The distinction is telling: this is not a wholesale sell-off but a surgical retreat from the very sector that had led the 2026 rally. In Asia, South Korea’s KOSPI — heavily dependent on Samsung and SK Hynix — fell nearly 20 percent from its own peak in just two weeks, and Japan’s Nikkei came under pressure too. The AI trade is global, and so is its unwind.
Why now: the four triggers
A bear market rarely springs from a single cause, and this one is no exception. Four forces converged. The first and most fundamental is simply valuation. After a doubling in three months, every conceivable piece of good news was priced in. The trade was crowded — fund managers, retail investors, and momentum algorithms were all sitting on the same side of the boat. When the boat began to tip, everyone tried to get out at once.
The second trigger is structural and is likely to haunt the market for a while. In early July, Bloomberg reported that Meta was quietly building a cloud unit, internally called “Meta Compute,” to sell surplus AI computing capacity to outside customers — a direct competitor to Amazon Web Services, Google Cloud, and Microsoft Azure. For the chip industry this was a cold shudder: if the largest buyer of graphics processors begins reselling its capacity, it fundamentally rewrites the supply-and-demand equation for AI infrastructure. Meta itself jumped more than ten percent on the news, but its suppliers cratered — Micron fell 10.6 percent that day, and chip designer AMD dropped 6.9 percent.
The third factor came from Intel. Reports that its new 18A-P manufacturing process would not reach profitable yields until late 2026 or even 2027 undercut the turnaround story that had powered Intel’s remarkable first-half rally. And fourth, the specter of a memory glut haunts the sector: China’s rising DRAM maker CXMT is heading toward a multibillion-dollar listing, and the fear that commodity memory could soon be abundant weighs on the entire pricing thesis.
The heart of the matter: sellers versus buyers
To understand what next week hinges on, you have to read the bear market correctly. What the market punished over the past few days are the sellers of AI infrastructure — the makers of the picks and shovels of this gold rush. Nvidia, Micron, AMD, Broadcom, the memory producers, the foundries: they all live on the fact that the big technology companies spend ever more money on data centers, year after year.
That is precisely the bet embedded in the bear market. Every share of a chipmaker that gets sold is, at its core, a statement: I no longer believe the hyperscalers’ gigantic capital budgets are durable. I do not believe they will pay off. The sellers’ market prices in a doubt about the buyers, without those buyers having yet said a word. And here is where it gets interesting: it is precisely next week that those buyers begin to open their books.
Next week on the stand: Alphabet and Tesla
On Wednesday, July 22, two heavyweights from the ranks of the “Magnificent Seven” report after the U.S. close: Alphabet, Google’s parent, and Tesla. Both are central buyers of AI infrastructure, and both embody the question that grips the entire market: does the spending pay off?
Alphabet has beaten both earnings and revenue expectations in every quarter of 2026, making it a model student of earnings season. This time the decisive line will be the cloud segment. Analysts expect Google Cloud revenue of roughly $22.8 billion, up about 67 percent year over year. If the cloud is growing as fast as this enormous investment pace demands, Alphabet delivers the first hard proof that the AI billions are actually converting into revenue. If growth disappoints, the chip bear market becomes a self-fulfilling prophecy.
Tesla faces a different but related test. The stock is down roughly 15 percent in 2026, and the company has raised its planned capital spending for the year from $20 billion to more than $25 billion — chiefly for the data centers meant to power self-driving, the robotaxi network, and chief executive Elon Musk’s AI ambitions. Analysts expect earnings up 22 percent and revenue up 12 percent from a year ago. Options pricing implies a possible swing of about seven percent in either direction. Tesla thus embodies the extreme case of the buyer: a company spending ever more on AI while its core business sputters and investors lose patience.
And this Wednesday is only the opening act. A week later, on July 29, Microsoft and Meta report; on July 30, Apple and Amazon follow. Chip giant Nvidia — whose results are considered the referee of the entire AI trade — does not report until late August. The witness stand is booked for weeks. But the first and in many ways most important word belongs to Alphabet and Tesla.
The ten-cent question
What exactly should investors watch for in these reports? Three figures: cloud growth, capital-spending guidance, and return. The crux is the gap between what is spent and what is earned. By Goldman Sachs’ math, Google, Amazon, Microsoft, and Meta alone plan combined capital expenditures of roughly $725 billion for 2026 — a 77 percent jump from last year’s already record-shattering $410 billion. Cumulatively through 2031, the figure could reach $7.6 trillion.
Against that stands an uncomfortable number: AI-related services generated an estimated $25 billion in revenue in 2025 — against more than $250 billion in infrastructure spending. That is roughly ten cents of revenue per dollar invested. This gap is the real battleground between bulls and bears. The bulls say the return is coming, just as the internet only monetized its infrastructure years after 2000. The bears say ten cents on the dollar is not a runway, it is a warning signal. And it is precisely here that next week’s reports deliver ammunition — for whichever side.
The stocks caught in the middle
For American investors, the chip bear market is anything but abstract. The core sellers — Nvidia, Micron, AMD, Broadcom — have led both the rally and the retreat, and any hyperscaler capex signal moves them directly. The buyers, meanwhile, split into two camps this earnings season: those that must justify their spending (Alphabet, Tesla, Microsoft, Meta, Amazon) and those, like Apple, that spend more cautiously and may benefit if compute gets cheaper. A third, often overlooked layer is power and infrastructure: the data centers driving all of this consume staggering amounts of electricity, and the utilities, grid operators, and cooling and equipment makers that feed them are quietly indispensable to every AI chip that hums.
For those holding a broad semiconductor or technology ETF, the swings of this sector show up immediately in the portfolio — on the way up and on the way down. In a tax-advantaged retirement account such as an IRA or 401(k), those swings compound without an annual tax drag, which is one reason long-term investors are often counseled to ride out sector volatility rather than trade it. The volatility that in early 2026 delivered nothing but gains is now showing its other face.
Risks, counterarguments, and an outlook
It would be a mistake to read the chip bear market too hastily as the bursting of a bubble. Several counterarguments deserve a hearing. First, the structural demand for compute is real and still growing — Goldman’s $7.6 trillion projection has not changed because of the slide. Second, the broad earnings season has been remarkably strong: 95 percent of the 47 S&P 500 companies reporting through Friday beat profit expectations, led by outliers such as health insurer UnitedHealth at $6.38 per share against an expected $4.94. The market, in other words, is by no means weak — it is merely rotating away from the most expensive segments. Third, the chip sector has historically followed a cyclical pattern of overshoot and disillusionment; a 20 percent decline after a 105 percent surge is a breather, not a breakdown.
It is worth being concrete about what a “good” and a “bad” report even look like on Wednesday. A good result for the bulls would be: Alphabet’s cloud growing in line or faster, management reaffirming its investment plans with reference to a bulging order book, and Tesla delivering the first tangible progress on robotaxi operations that justifies the capital outlay. A bad result would be the mirror image: decelerating cloud growth, a cautious tone on future spending, or a lowered forecast — precisely the signals the chip bear market has already bet on. Because expectations remain high after the rally, even a solid but not spectacular result can trigger a negative reaction; the bar is elevated, and in nervous phases the market punishes any disappointment out of proportion.
The biggest risk lies in the feedback loop. Should the hyperscalers even hint next week at throttling their investment budgets, the sellers’ bear market would jump to the buyers — and a sector correction could become a broader re-rating of the entire AI theme. Conversely, a trio of strong cloud numbers, reaffirmed spending plans, and the first evidence of return could make the chip sell-off look, in hindsight, like what many analysts suspect it is: a painful but healthy shakeout of a crowded trade.
Next week, then, is more than a cluster of quarterly reports. It is the moment the market stops speculating about the buyers of AI infrastructure and begins to listen to them. Between Wednesday evening, when Alphabet and Tesla open their books, and the end of mega-cap season, it will be decided whether 2026 is remembered as the year the AI trade underwent a healthy correction — or as the year the most expensive bet in market history failed its first serious stress test.
Try TradingView Free for 30 Days
Plus get a $15 discount on your first subscription through this link.
Read more in our topic hub: Topic Hub: Portfolio Crash Protection 2026
