NEW YORK — The number that detonated the semiconductor market last week was not a loss. It was not a warning. It was a forecast of 200 percent year-over-year growth in AI chip revenue — and Wall Street still sold everything in sight.
That is the particular cruelty of what happened to U.S. chip stocks on June 5, 2026. Broadcom Inc. reported a record-breaking quarter — revenue of $22.2 billion, up 48 percent year over year, adjusted earnings per share of $2.44 against a $2.40 consensus — and then offered guidance that projected its AI semiconductor revenue would nearly double from $10.8 billion to $16 billion in the current quarter. By any prior standard, that would have been a triumph. What it was instead was $1.2 billion short of what analysts had modeled, and markets responded as if the AI trade itself had been put under review.
The Philadelphia Semiconductor Index fell 10.3 percent on Friday, its steepest single-session drop since March 2020. Over two trading days, the index surrendered more than 10 percent. Nvidia, the world’s most valuable chipmaker, slid roughly 6 percent, removing more than $300 billion from its market capitalization in a single afternoon. Micron Technology tumbled 11 percent, evaporating $127 billion. Marvell Technology gave back 12 percent. Advanced Micro Devices fell 10.5 percent. The aggregate damage across U.S.-listed chip stocks reached approximately $1.3 trillion, ending a nine-week winning streak that had made semiconductors the undisputed engine of the 2026 market rally.
Dennis Dick, a proprietary trader at Triple D Trading, told Reuters the market had been running on reflexive optimism. “Blindly buying the dip had been winning you money,” he said. “But that ended today.”
The mathematics of the selloff require some unpacking, because on their face they are difficult to reconcile. Broadcom’s CEO Hock Tan had guided third-quarter AI semiconductor revenue to $16 billion — a figure that, taken alone, represents over 200 percent growth year over year and nearly double what the company shipped in the second quarter. On the earnings call, Tan maintained the company’s full-year 2026 AI chip revenue target of $56 billion and reiterated a fiscal 2027 target exceeding $100 billion. Bookings for AI semiconductors in the second quarter were over $30 billion against $10.8 billion shipped — a backlog that, under normal market conditions, would have sent the stock higher.
What undid Broadcom was the gap between what it said and what investors had quietly expected. Sell-side consensus compiled by LSEG put the Q3 AI chip revenue estimate at $17.2 billion. Some buy-side models had built in even more. Stacy Rasgon, an analyst at Bernstein, noted the shortfall bluntly: forward AI guidance fell short of what investors had modeled going into the print. Crucially, Tan did not raise the full-year 2026 AI chip revenue forecast during the call — and after a first half that put AI revenue approaching $19 billion, some investors had been pricing in an upgrade. When no upgrade came, traders read the maintained guidance as a ceiling rather than a floor.
There was a second, less-discussed fault line. Analysts at Macquarie flagged that Broadcom’s revenue share of the tensor processing units it supplies to Alphabet is expected to decline from around 95 percent in 2026 to 80 percent in 2027 and 65 percent in 2028, as Taiwan-based MediaTek gains ground in that custom silicon program. Broadcom had effectively been the exclusive supplier for one of Alphabet’s core AI chip builds. That exclusivity, analysts said, is ending. The market had not fully priced that transition.
CFRA Research senior vice president Angelo Zino framed the reaction as a product of positioning rather than fundamentals. “The bar was really high going into the print,” he told reporters. “Part of the response you’re seeing from the shares points to that.”
The macro environment did not help. Earlier in the session, the U.S. Bureau of Labor Statistics released the May jobs report, which came in substantially stronger than expected. Robust labor data raised the probability that the Federal Reserve would hold rates higher for longer — bad news for high-multiple growth stocks that derive much of their valuation from discounted future earnings. The S&P 500 fell 2.3 percent on the day. The Nasdaq Composite lost 4.18 percent, closing at 25,709.43 for its largest single-day decline since the tariff volatility of April 2025.
The breadth of the selloff underscored how tightly the AI trade had become correlated. Micron fell not because of anything specific to its own business — high-bandwidth memory demand for AI accelerators remains robust — but because it has become a proxy for AI capital expenditure sentiment. As a key supplier of HBM chips used inside Nvidia’s data center GPUs, Micron’s stock has increasingly moved in lockstep with the broader AI infrastructure narrative. When Broadcom’s guidance implied that hyperscaler AI chip orders might not accelerate as fast as hoped, Micron absorbed the signal as if it were its own. The company had crossed a $1 trillion market capitalization just days earlier, after UBS tripled its price target on AI memory demand.
What the two-day rout exposed is that the chip sector had been trading not on reported earnings but on the assumption of continuously accelerating AI infrastructure spending. When any single data point in that chain — a guidance figure, a softened outlook, a competitive market share shift — falls short of the most optimistic scenario, the entire complex reprices simultaneously. That dynamic, traders noted, is not unique to this week. It is the structural condition of a sector that has risen 75 percent year to date even after Friday’s losses.
Ohsung Kwon, chief equity strategist at Wells Fargo, said the sector had been “way overbought” but stopped short of calling it a structural break. “I don’t think it’s the end of the semiconductor bull market,” he told Reuters. Bernstein’s Rasgon took a similar view, writing that with Broadcom growing revenues and earnings per share above 50 percent in an environment only getting stronger, a wait of a quarter or two for the story to re-emerge was “OK.”
What neither analyst addressed directly is the question that the selloff actually poses: not whether Broadcom and Nvidia are good businesses — they demonstrably are — but whether the valuations attached to those businesses already assumed a rate of AI spending acceleration that may not materialize in the near term. Broadcom’s AI revenue is growing at over 200 percent annually, yet its stock fell because investors had expected more. That is not a company problem. It is a pricing problem. And pricing problems, in a sector that has run 75 percent in a single year, tend to resolve messily.
The VMware dimension adds a further layer. Broadcom’s infrastructure software segment, which includes VMware, posted revenue of $7.18 billion in the second quarter — slightly below the StreetAccount consensus of $7.32 billion. Software grew 9 percent year over year, following a 1 percent print in the first quarter. The deceleration matters because infrastructure software carries a gross margin near 93 percent and operates at close to 79 percent. It is the segment that funds Broadcom’s capital-intensive AI chip roadmap. A sustained VMware slowdown would raise questions about whether that funding engine can sustain the pace investors have priced in.
The two-day wipeout arrived in the same week that Elon Musk was preparing a blockbuster IPO for SpaceX at a $1.75 trillion valuation — adding institutional pressure on a tech market already scrutinizing premium pricing at the top of the stack. Whether the chip sector’s nine-week rally resumes or whether Friday marked the beginning of a more sustained reassessment remains, as of this writing, genuinely unclear. Broadcom’s bookings of $30 billion against $10.8 billion shipped suggest demand is real. But demand being real and valuations being justified are not the same question, and for the first time in months, the market appears to be asking both simultaneously.
Even after Friday’s losses, the PHLX Semiconductor Index remains up approximately 75 percent for 2026 — a fact that cuts both ways. For bulls, it confirms the underlying strength of the AI trade. For bears, it is the reason the sector had so far to fall.
