The $1 Trillion AI Semiconductor Selloff: When Good News Became Bad News

by

in

Estimated Reading Time:

13 minutes

Khan Capitals | June 2026


Key Takeaways

  • The AI semiconductor selloff erased more than $1 trillion in market value on Friday 5 June, with the Nasdaq falling 4% in its worst session since April 2025 and the S&P 500 slumping 2.6%, ending its run of nine consecutive winning weeks.
  • Broadcom’s guidance, not its results, lit the fuse: record Q2 revenue of $22.2 billion and AI semiconductor revenue of $10.8 billion (up 143% year on year) were overshadowed by a Q3 AI revenue guide of $16 billion against a $17.2 billion consensus, and a full-year AI forecast that was reiterated rather than raised.
  • A hot jobs report compounded the pain: the May employment report showed 172,000 jobs added against forecasts of 85,000, with 93,000 in upward revisions to March and April, pushing swaps markets to fully price a Federal Reserve rate hike by December.
  • The rates move was the transmission mechanism: the 2-year Treasury yield jumped 11 basis points to 4.16%, its highest since February 2025, repricing the discount rate beneath the market’s longest-duration growth assets at precisely the moment their earnings narrative wobbled.
  • Cross-asset deleveraging was immediate: bitcoin broke below $66,000 mid-week and lost roughly 12% over the week, with some $390 billion wiped from crypto market capitalisation, suggesting the unwind reflects positioning and leverage rather than a contained, single-sector earnings disappointment.

A Record Week That Broke on Friday

The AI semiconductor selloff of 5 June 2026 will be remembered for its violence, but the more instructive detail is what preceded it. On Monday 1 June the Nasdaq Composite closed above 27,000 for the first time. On Tuesday the S&P 500 finished above 7,600 for the first time, at 7,609.78, as the chip complex extended a rally that had carried the index through nine consecutive winning weeks. By Friday’s close, more than $1 trillion in semiconductor market value had been erased in a single session, the Nasdaq had registered its worst day since April 2025, and the S&P 500 had booked its first losing week in ten.

Two catalysts collided. Broadcom, the second-largest beneficiary of the custom AI accelerator boom, delivered a quarterly guide that was merely strong rather than accelerating. Hours later, the Bureau of Labor Statistics reported that the US economy added roughly double the number of jobs forecasters expected in May. Either alone would have been absorbable. Together they attacked the AI trade from both ends: the numerator of future earnings and the denominator of the discount rate. This piece examines the anatomy of the selloff, what it reveals about positioning in the world’s most crowded trade, and what the market may be underappreciating as the dust settles.

Anatomy of the AI Semiconductor Selloff

The sector-level damage was concentrated and brutal. More than $1 trillion was erased across the semiconductor complex on Friday alone. Nvidia fell 6.2%. Advanced Micro Devices dropped 10.9% to $466.38 and Intel fell 11.3% to $99.17. Micron Technology, the memory producer that had become the bull market’s latest standard-bearer, declined 13% on Friday after an 8% fall on Thursday. Broadcom and Micron together lost roughly a fifth of their value in two sessions. The Nasdaq-100 finished the week down 4.77%; the S&P 500 lost 2.64%.

AssetFriday 5 JuneWeekContext
Nasdaq Composite-4.0%-4.77% (Nasdaq-100)Worst session since April 2025; closed above 27,000 on Monday
S&P 500-2.6%-2.64%First losing week in ten; record 7,609.78 close on Tuesday
Nvidia-6.2%LowerLargest single-stock value loss of the session
Micron-13%Approx. -20% in two daysFollowed an 8% drop on Thursday
AMD / Intel-10.9% / -11.3%Sharply lowerClosed at $466.38 and $99.17 respectively
US 2-year yield+11bp to 4.16%HigherHighest since 25 February 2025
US 10-year yield+6bp to 4.54%HigherHighest since 21 May
BitcoinLowerApprox. -12%Broke below $66,000 mid-week; $390bn lost across crypto
Cross-asset scoreboard for the week ending 5 June 2026. Sources: CNBC, Seeking Alpha, Bloomberg, CoinDesk.

The breadth of the move matters as much as its depth. This was not a single-name accident. Every layer of the AI hardware stack repriced at once: the accelerator designers, the memory suppliers, the legacy foundries attempting to re-enter the race, and the networking complex that connects them. When a correction is that synchronised, the proximate trigger is rarely the full explanation. Crowding is.

Broadcom: When a $16 Billion Guide Costs a Trillion

On its face, Broadcom’s fiscal second quarter was exceptional. Revenue of $22.2 billion rose 48% year on year, non-GAAP earnings per share came in at $2.44, and the company described record revenue, operating profit and free cash flow in its results release. AI semiconductor revenue of $10.8 billion grew 143% year on year, driven by custom accelerators and AI networking.

The problem sat in the outlook. Broadcom guided current-quarter AI semiconductor revenue to approximately $16 billion against a consensus near $17.2 billion, and it declined to raise its full-year AI semiconductor forecast of roughly $56 billion, even while reiterating its expectation of more than $100 billion in AI revenue for fiscal 2027. In any normal market, reiterating 180% annual growth would be a victory lap. In this market, it was read as deceleration. The pattern echoes what we observed when Nvidia’s $91 billion quarter produced a muted reaction in May: when expectations embed perfection, the bar stops being the consensus estimate and becomes the whisper above it. Broadcom did not miss its numbers; it missed the market’s imagination.

There is an irony in the sequencing. Only days earlier, Dell’s $51.3 billion AI server backlog had been taken as evidence that the build-out was broadening down the stack. The demand signal did not change between Tuesday and Thursday. What changed was the market’s willingness to pay for it at any price.

The Jobs Report That Repriced the Fed

Friday morning’s employment report turned a sector correction into a market event. Nonfarm payrolls rose 172,000 in May, roughly double the 85,000 consensus, while the unemployment rate held at 4.3% and average hourly earnings rose 0.3% on the month and 3.4% on the year, according to the Bureau of Labor Statistics. The revisions were arguably more important than the headline: March was revised up by 29,000 to 214,000 and April by 64,000 to 179,000, adding a combined 93,000 jobs to the prior two months. A labour market that had appeared to be cooling through the spring, as we examined after the April employment report, now looks to have been re-accelerating for a full quarter.

The rates market did not hesitate. The 10-year Treasury yield rose more than 6 basis points to 4.54%, its highest since 21 May, while the policy-sensitive 2-year yield climbed 11 basis points to 4.16%, a level last seen in February 2025. Interest-rate swaps moved to fully price a quarter-point Federal Reserve hike by the December meeting, with roughly 60% odds of a move as soon as October. The repricing extends a journey that began with the spring’s hot inflation prints, when markets assigned just a 39% probability to a 2026 hike. That probability is now effectively 100%. The direction of travel for US policy is no longer in dispute; only the date is.

For equities, the mechanism is straightforward. The AI complex is the longest-duration asset in the public market: its valuations rest on cash flows projected years out, discounted at rates the market had assumed were heading down. A Fed that hikes into 2026 inverts that assumption. Good news for the economy became, in the space of one print, bad news for the multiple.

Memory, Margins and the Two-Day 20% Club

Beneath the index-level story, the memory segment is where the fundamental anxiety is most concrete. Micron entered the week as one of the market’s best-performing large caps, riding high-bandwidth memory demand that had been priced as structurally supply-constrained for years to come. Its 20% two-day decline, alongside sharp falls in Marvell, ARM and the broader complex, reflects a market beginning to interrogate two assumptions at once: that memory pricing power is permanent, and that smartphone and consumer demand can deteriorate without consequence for the AI supply chain that shares fabs, packaging capacity and capital budgets with it.

Neither assumption has been disproven. But the violence of the repricing in the names where the assumptions were most aggressively capitalised tells us where the leverage in the system sits. As we argued in our examination of the software-silicon divergence, the equity market has increasingly treated hardware cyclicality as a solved problem. Weeks like this one are a reminder that it is not.

Cross-Asset Spillover: The Deleveraging Tell

The clearest evidence that this was a positioning event, and not merely an earnings reassessment, came from outside equities. Bitcoin broke below $66,000 mid-week, notably before Friday’s equity damage, and finished the week down roughly 12%. Approximately $390 billion was erased from crypto market capitalisation and close to $7 billion in leveraged positions were liquidated. Strategy, the largest corporate bitcoin holder, disclosed its first bitcoin sale since 2022, a modest 32 coins, but a symbolic crack in the highest-conviction leveraged long in the market.

When the most speculative corner of the cross-asset complex starts deleveraging days before the equity catalyst arrives, the sensible inference is that risk budgets were already stretched. Rising front-end yields raise the cost of carry on every levered position simultaneously; the selling that follows is mechanical rather than thematic. Oil added a further complication for the inflation outlook, with Brent settling at $108.17 and WTI at $101.94 into this weekend’s OPEC+ ministerial meeting, leaving crude up roughly 78% year to date and underwriting precisely the price pressures that the Fed repricing reflects.

What the Market Is Underappreciating

First, the market is treating Broadcom’s guide as information about AI demand when it is more plausibly information about AI supply concentration. A $16 billion quarterly AI revenue guide, growing triple digits, with a reiterated $100 billion-plus fiscal 2027 target, is not a demand problem. The genuine signal is that custom accelerator revenue is lumpy, tied to a handful of hyperscaler deployment schedules, and therefore structurally incapable of delivering the smooth exponential curve that valuations had assumed. The cash flows are real; their quarterly choreography is not contractual.

Second, the consensus framing of the jobs report as unambiguously hawkish deserves scrutiny. The composition of May’s gains was narrow: leisure and hospitality contributed 70,000, local government 55,000 and health care 35,000, while financial activities shed jobs. Cyclical, rate-sensitive private hiring remains far cooler than the headline suggests. A Fed confronted with 3.4% wage growth and a 4.3% unemployment rate is under no obligation to hike quickly, and the gap between what swaps price and what the Federal Open Market Committee has signalled remains wide. Markets have moved from underpricing the hike risk to potentially overpricing its imminence.

Third, the speed of the crypto deleveraging suggests the system’s leverage is shorter-fused than in previous cycles, but also that it clears faster. Nearly $7 billion in liquidations in a week is painful precisely because it is fast; forced selling that completes quickly leaves cleaner positioning behind it. The historical parallel worth keeping in mind is early 2018: a melt-up punctured by a hot wage print, a volatility unwind that felt systemic, and an equity market that found its footing once the leverage had cleared, not when the macro data improved.

Investor Implications

Equities. The episode resets the burden of proof inside the AI complex. Positioning should reflect the distinction between businesses with contracted, multi-year revenue visibility and those whose valuations capitalise quarterly momentum. The two-day 20% declines were concentrated in the latter. Investors may wish to consider that the first losing week in ten is, historically, more often a positioning reset than a regime change, but that resets in crowded trades tend to come in waves rather than single sessions. Breadth beyond technology, which had been improving through the spring, becomes the key tell for whether this remains a sector event.

Fixed income. The front end has now done most of the work of pricing a 2026 hike, with the 2-year at its highest level in over fifteen months. The asymmetry at these levels is more balanced than it was in May: a Fed that validates the pricing delivers little further damage, while any softening in the data would unwind a fully priced move. Duration exposure at the long end remains hostage to supply dynamics and an oil price near $108, a combination we examined in our analysis of the synchronised sovereign bond rout.

Cross-asset. The simultaneous drawdown in chips and crypto is a reminder that the marginal buyer of both is the same levered risk budget. Diversification within the momentum complex is not diversification. Volatility markets had spent nine weeks pricing tranquillity; hedging costs reset higher on Friday but remain modest by the standards of past unwinds, and the option market’s repricing may lag the underlying regime if the rates story continues to build into the June FOMC meeting.

Conclusion

The week of 1 to 5 June 2026 delivered the AI trade’s sharpest stress test of the year: record index closes on Tuesday, a trillion dollars of semiconductor value gone by Friday, and a Federal Reserve hike moved from speculation to base case in a single payrolls print. The proximate catalysts, a Broadcom guide that merely met expectations and a labour market that refused to cool, matter less than what the reaction revealed: a market positioned for perfection in its largest sector and for dovishness in its policy outlook, and wrong on both in the same week. The structural AI investment case was not falsified on Friday; the price of believing in it carelessly was simply marked to market. The coming fortnight, with an OPEC+ decision, the SpaceX listing and the June FOMC meeting compressed into ten days, will determine whether this was the correction that extends the cycle or the first crack in it.

Sources: Bureau of Labor Statistics, The Employment Situation, May 2026; Broadcom Inc., Q2 FY2026 results (SEC Form 8-K); Bloomberg, US Bonds Slide as Hot Jobs Data Fuels Bets on 2026 Fed Hike; CNBC, 10-year Treasury yield surges above 4.5%; CNBC, Nasdaq falls 4% in worst day since April 2025; Seeking Alpha, Over $1T erased as chip selloff impacts Nvidia, Broadcom; CoinDesk, Bitcoin plunges below $66,000; CME FedWatch Tool.

Related Reading: For the earnings dynamics that set the stage for this week, see our analysis of Nvidia’s $91 billion quarter and the bar that became the beat. The downstream build-out context is covered in Dell’s $51.3 billion AI server backlog, while the capital expenditure cycle underpinning the trade is examined in the AI capex supercycle and the hyperscaler trade. On the policy repricing, our pieces on the PPI shock and the 39% hike bet and the synchronised sovereign bond rout trace how markets arrived at a fully priced 2026 hike. Readers following this story can also see our analysis of the SpaceX IPO pricing and its broken-playbook mechanics. For the leverage unwind that followed in digital assets, see the June 2026 crypto deleveraging. For the OPEC+ supply decision that landed as the war premium began to drain away, see OPEC Production Increase Meets the Peace Trade. For how the May inflation and jobs data flipped the rate-cut consensus into a hike debate before the June FOMC, see the Fed rate hike repricing. For the strategic backdrop to Apple’s position, see our analysis of the Apple Siri Gemini deal.

Keep reading Khan Capital

Join thousands of sophisticated investors receiving our institutional research direct to their inbox. Subscribers get a complimentary copy of The 2026 Geopolitical Portfolio: Defence, Energy, and Gold.

Depth over frequency. Unsubscribe anytime.

Disclaimer: The views expressed on Khan Capital are personal opinions of the author and do not represent those of any employer or institution. This content is for educational and informational purposes only and does not constitute investment advice. Past performance is not indicative of future results. Always consult a qualified financial adviser before making investment decisions.

About the author

Nauman Khan is an investment professional with experience across equities, fixed income, and alternative investments. He writes Khan Capital to provide independent, institutional-grade analysis of the events, policies, and structural forces shaping global financial markets. Read more


Comments

Leave a Reply

Your email address will not be published. Required fields are marked *