The Weekly Investor
AI and Tech

Chip Sector Loses $923B in a Day as Broadcom Misses AI Guide

Broadcom's below-consensus AI chip guide triggers a $923B single-session sector wipeout — Nvidia, Intel, AMD, and SK Hynix all suffer double-digit losses.

June 29, 2026

Key Points

  • Broadcom's Q3 AI chip sales guidance of $16 billion missed the $17.2 billion analyst estimate, triggering a single-session wipeout that erased approximately $923 billion in combined market value across the top ten semiconductor decliners.
  • The sell-off exposed a structural fault line in the chip sector — AI infrastructure names are priced for sequential acceleration, not steady growth, and any guidance that merely meets rather than exceeds expectations punishes stocks severely.
  • Traders should watch TSMC's forthcoming price hike implementation across all advanced nodes for the next margin-pressure catalyst, with Qualcomm's reported $8–10 billion Tenstorrent bid adding a potential M&A wildcard.


Broadcom beat on both revenue and earnings — $22.19 billion versus the $22.13 billion consensus, non-GAAP EPS of $2.44 versus $2.39 — and still sent the entire semiconductor sector into freefall. The company's Q3 AI chip sales guidance of $16 billion came in $1.2 billion below the $17.2 billion analyst estimate, and Broadcom declined to raise its full-year AI forecast. That was enough. Shares fell 14%, and the blast radius swallowed nearly $923 billion in combined market value across the sector's ten largest decliners in a single session.

The $923 Billion Rout, by the Numbers

The June 23 carnage was indiscriminate in its breadth and precise in its logic. Nvidia dropped 3%, erasing nearly $280 billion in market value on its own — a figure that exceeds the total market capitalization of most S&P 500 companies. Intel fell 7.6%. AMD lost 6.2%. Micron gave back 8.5% before its own earnings release subsequently reversed the damage. The iShares Semiconductor ETF declined 6.2% on the day, its worst single-session performance in months.
The damage extended well beyond U.S. borders. South Korea's Kospi index, heavily weighted toward semiconductor names, closed 10% lower. SK Hynix and Samsung — the two closest competitors to Micron in HBM memory — each finished the session down more than 12%. In dollar terms, TSMC, AVGO, and MU were each stripped of more than $100 billion in individual market value. The speed and uniformity of the decline reflected the degree to which the entire sector had been priced against a single assumption: that AI chip demand would accelerate in a straight line through 2026 and beyond.
The sell-off reveals a structural vulnerability in how AI infrastructure stocks are currently valued. Every major semiconductor name trading at elevated multiples is priced not just for strong demand, but for demand that consistently exceeds already-elevated expectations. The tech sector rout, as reported by CNBC, made clear that when a bellwether like Broadcom guides in-line rather than above consensus, the market reprices the entire supply chain simultaneously — not because fundamentals have deteriorated, but because the margin of safety embedded in elevated valuations is functionally zero. A $1.2 billion guidance miss against a $17.2 billion estimate is a 7% shortfall. A 14% stock decline in response tells you exactly how much excess optimism was already in the price.

TSMC Price Hikes and the Qualcomm Wildcard

While the Broadcom-triggered rout dominated headlines, two structural developments will have longer-lasting implications for how traders position in the chip sector through the second half of 2026. The first is TSMC's reported decision to hike prices across all advanced nodes. In Q1 2026, TSMC reported $35.9 billion in revenue at 66.2% gross margins, with 3nm alone accounting for 25% of wafer revenue and the full advanced-node portfolio — 7nm and below — representing 74% of the company's total wafer business. A price increase spanning 74% of TSMC's wafer revenue touches every major fabless customer simultaneously: Nvidia, AMD, Apple, and Qualcomm all face higher input costs with nowhere to go, since TSMC's Arizona capacity has been sold out through 2027 since early 2025.
CEO C.C. Wei stated plainly at a recent shareholder meeting that U.S. manufacturing capacity remains "far from enough," with additional production sites planned in Japan and Germany. TSMC's May revenue growth came in at 30.1%, and the company has already raised its 2026 full-year revenue growth target to more than 30%. Susquehanna maintains a $575 price target on TSM. The price hike calculus is straightforward for TSMC shareholders — constrained supply plus inelastic AI-driven demand equals pricing power — but for every downstream customer, it introduces a margin compression variable that wasn't fully in analyst models heading into the back half of the year.
The second development is Qualcomm's reported early-stage talks to acquire Tenstorrent for between $8 and $10 billion. Tenstorrent designs AI chips using the open RISC-V instruction set architecture and carries the engineering credibility of chip veteran Jim Keller, whose design fingerprints are on some of the most consequential processor architectures of the past two decades. The acquisition would give Qualcomm a direct entry point into the AI data center hardware market currently dominated by Nvidia and, to a lesser extent, AMD. Qualcomm has separately raised its fiscal year 2029 non-handset revenue target to $40 billion and launched the Dragonfly brand for AI data center solutions, framing the Tenstorrent deal as a strategic acceleration rather than a speculative bet. At $8–10 billion against Qualcomm's scale, the price is manageable — but the regulatory timeline and integration risk in a competitive market where Nvidia's moat is measured in software ecosystem depth as much as hardware performance make the outcome genuinely uncertain.

Nvidia's Bond Deal and the Google-SpaceX Revelation

Nvidia's response to the current environment was to raise $20 billion in a multi-tranche bond deal — the largest debt issuance in the company's history, dwarfing the $5 billion bond sale it executed in 2021. Capex is projected at $7.9 billion for 2026, up from $6 billion in 2025 and $3.2 billion in 2024, as the company moves to fund its own AI infrastructure buildout while retiring existing corporate debt. Nvidia is simultaneously projecting $1 trillion in AI infrastructure demand by 2027 and has announced 35 new AI and HPC supercomputer deployments across Europe, including the JUPITER exascale machine.
The most striking demand data point of the week came from a filing revealing that Google will pay $920 million per month to rent 110,000 Nvidia GPUs, CPUs, and memory from SpaceX, covering the period from October 2026 through June 2029. If SpaceX fails to make the GPU capacity available by September 30, 2026, Google receives a one-month grace period before it can terminate the agreement. That structure — nearly $1 billion per month in committed GPU rental payments from Alphabet to a SpaceX data center — illustrates the degree to which the scramble for AI compute has moved beyond anything the semiconductor industry's traditional demand models were built to anticipate. For context, this comes less than a month after SpaceX agreed to rent the entire GPU capacity at its Colossus 1 data center — more than 220,000 Nvidia chips — to Anthropic. As Yahoo Finance reported, the chip sector is navigating the paradox of record fundamental demand while markets punish any guidance that fails to exceed already-stretched expectations.
The divergence between AI infrastructure exposure and consumer-facing chip revenue is the most important positioning theme in the sector right now. IDC projects the global smartphone market will suffer its largest year-on-year volume decline on record in 2026, down 13% to the lowest level in a decade, as memory producers have strategically redirected output toward higher-margin data center customers. Against that backdrop, Deloitte's 2026 Semiconductor Industry Outlook projects global chip sales will hit a historic peak of $975 billion this year — but the gains are concentrated. Intel and AMD, with meaningful consumer and PC exposure, remain under structural pressure. Micron, TSMC, and Nvidia, with deep AI infrastructure positioning, are playing a different game entirely. The next definitive read on where Broadcom's AI revenue guidance lands relative to the $16 billion Q3 projection will be the sector's clearest near-term signal — watch that print for confirmation of whether the June 23 rout was a reset or the beginning of a more sustained derating.

The Weekly Investor

Daily market analysis for active traders. Free.

Keep Reading

View more