Key Bullet Points:
- Nvidia shares have fallen more than 11% in the past month as the semiconductor sector endured two brutal selloffs — a 7.9% SOX plunge on June 23 and a 6.3% crash on July 1 — yet the company's data center revenue is projected to hit $190 billion this fiscal year, up 65% year-over-year
- Meta's announcement that it's building a cloud business to sell "excess" AI computing capacity spooked the entire chip supply chain, raising fears that the biggest spenders may be overbuilding — but Nvidia CEO Jensen Huang told investors the selloff is a "buying opportunity" because "the buildout of AI has just begun"
- The global selloff rippled through Asia on Thursday as South Korea's Kospi was hammered: Samsung Electronics fell 9%, SK Hynix plunged 14.57%, and SK Square dropped 13.2% — with Samsung and SK Hynix now making up roughly half the Kospi's total weight
- Despite the carnage, Nvidia's fundamental story remains arguably the strongest in tech: GB300 systems are shipping in quantity, the Blackwell architecture is driving massive data center upgrades, and fiscal Q1 data center revenue hit $51.22 billion — 89.8% of total sales and a 66% year-over-year increase
- As markets reopen after the Independence Day break, investors face a defining question: was the first half's 75.5% surge in the semiconductor ETF (SMH) the start of something bigger, or the peak of a trade that's already priced in the future?
The AI King Under Siege
For the first time in years, the narrative around Nvidia is shifting.
Not the numbers. The numbers are absurd. $51.22 billion in data center revenue last quarter. A projected $190 billion for the full fiscal year. GB300 systems flying off production lines. A 66% year-over-year surge in the division that accounts for nearly 90% of the company's sales.
No, what's changing is the story around those numbers — and on Wall Street, the story matters as much as the spreadsheet.
As markets reopen Monday after the Independence Day break, Nvidia sits at the epicenter of a question that could define the second half of 2026: has the AI trade gone too far?
Two Selloffs, One Message
In the span of eight trading days, the semiconductor sector was hit with two of its worst sessions of the year.
On June 23, the Philadelphia Stock Exchange Semiconductor Index — the SOX — plunged 7.9% as investors questioned whether debt-funded AI spending by hyperscalers could be sustained. Nvidia itself fell 4.2% that day, dragging the Nasdaq to more than a one-week low.
Then on July 1, it happened again. The SOX dropped 6.3% after Bloomberg reported that Meta Platforms was building a cloud business to sell its "excess" AI computing capacity. The word "excess" sent shockwaves through every company that sells infrastructure to the AI buildout.
KLA Corporation dropped 12%. Lam Research fell 9.7%. Applied Materials sank 10%. Corning crashed 13.6%. Micron and SanDisk — the best-performing stocks in the S&P 500 for the first half — both dove over 10%.
The message was brutal in its simplicity: if the biggest AI spenders are building more than they need, maybe the companies supplying them are overvalued.
The Global Ripple Effect
The damage didn't stop at the U.S. border.
On Thursday, South Korea's market bore the full weight of the selloff. Samsung Electronics fell 9.06%. SK Hynix plunged 14.57%. SK Square, the largest shareholder of SK Hynix, dropped 13.2%.
The two chipmakers now account for roughly half of the Kospi's total weight — up from about a quarter at the end of last year. When Samsung and SK Hynix move, the entire Korean market moves with them.
"A sharp move in either name drags the whole index with it before the other roughly nine hundred listed companies get a say," noted Zavier Wong, a market analyst at eToro.
The selloff in Asia underscored something that's easy to forget: the AI trade isn't just an American phenomenon. It's a global supply chain, and when confidence cracks in Silicon Valley, it shatters in Seoul.
Jensen Huang's Counterpunch
Into this chaos stepped the one person who has the most to gain — or lose — from the AI narrative: Nvidia CEO Jensen Huang.
In comments that cut through the market panic, Huang called the selloff a "buying opportunity," insisting that "the buildout of artificial intelligence has just begun."
"Everybody should be very excited," Huang said. "They can now buy stock at a cheaper price and enter."
It's the kind of statement that either ages brilliantly or terribly, with no middle ground.
Huang's confidence isn't baseless. Nvidia's GB300 systems — the next generation of its Blackwell architecture — are shipping in significant quantities, offering performance improvements over the Hopper generation. Cloud computing companies, sovereign AI projects, and enterprise clients are placing orders at a pace that's kept Nvidia's data center segment growing at 66% year-over-year even as the broader market wobbles.
The company projects fiscal 2026 data center revenues of $190 billion, representing 65% growth. If those numbers hold, Nvidia isn't just growing — it's accelerating.
The $190 Billion Question
But projections and reality are different things, and the gap between them is where fortunes are made or destroyed.
The bear case isn't that AI spending will stop. It's that AI spending will slow — that the hyperscalers have pulled forward years of demand into a few quarters of frenzied building, and that the hangover is coming.
Meta's "excess capacity" announcement gave that argument its strongest evidence yet. If the company that has spent more on AI infrastructure than almost anyone else already has more than it needs, what does that say about the trajectory of capital expenditure across the industry?
The semiconductor ETF (SMH) surged 75.5% in the first half of 2026 — its best start to a year on record. That kind of performance doesn't happen without a massive consensus that demand will keep growing. If that consensus cracks, the reversal could be equally dramatic.
Then there's the macro backdrop. The June jobs report showed the economy added just 57,000 positions — roughly half what economists expected. Fed Chairman Kevin Warsh told the world at Sintra that inflation is "too high" and pledged to bring it down to 2%. The next Fed meeting is July 28, and no one — including Warsh — is saying which direction rates are going.
The Setup for the Second Half
When the bell rings Monday morning, investors will start pricing the second half of what has been the most extraordinary year for AI stocks in history.
Nvidia's stock is down more than 11% from its recent highs, trading around $198 — a far cry from the levels where it started the month. For a company generating $51 billion per quarter from data centers, the pullback has brought valuations closer to earth without making them cheap.
The bulls point to GB300 shipments, $190 billion in projected data center revenue, and Jensen Huang's unwavering conviction. The bears point to two semiconductor selloffs in eight days, "excess capacity" fears, and a global economy that may be slowing faster than anyone expected.
The AI trade's first half was a rocket ship. Its second half begins now — and Nvidia is in the cockpit.
Found this helpful? Share it with others.