FEATURED

NVDA Is Down 16% From Its High

May 14. Nvidia hits $236.54. A new all-time high. The company had just reported $81.6 billion in quarterly revenue, up 85% year-over-year. Jensen Huang called it the largest infrastructure expansion in human history. The board signed off on an $80 billion buyback. The quarterly dividend jumped from $0.01 to $0.25 a share. A 25-fold increase. And the stock fell.

Not a lot. Then a little more. Then June happened.

As of June 30, NVDA closed near $198. That is roughly 16% off the May peak. The stock dropped approximately 12% in June alone, which is how it ended up in what market commentators called correction territory. Volume stayed heavy throughout. This was not retail selling. The institutional footprint is all over it.

Here is the thing: the fundamentals did not change. What changed is the market’s read on three converging forces that are all hitting at the same time, and none of them are simple.

First, China is gone from the model. In April 2026, new U.S. export restrictions hit H20 chip shipments. Management confirmed this would cost roughly $8 billion in Q2 revenue. Then in early June, the Commerce Department moved to close a loophole that had let advanced AI chips flow to Chinese companies through offshore subsidiaries. Nvidia’s Q1 FY2027 filing stated it plainly: no Data Center compute revenue from China in the quarter, versus $4.6 billion in Q1 FY2026. Jensen Huang has said publicly that Nvidia has largely conceded China’s advanced AI chip market to Huawei. That number is not coming back. The $91 billion Q2 revenue guide already assumes a China-zero baseline.

Second, Vera Rubin is real, and that is creating a short-term problem. At Computex in late May, Nvidia confirmed Vera Rubin rack-scale solutions are ramping into full production across more than 350 factories in 30 countries, with initial shipments targeting Q3 2026 and volume in Q4. Huang has described the Vera Rubin supply chain as roughly twice the scale of Blackwell’s. That is bullish for 2027. For right now, some customers are holding off on Blackwell orders while they wait to see what Vera Rubin costs and delivers. That hesitation shows up in near-term demand signals before it resolves in the revenue line. Vera Rubin promises up to 10x agent throughput over Grace Blackwell. When a product that good is six months away, buyers pause.

Third, there is a UBS note from late June warning about a potential slowdown in hyperscaler capex growth. Worth watching. Not panicking over. But the entire $91 billion Q2 guide rests on Microsoft, Google, Amazon, and Meta staying on their current AI spending trajectories. Any softening in that language during upcoming earnings calls hits Nvidia’s forward model immediately and directly.

Slight tangent, but it matters: Nvidia also issued $25 billion in bonds in mid-June. Notes ranging from 2028 to 2056 maturities. The company generated $49 billion in free cash flow last quarter, so this is not a distress move. It looks more like Huang locking in cheap long-term capital ahead of what he thinks is a massive multi-year buildout. That is a signal worth noting, even if the market has not fully processed it yet.

What August 26 Is Actually Pricing

Nvidia reports Q2 FY2027 results on August 26 after the close. Confirmed. The analyst community covering 62 analysts is at a Strong Buy consensus with an average 12-month price target near $299, per StockAnalysis. The high target is $500. The low end runs from $180 to $215 depending on the source. At $198, the stock is trading at a meaningful discount to where the consensus expects it to go. Wide gaps like that either close fast or they close slowly while the thesis quietly erodes.

Getting to the $91 billion Q2 guide from $81.6 billion in Q1 requires roughly 11% sequential growth. That math works, but it works cleanly only if HBM4 supply from SK Hynix and Micron holds, Vera Rubin’s production ramp does not hit the kind of delays that plagued early Blackwell, and hyperscaler capex commentary stays firm. Three variables, all moving simultaneously. One stumble in any of them and the $91 billion number gets questioned on the call.

The options market is pricing a move of roughly 8 to 10% in either direction around August 26. Straddles centered near $195 to $200 reflect that. Implied volatility is running elevated relative to recent realized volatility, which is what you would expect when an event this binary is on the calendar. The $200 strike is the line. A sustained hold above $200 shifts dealer hedging flows bullish. A break below $190 opens the path toward the $170 area where longer-dated put open interest has been accumulating.

For traders expecting Q2 to come in at or above the guide, a bull call spread in the $210 to $230 range for the August expiration captures the earnings move with defined risk. If you think the GPU rental pricing softness and the capex slowdown warnings are telling a more structural story, a bear put spread with strikes laddering below $185 defines the downside without the open-ended risk of a short position. The neutral case is the most crowded and the most expensive. Long premium going into a high-IV event is a real cost. Short premium into an event the market is not treating as routine carries its own risk.

Here is where I am at with all of this: the stock is down 16% from a record high set six weeks ago. The company just posted what might be the most impressive single quarter in semiconductor history. The forward guide excludes an entire country’s worth of revenue and is still $91 billion. The analyst consensus is near $299. And the stock is at $198.

The market is not confused about the fundamentals. It is confused about the timing. Does the Blackwell-to-Vera Rubin transition create a temporary gap in demand, or does the upgrade cycle stack on top of existing demand rather than replacing it? August 26 answers part of that question. The part it does not answer is probably the more important one.