Jensen Huang drops a number. Why AI hardware came all the way back, and how I'm playing it with sell puts

Written Wednesday morning SGT, using Tuesday's US close.

The number

On September 17, Nvidia CEO Jensen Huang said the company will double the number of chips it sells next year. His exact words were: "I expect Nvidia to sell twice as many chips as this next year as we do this year." He said it at the AI summit King Charles III hosted at Dumfries House in Scotland.

The comment covers all of Nvidia's chips, not just GPUs. That includes the flagship Blackwell and upcoming Vera Rubin GPUs, Grace CPUs, optical networking chips, automotive platforms and Jetson robotics processors.

It also builds on guidance Nvidia already gave. On the August 26 earnings call, CFO Colette Kress laid out a 70% revenue growth forecast for fiscal 2028, which works out to about $673 billion. For context, Nvidia is on track for around $411 billion in revenue in its current 2027 fiscal year, which ends on Jan. 31, 2027, based on Wall Street's average estimate.

One detail matters here. Nvidia doesn't disclose how many chips it sells, so there's no public figure to double. "Twice as many chips" is a unit forecast, not a revenue forecast. Keep that in mind when you see headlines that treat them as the same thing.

Nine days: from selloff to full comeback

The comeback only makes sense if you remember how bad the week before was.

Saturday, Sept 12: Anthropic CEO Dario Amodei called on AI companies to slow the pace at which they advance model capabilities. Elon Musk and Sam Altman said they agreed.

Monday, Sept 14: Warnings from leading AI executives triggered a global semiconductor selloff. At the same time, the 10-year Treasury yield hit 5%, oil was above $100, and markets expected a Fed rate hike. The worry was simple: if the people building AI want to slow down, what happens to chip demand?

The same week: The Fed raised rates by a quarter point to 3.75% to 4.00%. NVDA fell below $210.

Thursday, Sept 17: Huang gave the "double the chips" forecast in Scotland.

Monday, Sept 21: Intel and Arm surged over 12%, the PHLX semiconductor index jumped 4.3%, and AMD closed 10% higher to reach a $1 trillion market cap for the first time. The Nasdaq posted its first record close since June.

Tuesday, Sept 22: The SOX rose another 2.06%, with 27 of its 30 constituents up. Memory led: SanDisk rose 6.82% and Micron gained 5.00%.

NVDA is now back near $230, close to its May record of $235.74. That's the full comeback in the topic title.

Why the comeback happened: four reasons

1. The number answered the scary question.

The selloff was about whether AI demand might slow. Huang's answer was that demand is running ahead of what Nvidia can supply. He explained it by saying that AI contributes so much to so many industries and economies that, in almost every country Nvidia operates in, people want to invest in AI.

2. Supply is the limit, not demand.

Kress said the 70% growth target was constrained by available supply, which means customer demand alone could support even faster growth. Huang said on the same call that the growth rate would be "a lot higher" without those production limits.

3. Memory went along for the ride.

Memory scarcity is the biggest bottleneck Kress identified. That's bad for Nvidia's costs but great for memory makers, which is why SanDisk, Micron, Seagate and Western Digital led Tuesday's rally.

4. Macro got out of the way, and shorts had to cover.

On Tuesday, oil fell to a two-week low as Iran signalled it could reopen the Strait of Hormuz soon, which eased pressure on rates and growth stocks. China is another potential upside: Nvidia's forecasts don't assume any China sales, and Xi Jinping visits the US on Thursday.

The bear case

A fair write-up has to cover what could go wrong.

There's no track record. Bernstein's Stacy Rasgon has noted that Nvidia has never issued guidance this far out. Because this is its first year-ahead forecast, there's no history to show how reliable it is.

Margins are under pressure. Gross margins are expected to bottom at 71% to 72% in the fourth quarter of fiscal 2027, then settle at 72% to 73% in fiscal 2028, as rising memory costs weigh on profitability.

Supply still decides. The stock is still below its May record, and Huang's volume promise now runs into the same supply bottlenecks Kress described.

The Fed is still hiking. The 10-year yield is near 5%, and the AI safety debate hasn't gone away.

The valuation argument, to be fair, is on the bulls' side. Nvidia trades at less than 17 times expected earnings, about half its 2025 multiple.

How I use this in my sell-put strategy

NVDA is one of my two core names, which means it gets my largest position size. It's also where I'm most careful about timing.

The main lesson from this episode is simple. The best time to sell a put on NVDA was during the fear of September 14, not after this rebound. Put premiums are highest when people are scared. Now the stock is back near its high and the fear is gone, so the same trade pays less and gives less margin for error. I don't chase the comeback.

This is how I run NVDA through the checklist tonight:

Gate 1: Trend. The stock must be above its 200-day moving average. This is a hard rule. After a rebound to near the record, I expect this to pass, but I check it on the chart.

Gate 2: Premium. IV Rank of at least 40%, IV Percentile of at least 45%, and implied volatility at least 1.2 times historical. After a rebound like this, these readings usually come down. If they fail, I don't trade, however bullish the story.

Gate 3: Earnings. Nvidia last reported on August 26, so the next report should be in late November (check the confirmed date). My 30 to 45 day window covers expiries from late October to early November, which should clear it with room to spare for the 7-day blackout.

Gate 4: Strike. I use my two-anchor method and take the lower of:

the one standard deviation move below the current price, or

the nearest key moving average × 0.98.

For example, with made-up round numbers: spot $230, IV 35%, 35 days to expiry. One standard deviation is 230 × 0.35 × √(35/365) ≈ $25, so the statistical floor is about $205. I then check the delta. It has to be between 0.15 and 0.25.

Gate 5: Size and exits. One $205 put ties up $20,500 in cash. My core-name cap is 5% of the account per position. The exits go in the moment the trade fills:

a take-profit order to buy the put back at 50% of the premium,

a stop at 2 times the premium,

one roll at most.

Thursday. Xi's visit is an event. It could bring China upside or new export headlines. I'd rather let it pass and sell my puts once the news is out and volatility tells me what it's worth.

The rest of the AI hardware group

AMD gapped up 10% to $1 trillion. I don't sell puts right after a gap like that.

Micron reports September 30, which puts it inside my 7-day earnings blackout from today.

SanDisk is near $1,900 a share, so one cash-secured put needs about $190,000. That breaks my sizing rules before I even look at the chart.

This is the same approach that gave me my NVDA $200 put for the June 20 expiry, which kept 100% of its premium. I sold it when the gates were green and fear was high, not when everyone was celebrating.

My verdict

Huang's number is a strong signal. But it's a unit forecast with no track record, and whether it lands depends on supply. For a put seller, the comeback is already priced in.

I let Thursday pass, run the gates, and only sell when the premium is worth the risk. If the gates are green, the trade happens. If they're not, there will be another selloff like September 14.

Did you buy the September 14 dip, or are you waiting for the next one? Let me know in the comments.

This is my personal trading framework, shared for education. It is not investment advice. Options carry significant risk, including assignment and losses larger than the premium collected. Always verify live chain data and confirmed earnings dates before trading.

$NVDA $AMD $MU $SNDK $SOXX