Jensen Huang's "Buy at a Discount" Call: Nvidia Says Yes, the AI Sell-Off Says Not Yet

Jensen Huang's call still looks directionally right

In Seoul last month, Huang did not hedge when asked about the rout. After US tech shares dropped Friday on rate-hike fears and South Korea's Kospi tumbled on Monday, he told reporters that the pullback was an opportunity and that investors should be happy to buy at a discount. Hours later, Nvidia and SK Hynix announced a multi-year deal to co-develop next-generation AI memory chips, putting action behind the commentary.

Bulls can still point to that consistency: Huang has argued that the AI infrastructure buildout is early, and the SK Hynix partnership reinforces that view rather than contradicting it.

Why the discount trade has not fully worked

The market reaction was broader than a simple rerating of individual names. SK Hynix plunged more than 10%. Samsung Electronics fell more than 8%, Kioxia more than 15%, and Tokyo Electron more than 9%. The VanEck Semiconductor ETF dropped more than 2%, while AMD and Teradyne fell 5% and 4%. When chipmakers, equipment suppliers, memory makers, and U.S. semis all sell off together, company-specific fundamentals often get drowned out by broader risk aversion.

That is the gap Huang is betting against. He is asking investors to treat the pullback as noise inside a long buildout cycle. So far, the market has treated it more like a warning about financing, expectations, and exposure.

The sell-off reflected financing anxiety more than a clear demand break

Huang's "buy the discount" call works only if the market is overreacting to temporary financing or macro stress, not to a durable slowdown in AI demand. On that measure, the evidence is mixed.

Fundamentals held up even as sentiment cracked

SK Hynix is a useful example. It posted record quarterly profit and revenue, yet its shares still fell more than 10%. That suggests investors were not waiting for another quarter of weak operating results before selling; they were reacting to worries about how AI expansion would be funded.

Those worries showed up clearly elsewhere in the tape. South Korea's market was pushed to its lowest level in three months as investors focused on financing risks tied to AI infrastructure spending and on increasing competition from China.

The same pattern returned later in the summer. Korean chip stocks sold off again as investors revisited concerns about AI financing and Chinese competition, with Samsung and SK Hynix falling as much as 9.5% and 10.9%, respectively, while the benchmark KOSPI was down 7.3% as of 0032 GMT. The takeaway is not that the long-term AI story collapsed. It is that sentiment kept resetting faster than supply or demand did.

Broadcom showed how quickly the pressure broadened

The sell-off also spread beyond memory. Broadcom's softer-than-expected outlook triggered a selloff in AI-linked stocks, chiefly semiconductor names. After that, SK Hynix and Samsung fell again, with SK Hynix down more than 8% and Samsung down more than 5%. Retail sentiment tied to DRAM remained bearish, which helped keep pressure on the broader memory chain, including Micron-related exposure.

For investors, the setup is straightforward:

• If financing panic fades while fundamentals hold, Huang's discount call starts to work more often.
• If those concerns become a more permanent repricing of AI spending, the pullbacks may not look like clear buying windows.

Has Huang's call paid off? Directionally, yes. On timing, not yet.

Huang has been right about the direction of the market, but not yet about the timing.

The clearest sign came from the sector's rebound. The SOXX plunged 10% before rebounding sharply, as renewed confidence in AI infrastructure spending helped reverse part of the damage. That pattern has repeated: fear spreads quickly, then eases when investors again see signs that funding and demand are still intact.

Nvidia sits at the center of that tension. It delivered record fiscal 2026 revenue of $215.9 billion, yet it closed below the $200-per-share mark. Meanwhile, NVIDIA maintains a more attractive valuation at 25.4 times forward earnings versus AMD at 84.4. That does not prove the market is wrong; it does suggest that fear has been forcing sharper resets in price than in reported results.

So Huang's call has been more validated in conviction than in execution. The long-term AI buildout still looks intact. The short-term discount trade, however, remains intermittent.