Nvidia has sold its output for this year, and on Jay Goldberg's account probably for next year as well.
Most of the market reads a sold-out order book as proof of demand. Goldberg reads it as the end of the upside, which is why he has a sell rating on the largest company in the semiconductor industry while calling this the strongest chip cycle of his career.
"And once you're sold out, what's the upside?"
Goldberg covers semiconductors at Seaport Research Partners, and the rating he carries on Nvidia is one almost nobody else on the sell side does.
The full segment is covered here so you can skip it.
Here are the 4 insights that matter.
👤 Guest: Jay Goldberg, who covers semiconductors and electronics at Seaport Research Partners and holds a sell rating on Nvidia
🎙️ Hosts: Tom Keene and Paul Sweeney, who anchor Bloomberg Surveillance
🧩 Other segments: Steven Major, Global Macro Advisor at Tradition; Alicia Levine, CIO of BNY Wealth; and David Rosenberg of Rosenberg Research
📰 Published: 16 September 2026 on the Bloomberg Surveillance feed
🟣 Apple Podcasts | 🔗 Episode page | ⏱️ length not available
Key Takeaways
He calls this the strongest semiconductor upswing any of us have seen, and still says the industry is cyclical
He can follow the argument that structural change has smoothed the cycle; he does not accept it
His sell on Nvidia is a relative call, not a short — he expects it to lag the rest of the AI complex
Being sold out for this year and probably next removes the source of upside surprises
He expects coordination on AI safety whether or not governments impose it, because the cybersecurity risk is already demonstrated
Moore's law has slowed considerably, and the two-nanometer Apple chip is evidence of effort rather than of pace
1. The Biggest Upcycle Yet
Keene asked how the semiconductor business differs from the boom-and-bust stereotype the sector carried for decades. Goldberg said the stereotype still holds, with one qualification.
"Semiconductors are still cyclical. They go up and they go down. This time it may be a while before it goes down again. I think this is the strongest semiconductor upswing cycle any of us have ever seen."
He gave the this-time-is-different case a fair hearing and declined it. "And I think there are some people who think that maybe this time is different." There are, he said, some signs of structural changes in the industry that will even out the cyclicality in future.
His own position: "I'm not quite there yet, but I can see their reasoning."
2. The Sell on Nvidia
Asked to walk through the logic of a sell rating on the company that sits in the middle of every diagram of the artificial-intelligence economy, Goldberg first narrowed what the rating means.
He frames it to clients as an underperform rather than a short: "I'm not telling people to go out and short Nvidia."
The claim is relative. "I think it's just going to underperform the rest of the semis, the rest of the AI complex."
The first reason is size. The company has become very big, so it is harder and harder for anything to move the needle.
The second is the bar. "Expectations are always very high and it's increasingly hard to beat them."
The third is the order book, and it is the one that inverts the usual reading. "And on top of that, they're sold out, right? They're sold out, they said, for this year, probably for next year. And once you're sold out, what's the upside?"
His conclusion is that the stock will simply find it hard to perform, not that the business is in trouble.
3. Guardrails, Not Drama
Asked how he frames the safety debate — the program's shorthand was the Altmans and the Elon Musks of the world calling for guardrails — for institutional clients, Goldberg started with a disclosure.
"Well, just to be honest, I didn't take any investor calls last week because I was in Italy. But I've been following this debate more than I should have."
He said the two largest model developers have already conceded the substantive point. "There are very serious cybersecurity threats capable from their tools. That's already very clear." The companies he named are OpenAI and Anthropic.
From there the question is form, not principle. "And I think some degree of coordination is going to have to take place. And the rest is just debating exactly what that coordination looks like."
The range he sets out runs from statute to trade bodies. "There are some who are calling for full-on government regulation. There are industry standards."
Strip out the existential framing, he said, and this is ordinary industrial practice: "Industries regulate, right?" Keene added that industries standardize, and Goldberg agreed there are lots of ways to do it.
"It doesn't have to be big and ugly, but they're going to have to do something here."
4. Moore's Law Is Slowing
Keene asked about the A20 chip in the new iPhone, made by Taiwan Semiconductor on a two-nanometer process rather than three, and whether it simply represents continued innovation.
Goldberg explained what the numbers refer to — semiconductor manufacturing processes, and the idea that more can be packed into each chip.
That is Moore's law, and he said he is not sure how much longer it runs. "It's Moore's law, this idea that compute gets denser, doubles every two years. That's slowed down considerably."
The advances continue, but the cost of getting them has changed. "We're still getting advances, but we're having to work harder for each one."
"And the Apple chip is remarkable for a lot of reasons, but I think we're having to work ever harder to get those strengths."
Bonus Insights
The setup for the Nvidia question was a picture. The program described the diagrams of the artificial-intelligence economy that show a flywheel of companies doing deals with each other, and noted that Nvidia sits in the middle of almost every one of them.
Keene introduced Goldberg as having huge experience in semiconductors out of Berkeley, and called him hugely esteemed.
The desk's own summary of the cycle, offered while Goldberg was answering, was that it is an incredible massive cycle with growth everywhere.
Goldberg's bottom line is that the strongest chip cycle anyone has seen does not make the best-known chip stock the best investment: Nvidia's problem is that it has already sold everything it can make, and a company with no room to surprise cannot beat expectations that assume it will.
Products, Companies & Tools Mentioned
Nvidia (The sell rating, on the argument that size, expectations and a sold-out order book leave no upside)
OpenAI and Anthropic (The two companies he says have already accepted that their tools carry serious cybersecurity risk)
Apple and Taiwan Semiconductor (The A20 chip and the two-nanometer process behind it)
Seaport Research Partners (Goldberg's firm)
Listen to the full episode:
If this was worth your time, send it to someone closer to the industry than you are.
Get the latest market chatter as it happens: