Nvidia has announced plans to invest up to US$100 billion in OpenAI to support the AI lab’s massive build-out of computing power. The deal, tied to OpenAI’s goal of adding 10 gigawatts of capacity, could see the first deployments roll out next year.
BNN Bloomberg spoke with Jay Goldberg, senior analyst at Seaport Research Partners, who remains the only Wall Street analyst with a sell rating on Nvidia, to talk about the deal.
Key Takeaways
- Nvidia plans to invest up to US$100 billion in OpenAI, with deployments starting next year.
- Strategically, Nvidia positions OpenAI as a counterweight to Google, Amazon and Microsoft.
- Critics say the structure resembles vendor financing, raising doubts about true demand.
- Analyst Jay Goldberg cautions that Nvidia’s returns could fall short if OpenAI fails to grow at extraordinary levels.
- Nvidia’s valuation, at nearly US$4.4 trillion, may limit upside as capacity constraints remain.
Read the full transcript below:
ANDREW: Nvidia is investing up to US$100 billion in OpenAI, with the first deployments of this equipment expected next year. Our guest has questions about this deal, as do many other observers. He is the only analyst on Wall Street with a sell rating on Nvidia. We’re joined by Jay Goldberg, senior analyst at Seaport Research Partners. Great to see you, and thanks very much. You believe part of this deal is that Nvidia wants to bolster OpenAI as a player because it is concerned the likes of Amazon and Google might turn out to be fickle friends. They, in turn, are wary of Nvidia’s power.
JAY: That’s right. For about the last 10 years, data centre semiconductors have been dominated by the buying power of a handful of companies — the hyperscalers, Internet giants like Google, Amazon and Microsoft. They’ve had a lot of negotiating leverage against the industry. That has shifted with AI. Now Nvidia is very much in the driver’s seat. They have the scarce resource that everybody wants. One way to view this deal with OpenAI is strategic: Nvidia is positioning OpenAI as a counterweight to the power of those incumbents. It’s an interesting move.
ANDREW: That’s interesting, because down the road, if these incumbents do develop their own chips or find alternative sources, Nvidia would want OpenAI as a powerful ally.
JAY: That’s right. All of those incumbents are developing their own chips. They’re working with Broadcom to bring out their own semiconductors. That’s really the only competitive threat to Nvidia at the moment — hyperscaler custom silicon. OpenAI appears to be going very much in the other direction, leaning on Nvidia almost exclusively.
ANDREW: OpenAI will be massive, though, presumably, if all this investment and spending goes through.
JAY: That’s a big if. Where will they get the money? They’ve signed up with Oracle for a US$300-billion deal. Nvidia is bringing US$100 billion. Where’s the rest going to come from? I think they can raise it — there’s a lot of interest in the sector. But it’s still hard to raise this much money. It will be interesting to see how they do it and what concessions are required to get the deal done.
ANDREW: And you say it could be tough for Nvidia to justify this massive investment if the return isn’t high enough.
JAY: I see this as somewhat circular. They’re providing finance and creating their own demand. It’s basically: “Here’s US$100 billion, OpenAI, now you can use it to buy our product.” We see this kind of thing in bubbles. I’m not saying there’s anything wrong with the accounting, but it raises questions about the true level of demand and what multiple we assign to Nvidia revenue when they are effectively paying for it.
ANDREW: It does seem kind of circular — I give money to you as long as you buy my equipment.
JAY: That’s right. This kind of thing works until it doesn’t. We’ve seen this in past cycles. Companies provide vendor financing, and it works on the upswing. But when there’s a downturn — and there always is one — it amplifies the decline. When you’re going up, it’s good. When you’re going down, it’s very bad.
ANDREW: You have a sell on Nvidia, which is unique. Your thesis has yet to play out. Why are you pessimistic on the stock?
JAY: I think of it more as an underperform. Nvidia’s upside is largely priced in. The Street has a good sense of how many chips they can produce, given capacity constraints at TSMC. We know what Nvidia’s Blackwell output should look like. Once you’re sold out, and everyone knows the number, what’s the upside? It doesn’t seem like there’s much left this year.
ANDREW: With a valuation of almost US$4.4 trillion, is it outlandish?
JAY: It’s a big number, but the one big difference between this AI cycle and the dot-com bubble in the 1990s is the multiple isn’t that high yet. So there is some room to grow. But it’s getting stretched. I think Nvidia will underperform the sector. Longer term, there are questions about AI adoption and demand. I think AI will be fundamental to our lives, but we don’t yet know what that looks like. And historically, disruptive technologies are adopted in waves, not in a straight line. At some point, we’ll pull back.
ANDREW: Jay, thank you very much for joining us.
JAY: Thank you.
ANDREW: Jay Goldberg, senior analyst at Seaport Research Partners.
---
This BNN Bloomberg summary and transcript of the Sept. 23, 2025 interview with Jay Goldberg are published with the assistance of AI. Original research, interview questions and added context was created by BNN Bloomberg journalists. An editor also reviewed this material before it was published to ensure its accuracy and adherence with BNN Bloomberg editorial policies and standards.
