What’s the Deal With Circular AI Deals?

Summary

WSJ Tech News Briefing

Episode: What’s the Deal With Circular AI Deals?
Date: October 28, 2025
Host: Bell Lin
Guests: Katie Bindley (WSJ Tech Reporter), Jonathan Weil (WSJ Heard on the Street Columnist)

---

Episode Overview

This episode explores two interwoven themes:

1. San Francisco’s resurgence following the pandemic downturn—focusing on the city’s economic and social recovery, fueled in part by the AI boom.
2. The rise of “circular deals” in the AI industry—unpacking how major tech companies (OpenAI, Nvidia, Microsoft, AMD, Oracle, etc.) are entangled in mutual mega-investments and product purchases, and assessing the risks and stakes of this trend.

---

San Francisco’s Rebound: From ‘Doom Loop’ Fears to AI-Driven Recovery

The Pandemic Years and the ‘Doom Loop’ (01:38)

• Katie Bindley recounts how the pandemic emptied San Francisco:
  • Remote work decimated “the high concentration of knowledge workers” in the city.
  • Commercial real estate and rents declined sharply, crime and visible homelessness increased.
  • “We technically weren’t in a doom loop... but people had concerns the city was in a pretty bad place.” (Katie Bindley, 01:38)

Signs of Recovery (02:39)

• Return to work: More workers are back in offices multiple days a week.
• AI effect: “You have more people wanting to be here as a result of a lot of the AI companies.” (Katie Bindley, 02:41)
• City initiatives: Starting in 2023, policies addressed empty storefronts and eased business permitting.
  • Example: A coffee shop founder succeeded with a lease-sharing agreement despite naysayers. “So far feels like it’s been a fantastic decision.” (Katie Bindley, 03:23)

Progress on Crime and Homelessness (04:00)

• Crime rates, especially burglaries, are down 28% in 2025.
• Visible homelessness has decreased due to tent encampment removals.
  • “Some residents have definitely noticed improvements on both.” (Katie Bindley, 04:10)

Ongoing Challenges (04:21)

• Shortage of affordable housing and persistent issues with the drug crisis remain high on the mayor’s list.
  • “He pointed in particular to affordable housing shortage... and the issues tied to the drug crisis.” (Katie Bindley, 04:27)

---

Circular AI Deals: Hype, Risk, and Historical Echoes

What is a Circular Deal? (05:48)

• Definition: Company A pays Company B; Company B then buys from Company A—often relying directly or indirectly on funds provided by the initial investment.
  • “Without that initial transaction, company B might not be able to make the purchase.” (Bell Lin, 05:51)
• Current context: This dynamic is widespread among AI giants.

Examples in AI (06:32)

• Nvidia & OpenAI:
  • OpenAI announced a deal to buy millions of Nvidia chips.
  • Nvidia promises to invest $100 billion in OpenAI—possibly enabling OpenAI’s purchases.
  • “Is that a quid pro quo? Maybe, maybe not... If the $100 billion from Nvidia doesn’t come through, then maybe whatever OpenAI’s hoping to buy from Oracle, it may be less.” (Jonathan Weil, 06:40 & 07:17)
• Downstream ripple: OpenAI’s future deals—like potentially spending $300 billion with Oracle—may hinge on proceeds from such investments.

Dot-Com Bubble Comparisons (07:54)

• Historical parallel: The dot-com era’s telecom infrastructure build-out relied on similar circular financing—vendors lent money so buyers (customers) could purchase their products.
  • Lucent and Windstar are cited as cautionary examples: “Lucent lent as much money as Windstar needed for a long time for Windstar to be able to buy Lucent’s stuff.” (Jonathan Weil, 08:27)
  • When demand collapsed, both companies suffered heavily.
• New twist: AMD is offering OpenAI warrants for up to 10% ownership to secure chip sales—a highly unconventional move for an established public firm.

Best-Case and Worst-Case Scenarios

Best-Case: Massive AI Demand Materializes (09:35)

• For tens or hundreds of billions of investment to be sustainable, similarly massive customer demand must emerge—beyond just “average retail people paying $20 a month.”
  • “If you’re talking about trillions of dollars of infrastructure..., the best case scenario is you need like trillions of dollars of cash flow.” (Jonathan Weil, 09:44)
• Skepticism prevails over whether this is realistic: “That’s where the biggest source of skepticism is.” (Jonathan Weil, 10:25)

Worst-Case: Pullback Triggers a Cascade (10:38)

• If one or several major investors decide the returns aren’t coming and cut back spending, others may follow, rapidly deflating the AI ecosystem.
  • “That would be a scenario with a lot of ripple effects.” (Jonathan Weil, 11:06)

---

Notable Quotes & Memorable Moments

• “We technically weren’t in a doom loop... but people had concerns the city was in a pretty bad place.”
  — Katie Bindley, (01:50)

• “Is that a quid pro quo? Maybe, maybe not.”
  — Jonathan Weil, (06:40)

• “The notion that you would have an established large, mature publicly traded company giving away 10% of itself in the form of warrants, that’s a new twist.”
  — Jonathan Weil, (09:21)

• “That’s where the numbers are kind of blowing people’s minds... it’s hard to get your head around the notion that there’s going to be those types of future cash flows.”
  — Jonathan Weil, (10:17)

---

Key Timestamps

• San Francisco’s past reputation: 01:38
• Factors driving city’s recovery: 02:39
• Crime and homelessness update: 04:00
• Lingering city challenges: 04:21
• Explanation of circular deals: 05:48
• AI mega-deals and examples: 06:32
• Dot-com parallels and AMD’s “new twist”: 07:54–09:21
• Best-case and worst-case scenarios: 09:35–11:06

---

Tone & Language

The episode uses clear, accessible language, balancing skepticism with optimism. Both Katie Bindley and Jonathan Weil offer context and analogies that help listeners understand why the AI industry’s current deal-making matters, and how it could shape the market—and cities like San Francisco—in the years to come.

---

For listeners interested in the intersection of tech investment, urban economics, and market risk, this episode distills a complex, fast-changing landscape into clear takeaways about today’s AI ecosystem and its echo with tech’s past booms and busts.

Transcript

[Katie Pizzolato]

No longer a far off abstraction, quantum computing is becoming top of mind for businesses. And the exploration of how it can be used within industries is expected to increase significantly over the next five years. At the break, join Katie Pizzolato, vice president, IBM Quantum Platform, to learn how companies are beginning to discover how quantum computing could one day solve their biggest challenges.

[Bell Lin]

Welcome to Tech news briefing. It's Tuesday, October 28th. I'm Bell Lin for the Wall Street Journal. San Francisco is back, or at least it's on its way. We look at how the city at the center of the AI boom is coming back from the brink of a doom loop. Plus, with so many megadeals being struck between OpenAI, Nvidia, Microsoft and others, circularity has become one of the buzziest words in the AI vernacular. We dig into the best and worst case scenarios for these deals. But first, not too long ago, it seemed like San Francisco was emptying out as locals decamped for suburbs and cheaper locales. Yet if you look at the city now, things are pretty different. Our tech reporter Katie Bindley tells us how San Francisco is bouncing back. So, Katie, as you write in your story, San Francisco is showing signs of a rebound. Crime rates have dropped, the homeless tent encampments have shrunk and rents are up. These are all good signs. But let's go back in time a little bit. What was the reputation of the city prior to this recent turnaround?

[Katie Bindley]

Yeah, so basically what happened is you had the pandemic hit. People started working from home. The city has, you know, a high concentration of knowledge workers, people who work in tech, people who could do their jobs from home. And of course, tech was one of the earliest adopters of remote work. And so you just had way fewer people going into the office, fewer people just walking around the streets and everything. Rents certainly declined for a while. The commercial real estate market suffered quite a bit. And then you also had certain types of crime went up at the same time that this was going on and a lot of visible homelessness and drug use. And we technically weren't in a doom loop because there's this economic definition of that. The economists told me back at the time we were not meeting yet, but people had concerns that the city was in a pretty bad place. And then there were a lot of initiatives to start to try to climb out of it.

[Bell Lin]

Katie, what's behind the recent change in San Francisco's fortunes?

[Katie Bindley]

There's several different factors going on. You do have more people who are going back into work more days a week. You have more people wanting to be here as a result of a lot of the AI companies. And then there were a lot of initiatives dating back back in 2023, certainly to try to clean up downtown, just like various efforts to make it so there weren't so many empty storefronts. You know, a lot of restaurants and businesses did close during the pandemic. There've been some permitting changes that people have mentioned. I spoke with, for example, this founder of a coffee shop who he'd opened a coffee shop around the civic center during the pandemic. It didn't last a year. And then he had this opportunity to partner with like a foundation through the SF Chamber of Commerce and to have a lease sharing agreement in a particular storefront downtown. And so he went ahead and did that and he opened about a year ago and so far feels like it's been a fantastic decision. But when he was in the process of making that decision, a lot of people were telling him like, hey, this probably isn't such a great idea, but it turns out it has been a good idea.

[Bell Lin]

Two other issues that you bring up in your story are around crime and drug use, issues that residents have cited as detracting from the quality of life. And have those problems started to recede in San Francisco as well?

[Katie Bindley]

Crime rates definitely are down. Burglaries are down 28% this year. There's also less visible homelessness, and part of that is from the removal of tent encampments. So, yes, some residents have definitely noticed improvements on both.

[Bell Lin]

The bright spots in the city aside, what are the sorts of issues that still remain in San Francisco?

[Katie Bindley]

Yeah, I actually spoke with the mayor about this, and as far as some of the areas that he still feels like there's a lot of room for improvement on, he pointed in particular to affordable housing shortage. And he also mentioned, you know, really wanting to continue to tackle the issues tied to the drug crisis, really. So those were two points that he made that apparently are high on his list for making more progress on that.

[Bell Lin]

Was WSJ tech reporter Katie Bindley. Coming up, bullish investors might think circular AI deals are a win win, but there are plenty of skeptics who suspect that the AI Ecosys is in a bubble. We'll dig into how these deals actually work after the break.

[Katie Pizzolato]

Quantum computers are poised to help businesses drive innovation through more expansive and multidimensional computations, says Katie Pizzolato, vice president, IBM Quantum Platform, which sees a roadmap to these use cases.

By 2029, they unlock a new set of mathematics and algorithms to tackle applications outside the reach of classical computers. Working alone, it really is a fundamentally different way to process information.

Different, but not impenetrable.

It does feel more complex. For sure it is, but it is much more accessible and the barrier to entry is much lower than people anticipate.

[Bell Lin]

What exactly is a circular deal? Broadly speaking, company A pays company B and then company B turns around and buys company A's product or services. But without that initial transaction, company B might not be able to make the purchase. WSJ Heard on the street columnist Jonathan Weil is here to explain how this dynamic is at play in the AI realm and just how interconnected all these AI companies are. John, it sounds like there are plenty of risks to this type of circular deal. So to exemplify it for us, can you describe some of the relationships between the biggest AI players? Right now I'm thinking OpenAI, Oracle, Nvidia, AMD and Core Weave.

[Jonathan Weil]

Well, just take Nvidia and OpenAI, for example. There was an announcement by both companies in late September that OpenAI would be buying millions of chips from Nvidia. And Nvidia was going to invest $100 billion into OpenAI. Now, is that a quid pro quo? Maybe, maybe not. But then the question comes up, well, would Nvidia be able to sell all these chips to OpenAI? Would OpenAI be able to buy all these chips from Nvidia if it wasn't for this promise of $100 billion equity investment? And then that whole relationship enables all sorts of other relationships? Well, now there's $100 billion that potentially down the road, OpenAI may have in the form of an equity investment from Nvidia. Now OpenAI can go out and sign contracts with Oracle to buy $300 billion worth of stuff from Oracle. Where are they going to get the $300 billion? Nobody really knows. Maybe it'll come down the line somehow. But if the $100 billion from Nvidia doesn't come through, then maybe. Whatever OpenAI's hoping to buy from Oracle, it may be less than $300 billion if that $100 billion doesn't come through.

[Bell Lin]

So there are so many contingencies here, which makes me think of the dot com bust. Why are so many folks drawing comparisons between today's big AI deals and what happened during the dot com bubble?

[Jonathan Weil]

The portion of the dot com bubble that people are thinking of or referencing by and large, has to do with the telecom portion of it. There was a lot of infrastructure that needed to be built at that time. To support the creation of the Internet as we knew it. Then there was a lot of fiber that got put into the ground and there were a lot of telecom equipment companies that were selling gear to startup telecom providers that didn't have the money unless they raised it from somebody or borrowed it from somebody. The poster child at that time was Lucent Technologies, and its poster child customer was Windstar Communications. And Lucent lent as much money as Windstar needed for a long time for Windstar to be able to buy Lucent stuff. And then when it became clear that the infrastructure had gotten completely overbuilt for the time and there was not the demand to be able to use all of that at the time Windstar went broke, Lucent came close to going broke, but in the end got bought. And so the similarities that people see are that, well, we've got new build out of infrastructure that's happening for a new technology and it's foreseeable that things could go wrong. But there's a key difference too, and an example of something that I had never seen before. AMD is actually planning to give warrants for up to 10% of itself to OpenAI as an inducement to get OpenAI to be a customer of AMD and buy its chips. The notion that you would have established large, mature publicly traded company giving away 10% of itself in the form of warrants, that's a new twist.

[Bell Lin]

What do you think would have to happen in order for the best case scenario to play out?

[Jonathan Weil]

Best case scenario is that you're going to have massive applications and demand for things like ChatGPT, Claude, Gemini, all of these different AI products. They haven't really even figured out who those customers are going to be or what the demand structure is going to be. It's probably not going to be just average retail people paying $20 a month subscriptions for them. They're going to have to figure out something. But if you're talking about trillions of dollars of infrastructure, the best case scenario is you need like trillions of dollars of cash flow in the form of customer demand to support that. If you don't have that, and this is where the numbers are kind of blowing people's minds, it's hard to get your head around the notion that there's going to be those types of future cash flows on the demand side to support the type of infrastructure that we've had. And that's where the biggest source of skepticism is.

[Bell Lin]

What do you think might trigger the worst case scenario?

[Jonathan Weil]

One day one of the big companies investing hundreds of billions of dollars into developing data centers. Whatever capex is necessary to develop these technologies and products wakes up and says, we've heard from our investors. They said enough. We need to be able to justify the capex. We're cutting back. And then if that led to pressure from all their peers to say, we've got to cut back too, we can only justify the levels of capital expenditures that are going to produce a return for us in the foreseeable future. That happened across the board. That would be a scenario with a lot of ripple effects.

[Bell Lin]

That was WSJ heard on the street columnist John Weil and that's it for Tech News Briefing. Today's show was produced by Julie Chang, with Deputy Editor Chris Insinsley logging off. I'm Bell Lin for the Wall Street Journal. We'll be back later this morning with TNB Tech Minute. Thanks for listening.

[Katie Pizzolato]

The time to develop strategies around quantum computing is now, says Katie Pizzolato, vice president, IBM.

Quantum platform, just like AI, was experimental in its early days and is now foundational. Quantum is moving through a very similar arc, so building quantum literacy is important. Educate leaders and technical teams on the basics and implications of the technology and run experiments and identify where you define value.

Quantum, she says, should be seen as a new part of your ecosystem, not a substitute for it.

Think about how this technology is going to integrate with AI and classical systems. Quantum is not going to replace those things. It will augment them. We've got to find the parts of the workforce where Quantum is Most valuable.

Visit IBM.com to learn how quantum computing will accelerate business innovation and growth.

[Katie Bindley]

Custom content from WSJ is a unit of the Wall Street Journal Advertising department. The Wall Street Journal news organization was not involved in the creation of.