Broadcom in Early Financing Talks for OpenAI Chips
Read the Portuguese translation →
Author: Bloomberg Television · Watch on YouTube
Video authored by the channel shown, not by IANIA; all rights belong to the creator. Transcript and summary generated automatically; the original-language content is preserved.
Original language: English
Original video: Broadcom in Early Financing Talks for OpenAI Chips · Channel: Bloomberg Television · 02:42 · Watch on YouTube
Topics: AI infrastructure and hardware, Markets and economy
Summary
Section titled “Summary”Broadcom has held early talks about raising $30 billion in debt to help OpenAI pay for custom AI chips the companies are developing together. Bloomberg tech editor Sarah Fryer describes the high costs of building AI infrastructure, including developing and buying chips, building data centers, and securing energy. She says OpenAI needs financing to fund growth, while demand, chip supply, and costs are driving computing-capacity plans higher. Fryer says the financing boom has not yet hit its limits, but questions remain about whether private debt deals and public markets can sustain it.
Key points
Section titled “Key points”- Broadcom has held early talks about raising $30 billion in debt for OpenAI’s custom AI chips.
- Broadcom and OpenAI are developing the chips together.
- Building AI infrastructure requires ongoing chip development and purchases, data centers, and energy resources.
- Fryer describes the business as extremely expensive and says OpenAI needs more financing to fund growth.
- The discussion mentions Apollo and Blackstone and a target of 20 gigawatts of AI computing capacity by 2028.
- Capacity growth is shaped by AI demand, chip costs, limited supply, and increasing computing needs for each training round.
- Fryer says the financing expansion may eventually meet limits in private debt and public markets, but she does not think that point has arrived.
- The speakers note that consumers may not want to pay more for AI products.
Chapters
Section titled “Chapters”| Start | Chapter | Summary |
|---|---|---|
| 00:00 | Broadcom’s financing talks | The segment introduces Broadcom’s early talks to raise $30 billion in debt for chips being developed with OpenAI. |
| 00:32 | The cost of building AI infrastructure | Sarah Fryer explains the recurring costs of chip development and purchases, data centers, and energy, alongside OpenAI’s financing needs. |
| 01:37 | AI capacity, demand, and chip supply | The discussion turns to a 20-gigawatt capacity target and the demand, costs, and supply constraints behind rapid growth. |
| 02:10 | Financing limits and market skepticism | Fryer considers whether private debt deals and public markets can sustain the spending, while noting that she does not think the limits have been reached yet. |
Quotes
Section titled “Quotes”The company has held early talks about raising $30 billion in debt — 00:00
this is an extremely expensive business to build. — 00:32
There just aren’t as many of them to go around. — 01:37
I don’t think that we’re at that point yet. — 02:10
Concepts and entities
Section titled “Concepts and entities”Broadcom, OpenAI, Sarah Frier, Apollo, Blackstone, Anthropic, Google, NVIDIA, financiamento circular, capacidade de computação de IA
Full transcript (original language)
Section titled “Full transcript (original language)”00:00 Let’s talk about the financing and start with that tech story because as we mentioned Broadcom already looking at another massive AI financing deal. This one tied to open AI. The company has held early talks about raising $30 billion in debt to help open AI pay for the custom AI chips that the two companies are developing together. Let’s discuss with Bloomberg tech editor Sarah Fryer. And Sarah it’s another day with this type of deal. Nvidia does this a lot too, the sort of chip financing circular financing concern type deals. Well, what are the details behind Broadcom yet again arranging one of
00:32 these deals? >> to fund the purchase of chips that it’s co-developing with open AI. I mean, we’re seeing this over and over. They’re working also with Anthropic. They’re working also with Google. This is I’ll say it again, an extremely expensive business to build. And you need new chip your chips only have a certain life cycle. You need to keep developing new ones. You need to buy chips to to house in the data centers. You need to furnish the data centers in in different communities. Make sure that
01:04 they are that you have the energy resources. It is a lot. And and this is this is open AI fresh off of deciding or declaring that they’re going to IPO next year. I know that happened a couple months ago. But in in our in our minds, you know, that’s still pretty recent. They have the need for more money for more financing to fund their growth. If they’re not going to IPO this year. >> So they’re working here with Apollo and Blackstone. I’m just looking at my notes here. 20 gigawatts of AI computing
01:37 capacity by 2028. This number keeps changing. Like how much confidence does the market have? I mean, every week it just keeps growing. It’s not even monthly anymore. It’s just going so quick. >> It’s growing quickly and I think we also need to think about the different factors involved in that growth, right? It’s not just the demand in the market for AI. It’s also the cost of these chips. It it’s it’s the the supply calculation. There just aren’t as many of them to go around. They need more and more each training round. They need to get get their product to be better and better. And this is you’re right. It’s going to
02:10 accelerate and then you get to a question of of when does this collide with reality? When are when do we just are we just fundraising until there’s no there’s no more funders willing to do private debt deals and and complicated creative financing deals and the public market can’t withstand it. You know, where when do we get to that point? You know, I don’t think that we’re at that point yet. But it’s certainly something that that we’re looking at. We’re looking at, you know, when people are skeptical of
02:42 public market events or example if there’s an earnings where a company does extremely well, but it’s not good enough as we saw recently. You know, I think we’re going to see a lot more of that this quarter or this earning season’s going to be very telling. >> Yeah, it’s very expensive and then consumers don’t want to pay up for this stuff, which doesn’t help either. Sarah, thank you so much for joining us. Bloomberg’s Sarah Frier.