Nvidia Makes $3.5 Billion Bet on MediaTek

31 Aug 2026 · 43 min · 24 chapters

Ask about this episode

Ask anything about it. ChatGPT or Claude reads this page and answers with the times it was said.

Connect VO and ask about every podcast you hear, including the moments you saved. Add to ChatGPT · Add to Claude

In short

NVIDIA invests $3.5B in Taiwan’s MediaTek via convertible bonds, aiming to bring MediaTek custom AI chips (“XPUs”) into NVIDIA’s AI data-center stack using NVIDIA networking/memory tech (MVLink Fusion) and its “AI factory” architecture. The episode also touches on Tim Cook stepping down as Apple CEO, China’s AI policy demands ahead of U.S. talks, and Anthropic’s IPO timing.

Guests (backgrounds)

Jensen Wang, NVIDIA CEO; Rick Zai, MediaTek CEO; Brad Warden, Nomura Asset Management portfolio manager; Ian King, Bloomberg reporter; Kunjan Sabani, Bloomberg Intelligence; Mark Gurman, Bloomberg Apple reporter; Mike Shepard, Bloomberg reporter; Bailey Lipschütz, Bloomberg Equities reporter; Matt McElwain, Madrona investor; Tori Shivanandan, Radiant COO.

Key claims

NVIDIA says XPUs are already in the market and that GPUs accelerate the full AI lifecycle; MVLink Fusion makes custom silicon easier to deploy inside NVIDIA data-center systems; partnership is “decade long” and not “circular financing”; MediaTek will use bond proceeds for supply chain and more ASIC/advanced packaging capability.

Notable examples

AWS deploying NVLink Fusion with Marvell; MediaTek’s July plan to use $5B in bonds for supply chain; Apple iPhone roadmap continuity under John Ternus; China singling out Anthropic for alleged anti-China behavior.

Written by AI. May contain mistakes. Listen to the episode to check what was said.

Chapters

Tap a time to open that second in VO

NVIDIA's $3.5 Billion Investment Explained

0:36 to 1:05

Explore NVIDIA's significant investment in MediaTek and its implications.

“It's designed to help you move from a chaotic starting point to a reviewable first version.”

NVIDIA's $3.5 Billion Investment Explained

2:32 to 4:10

Explore NVIDIA's significant investment in MediaTek and its implications.

“NVIDIA is making another multi-billion dollar bet on the AI ecosystem, investing$3.5 billion in Taiwan's Mediatek.”

MediaTek's Strategic Plans with NVIDIA

4:11 to 6:39

Discussion on MediaTek's use of funds for supply chain and technology enhancements.

“Earlier in our conversation, you talked about the benefit of working with NVIDIA on MVLink Fusion is speed to market.”

Integration of XPUs into NVIDIA's Architecture

6:40 to 8:30

Insights into how XPUs and NVIDIA’s architecture work together in data centers.

“It's not coming into the market, it's in the market.”

Understanding MediaTek's Role in the Market

8:31 to 10:21

A look at MediaTek's position in the tech industry and comparison to U.S. companies.

“isn't just how good are the individual components, it's how good do they all link together.”

Custom Silicon and Its Impact on AI

10:22 to 14:00

Analysis of the trends in custom silicon development and their implications for AI.

“NVIDIA's CEO, Genswan, also talks about how the specialist platforms are largely inference platforms.”

NVIDIA's Strategic Partnerships

14:00 to 16:25

Learn about NVIDIA's strategy in leveraging partnerships for technology advancement.

“And so there's a place for all of these.”

Sponsor: Superhuman Go

16:25 to 18:29

Discover how Superhuman Go can enhance your productivity with AI.

“Namira Asset Management International Senior Portfolio Manager, Brad Warden.”

Sponsor: Superhuman Go

18:36 to 19:32

Discover how Superhuman Go can enhance your productivity with AI.

“You already know how AI is changing how everyday work gets done, how much ground you can cover, and how fast a team can scale.”

China's AI Guidelines and Anthropic

19:32 to 23:26

Understand China's stance on AI guidelines and its implications for Anthropic.

“First up, shares of Shein on the gray market are under pressure ahead of the fast fashion giant's Hong Kong debut Tuesday.”
Show all 24 chapters

Transition of Apple's Leadership

23:26 to 26:47

Explore the leadership changes at Apple and their potential impact.

“Bloomberg's Mike Shepard with the reporting.”

NVIDIA's Investment in MediaTek

26:47 to 28:00

Analyze NVIDIA's significant investment in MediaTek and its implications.

“Generally speaking, technology stocks are lower.”

NVIDIA's Role in AI Factories

28:00 to 29:26

Learn about NVIDIA's diverse chip offerings and their integration in AI factories.

“And now the amount of chips, there's like seven different types of NVIDIA chips that has to go into an AI factory.”

Anthropic's IPO Impact

29:26 to 30:52

Explore the implications of Anthropic's IPO on the broader market and other companies.

“Anthropics IPO plans are casting a bit of a shadow over other companies' U.S.”

Investor Trends and IPO Timing

30:52 to 32:34

Understand the current investor sentiment and the timing of upcoming IPOs.

“And in your report, you detail that like a lot of those other IPOs or potential IPOs that you're mentioning are finding it hard to get the attention of some investors.”

Madrona's Annual AI40 List

32:34 to 34:29

Discover Madrona's AI40 list and the funding landscape for AI companies.

“Madrona has released its sixth annual IA40 list of the top private companies building and enabling AI.”

Durability vs. Disruption in AI

34:29 to 36:28

Examine the challenge of durability versus disruption among AI companies.

“But of that$410 billion, just three names account for most of it.”

The Role of Builders in AI

36:28 to 38:27

Learn about the importance of builders and collaboration in AI development.

“That is the core question that's increasingly being asked by enterprises, by the businesses that leverage all these tokens.”

Impact of Anthropic's IPO

38:27 to 39:10

Analyze how Anthropic's IPO could influence future IPOs in the tech landscape.

“Matt, you heard Bailey's reporting there very quickly.”

Impact of Anthropic's IPO

40:39 to 41:14

Analyze how Anthropic's IPO could influence future IPOs in the tech landscape.

“ChatGPT Work is a new way of working in ChatGPT that can take action across your apps and files, stay with a project for hours if needed, and turn a goal into finished work.”

Impact of Anthropic's IPO

41:21 to 42:02

Analyze how Anthropic's IPO could influence future IPOs in the tech landscape.

“You already know how AI is changing how everyday work gets done, how much ground you can cover, and how fast a team can scale.”

Tim Cook's Transition at Apple

42:25 to 43:32

Discussing Tim Cook's final day as CEO and transition to the executive chair role.

“Tim Cook says he has enormous comfort with John Ternus taking the helm as CEO and said that he will miss this work as he has his final day as CEO of Apple.”

Radiant's Nuclear Micro-Reactors for the Military

43:34 to 47:00

Tori Shibandan explains Radiant's innovative portable nuclear reactors and their military applications.

“Radiant president and chief operating officer, Tori Shibandan, joins us now.”

NVIDIA's $3.5 Billion Investment in MediaTek

47:07 to 50:09

Analyzing NVIDIA's investment strategy and its implications for the tech industry.

“NVIDIA investing$3.5 billion in Taiwan's Mediatek.”
Hear the part that matters, and keep it.Open this episode in VO. Double tap your headphones to save a moment as you listen.
Get VO free

Transcript

Automatic transcript. May contain errors.

0:00The most effective people at work aren't working harder than everyone else. They're working smarter, inside better systems. Superhuman Go, from the makers of Grammarly, is the AI assistant that works inside every tool you already use, always ready, already aware of what you're working on. It's a teammate whose only job is to help you be better at yours. With Go working with you, you can show off what you do best. See what Superhuman Go can do at superhuman.com. Some people treat ChatGPT like some kind of smart search engine, and some use it to get work done. ChatGPT Work is a new way of working in ChatGPT that can take action across your apps and files, stay with a project for hours if needed, and turn a goal into finished work.

0:46It's designed to help you move from a chaotic starting point to a reviewable first version. So all the source materials, briefs, and scattered information that you have to grind through to turn into something useful can just become something useful. Put ChatGPT to work on your most ambitious ideas and projects. Get started at ChatGPT.com by selecting Work Mode, available on Plus and Pro plans. At Venture Global, we think about what can be done, not what's usually done. Through innovation, Venture Global is not only building some of the largest energy facilities in the world right here in the United States, but delivering American energy at a fraction of the cost and a fraction of the time.

1:31So while others are busy talking, we're busy building. That's Venture Global. That's unstoppable energy.

1:43Bloomberg Audio Studios. Podcasts, radio, news.

1:54Bloomberg Tech is live from the heart of Silicon Valley with Ed Ludlow in San Francisco.

2:04Ed Ludlow:This is Bloomberg Tech. Coming up, NVIDIA makes a$3.5 billion investment in media tech, deepening its collaboration with the Taiwanese chip maker. We'll hear from both CEOs. Plus, today is Tim Cook's last day as CEO of Apple, before John Ternus takes the reins tomorrow. And all eyes on Anthropik as it prepares to file publicly in the coming weeks. This is other companies' either own US listing plans in the second half of the year. Okay, let's get straight to our top story. NVIDIA is making another multi-billion dollar bet on the AI ecosystem, investing$3.5 billion in Taiwan's Mediatek. The partnership brings Mediatek's custom AI chips into NVIDIA's architecture, plugging so-called XPUs into NVIDIA high-speed networking and memory technology.

2:52Ed Ludlow:The deal could also revive concerns about circular financing as NVIDIA invests across its own ecosystem. I sat down with NVIDIA CEO Jensen Wang and MediaTek CEO Rick Zai earlier today. This is not circular because obviously they do their own business and we do our own business. And MediaTek is already incredibly profitable, incredibly successful. They are the world's best SOC maker. They make more SOCs than any company in the world. It's a great pleasure for us to invest in a technology company in Taiwan. It's a great pleasure for us to invest in MediaTek. This is going to turn out, of course, the investment is grounded on a deep partnership that's going to be multiple years.

3:37This is a part, the roadmap that we're executing on right now is a, it's not one or two years. It's a decade long. And so this partnership is extremely important to us. And we want to make sure that we have we demonstrate every opportunity, every take every opportunity to demonstrate the importance of it, the seriousness by which we take it, the confidence we have in the partnership and the investment that we're making here demonstrates that. But ultimately, this is going to be a phenomenal investment for us. It's going to generate incredible returns. I have every confidence.

4:10Ed Ludlow:Rick, in July, MediaTek said that it would use its bond program,$5 billion of dollar bonds, basically to support and grow supply chain. Earlier in our conversation, you talked about the benefit of working with NVIDIA on MVLink Fusion is speed to market. What do you plan to do with the proceeds? And when you talk about supply chain, do you need to do more on Taiwanese suppliers or U.S. suppliers in the context of this relationship? I think the, certainly, well, back in the end of July, we would not, we were not able to discuss the strategic investment part of the bond. So, but the proceeds, I think, will be for multiple purposes.

5:00Supply chain definitely being one of them. And the supply chain is a global. I mean, whoever needed to enhance and enable the customers. But on the other hand, I think, as Jensen said so eloquently, the investment that MediaTek is a tech company, it's a very strong tech company, and we will definitely invest, if anything, more into the technology, in our capability to design, in our capability to make the ASIC chip in the IPs for interconnects, in advanced packaging. I think the key here is really the collaboration agreements that we have built and expanded with Somalia, as described in our press release.

5:55Those are the key things. The proceeds will basically go in to make those things happen better and further.

6:04Ed Ludlow:Jensen, on the earnings call last week, you talked about, with confidence, right, that the GPU-based architecture has superior economics. You were talking about that in the context of the frontier labs, largely because you were asked about custom silicon. You know, the simple terms of it is that MVLink Fusion just makes it easy to take XPU and just drop it into the data center. That would show that, you know, NVIDIA does see XPU coming into the market. Just the difference of your thinking this week to what you outlined last week. I see XPUs in the market. It's not coming into the market, it's in the market.

6:44Look, an XPU is a specialized chip. But NVIDIA's GPU is a general purpose accelerator. We accelerate the entire lifecycle of AI from data processing, which data processing to pre-training to post-training to agentic AI inferencing. We accelerate every single AI model in the world, from closed models, every closed model, every open model, video models and language models, biological models and physics models and robotics models. We literally accelerate every single area of AI. That explains why it is that we are the most fungible and the most durable and therefore the most rentable GPU, the most rentable computer infrastructure in the world.

7:36We're in every cloud, we're on-prem, we're at the edge. And so we're able to address markets that, frankly, nobody else can because of the architecture of our company and the entire full-stack AI factory platform that we have.

7:54Ed Ludlow:Let's get more on the story with Bloomberg's Ian King, who helped me break it this morning. It's really interesting. For some time now, Gentleman's been asked about, what about all these custom chips that hyperscalers are working on, that the Frontier Labs are working on? And to this point, he's been a little bit dismissive of the threat. Today, with MediaTek, they found a mechanism for those same AI companies to drop an XPU or a custom chip into an existing data center to work with NVIDIA's technology. What else do we need to know here? Yeah, I mean, taking a step back and looking at this from a technology perspective, the way that the data center works and how effective it is isn't just how good are the individual components, it's how good do they all link together.

8:37And what NVIDIA is saying is, hey, use our link, use this fabric, they call it, to bind everything together, and it'll all work better. So what he's trying to do is, even if you want to go it alone, even if you want to use your own stuff, use it inside our framework, Use it inside our house and that'll help you. And of course, that means in terms of business that, look, hey, if companies want to do their own thing like Amazon did with that announcement we saw during earnings, they're going to do it. And they're still going to be using at least some of his chips. So it works business wise.

9:09Ed Ludlow:We haven't said it to this point, so I'll say it now for the audience. This is NVIDIA buying convertible bonds that convert into equity. $3.5 billion, we confirmed, is NVIDIA's biggest non-US direct investment. So kind of interesting, but probably talk a little bit more about MediaTek, right? What is MediaTek? What is it comparable to here in the U.S.? Yeah, no, that's a very good question. This is a very important company that nobody really pays that much attention to. Think they're the Taiwanese equivalent of Qualcomm. They make smartphone processors, smartphone modems. You know, the real guts of that, they have a large presence in the China market and large presence worldwide.

9:47Why they, by numbers, they're one of the biggest companies. And that's really what they've done, and they're looking to expand.

9:53Ed Ludlow:But in ASICs, they're more like Marvell or Broadcom, aren't they? That's right. This is what they're trying to do. They're trying to basically take that expertise that they've developed in smartphones and expand that into the data center, which is a more lucrative business. And what they're doing is trying to replicate what Broadcom does, what Marvell does, which is help, say, Google, say, Amazon. Hey, you've got this great idea for a chip. You've got these specifications. We'll get you all the way there. Bloomberg's Ian King, who broke the story with me this morning, and it's one of the most read today on the Bloomberg platforms.

10:21Ed Ludlow:Go and check out the details. NVIDIA's CEO, Genswan, also talks about how the specialist platforms are largely inference platforms. But as you heard, NVIDIA is across every phase. Listen to this. An XPU is a specialized chip. NVIDIA's GPU is a general purpose accelerator. We accelerate the entire lifecycle of AI from data processing, which data processing to pre-training to post-training to agentic AI inferencing. We accelerate every single AI model in the world. Let's get more on NVIDIA. Brad Warden, Senior Portfolio Manager at Nomura Asset Management International joins us. And quietly, Brad, in advance, you outline the different phases of AI.

11:10Ed Ludlow:You've heard from Jensen there about why they did this with MediaTek to allow XPUs to come into the ecosystem. What does that reflect and which phase of AI you think we're in right now? Right. Thanks, Ed, and thanks for having me this morning. I think, you know, if you think about the phases of AI, right, we've built this foundation, right, which was, you know, I think we started thinking about GPUs, right, as the base of that, really saw that part of the ecosystem develop. And then a lot of things develop around that. And I think that's what you're seeing right now. Right. So you've had this foundation that's gotten built for AI.

11:47Now we've kind of moved into this phase. Right. Of monetization. Right. So now and a lot of the questions, I think, which go back to what you were outlining earlier, just around financing and thinking about what sort of are the next steps of that. And I think if you really step back and think about the relationship that was announced that you talked about, really is this idea of basically expanding the ecosystem. And the strategic aspect of what NVIDIA is doing is making sure that they play across that ecosystem. And you've seen it in a number of different investments that they've made. But I think this one obviously extends some of that.

12:28Yeah.

12:29Ed Ludlow:Brad, I really value your perspective and reaction here, because in the Nomura Transformational Technologies ETF, the top holdings are NVIDIA, AMD, TSMC, Seagate Land Research, but also Broadcom's in there as well, right, on the custom ASIC side. What's the biggest motivation right now that you're seeing from those pursuing custom silicon? The economics on a dollar per token basis or diversity of supply is a kind of like safety net. That's a fantastic point. I think if you look at each of the adoption that each of these big hyperscalers are doing in terms of looking at XPUs, looking at ways to really optimize that infrastructure, which fundamentally, if you look at technology, it's the cost per bit of output.

13:15Right. So in this case, it's token per watt, if you want to think about that. And if you know your infrastructure and your operating requirements, the best out of anybody, which these hyperscalers do, their ability to optimize that and do it in their own silicon, I think, is something that's very, very powerful. Now, there are a lot of reasons why you use a general compute, a GPU as well. And so both of those are valuable. So how you leverage that, again, goes to optimizing that infrastructure. And if we think about it, too, recently we've been talking about open source versus closed source models and how those both develop.

13:53And I think the idea is that you optimize the usage of that. And that's what technology does is over time we figure out ways to optimize. And so there's a place for all of these. And it goes back to the strategy of being able to touch all of those points. And I think that was the point Jensen was making in your interview was really like, look, our customers are asking for all these different architectures and capabilities. And if we can be a partner in that and work on that, that's an important strategic move on the part of NVIDIA to stay at the heart of that and to make sure that their advantage stays there.

14:27Plus, they get a view into how the whole ecosystem is developing at every touchpoint that they've been involved in. And that's critical from a competitive standpoint and very smart from a strategic perspective as well.

14:40Ed Ludlow:So that ending there, interesting. Where do you sit on the concern or lack of concern in what is or is not circular financing in this case? I think it's something that's worth watching, right? I mean, we're watching very closely the amount of leverage that's being put on the cash flows. But we're watching the monetization, right? We're watching very carefully the revenues that are accruing from those types of financing. And I think the market, too, is clearly skeptical, right? Because we've seen really powerful earnings power out of a number of these companies and bidding, including the ones we've just talked about.

15:13But but really the valuations continue to fall. Right. And part of that is that skepticism around that and the circularity.

15:21Ed Ludlow:You watch the CDS. May I just jump in and ask? I'm sorry, we'll run out of time. In this MediaTek case study, the argument that both CEOs made was, in a custom ASICs context, you can get it into the real world quicker. You've got your custom accelerator. You just drop it into an existing architecture. But the counterargument to that was, well, NVIDIA has been able to find its own demand, therefore. I'm trying to work out the real world mechanics of how that would go. Well, I'm not sure specifically. I think what this does is it does definitely speed up the capability to pull that in from a customer standpoint.

16:01And I think that is the idea is the speed to market and the speed with which you can connect all these systems together is becoming more critical because we're moving at light speed, right, in terms of how this is happening. So as we do this, it's actually a very smart thing.

16:16Ed Ludlow:I'm enjoying this. Some people wake up and talk about monetary policy and inflation expectations, and we wake up and talk about compute and architecture. Thank you very much indeed. Namira Asset Management International Senior Portfolio Manager, Brad Warden. Coming up, China laid out the terms for AI talks with the U.S. and accused anthropic of bad behavior. We've got more on that story next. This is Bloomberg Tech.

16:46The people who seem to get more done than everyone else. They're not working longer hours or running on more caffeine. They've just stopped wasting time on the stuff that doesn't move work forward. Switching apps, re-explaining context, hunting for files. Those aren't small inefficiencies. They're hours wasted every week. Superhuman Go gives you those hours back. From the makers of Grammarly, Go is an AI assistant that sits inside every tab and tool you already use, always available and already aware of what you're working on. Ask it to draft something, summarize a long thread, pull up a file, or prep you for a meeting.

17:27Go handles it without you ever leaving the page you're on. This is what it looks like when AI actually fits into your work, instead of adding to it. It's like having a teammate whose only job is to help you be better at yours. Go keeps up so you can move forward. With Go working with you, you can show off what you do best. Superhuman Go. Find out more at superhuman.com. That's superhuman.com. Some people treat ChatGPT like some kind of smart search engine, and some use it to get work done. ChatGPT Work is a new way of working in ChatGPT that can take action across your apps and files, stay with a project for hours if needed, and turn a goal into finished work.

18:10It's designed to help you move from a chaotic starting point to a reviewable first version. So all the source materials, briefs, and scattered information that you have to grind through to turn into something useful can just become something useful. Put ChatGPT to work on your most ambitious ideas and projects. Get started at ChatGPT.com by selecting Work Mode, available on Plus and Pro plans. You already know how AI is changing how everyday work gets done, how much ground you can cover, and how fast a team can scale. To stay ahead, you need the tools that give you a competitive advantage built for this new era.

18:47Welcome to Agentic Revenue. Adio is the CRM for this world. It meets you where you work, compounds every customer signal into context, then axon it across your pipeline to let you move it on match speed and scale. With agents and automations for every job in revenue, Adio orchestrates your work around the clock. Built to handle the scale of your workloads, extensible with API and MCP, and with the infrastructure to keep up with your most ambitious agents. Loved by high-growth startups like Granola, Modal, and Etched, Adio runs the work behind every win. That's Adio, the agentic CRM, the intelligent system that never sleeps.

19:23Picks up leads at 2 a.m., catches renewals before they slip, hands you the answer before you ask. Try Adio free at adio.com slash iHeart. That's adio.com slash iHeart.

19:40It's time now for Talking Tech. I'm Yahaira Anand. First up, shares of Shein on the gray market are under pressure ahead of the fast fashion giant's Hong Kong debut Tuesday. The stock fell as much as 17 % in gray market trading after its$1.7 billion IPO. That values Shein at about$26 billion, well below its nearly$100 billion valuation back in 2022. Plus, SK Hynix is eyeing Japan as it races to meet booming AI demand. The memory chip giant is studying potential manufacturing sites across the country with its chairman, telling Bloomberg it wants locations with good power and good water. According to a source, a small acquisition or partnership with a Japanese company is also on the table.

20:29And Z.ai fell short of expectations as China's AI battle heats up. First up, first half revenue grew nearly fivefold to$142 million, but still missed analyst estimates by about 30%. That's despite the company's market value surging more than 800 % since its January listing. Ed?

20:52Ed Ludlow:Okay, thank you, Yuhaira. China called for greater guidelines for AI models and accused Anthropic of bad behavior ahead of AI talks with the U.S. China said on Sunday the U.S. to prove its AI companies are subject to the same guidelines before any, quote, substantive talks with China could take place. Mike Shepard joins us with the latest. This is China trying to set the agenda, right, Shep, ahead of what are pretty key conversations. Very much so. And this is a conversation, of course, that would set the stage for a dialogue on artificial intelligence and AI policy between President Donald Trump and Chinese leader Xi Jinping September 24th here in Washington.

Read the full transcript

21:33And Ed, before the two leaders sit down, though, they need to sort out exactly what will be on the table before them. And there needs to be this dialogue between lower level officials from the two countries. So far, we have seen no date or agenda set for such a conversation to take place. And we see here in this post from a social media account that's affiliated with the Chinese state broadcaster CCTV, a version of what it would like. Before any conversation can happen on AI, the U.S. really does, according to China, need to meet these terms that it is setting. American AI developers need to be following a common set of rules when it comes to disclosure, when it comes to audits, and when it comes to safety.

22:18And interestingly yet, it also singled out Anthropic as perhaps its main villain in the whole opera here.

22:23Ed Ludlow:So that is interesting. I mean, they are singling out a single U.S. company or entity. Do we know why? Well, you know, the post itself was titled Anthropic has contracted the American disease and then used Anthropic really as a way of explaining all of its issues. Anthropic, though, why them and not OpenAI, which, of course, has had its own safety issues that we have talked about extensively on this show with the hack on hugging face. Anthropic, though, has really taken a position that is very much in the view of officials from Beijing as anti-China. Dario Amadei has supported the export controls that past administrations have pushed to prevent Beijing and Chinese companies from acquiring advanced AI chips.

23:08And he has also warned against the prospect of China actually taking the lead in artificial intelligence development and software. And the reason is that they see it as possibly being used as a tool by Beijing. And of course, we're seeing China now react against Anthropic.

23:26Ed Ludlow:Bloomberg's Mike Shepard with the reporting. Thank you very much. Tomorrow, Apple gets a new CEO. But how will that affect the iPhone maker's priorities? Bloomberg's Mark Gurman, who leads our coverage of Apple and consumer tech, has been looking into that. And what John Turnus brings to the table, Mark, is with us now, this was the subject of this weekend's Power On, but a really detailed piece in Bloomberg Businessweek. And let's start there. OK, so today is Tim Cook's last day. Tomorrow is day one of John Ternus as CEO. What changes overnight? I would say pretty much nothing changes overnight of any major significance.

24:03This is going to be a gradual multi-year changing of the guard. Tim Cook is not going anywhere. He's still going to be around. There are questions about how much he's going to be around. I'm led to believe he will be very present and intimately involved with Apple in the several years to come, serving as an advisor to John Ternus, the new CEO, but also having a major impact on things like supply chain, operations, procurement, but most notably working with the Trump administration and the government in China and other governments around the world where he owns those hands-on relationships.

24:42Ed Ludlow:Timing is so interesting. Next week, next Wednesday, is the Apple full iPhone event. How much has John Ternus had a hand in this generation of iPhone that's likely to be announced when we get them next week? Well, my answer is the same for the last five generations. John Ternus has been leading the development of the iPhone since the release of the iPhone 12 in 2020. He spearheaded that. That was his final test before being named to the CEO role earlier this year or being named to the senior VP of hardware engineering role in 2021. That iPhone was a big success at the height of COVID with 5G and the new design.

25:24And then they rolled out subsequent iPhone after iPhone after iPhone under Ternus. Right. And the timing of the transition is not a coincidence. Typically, they like to do these big transitions either in October, December, or at the very beginning of the year. September 1 was picked for a very specific reason, to have John Ternus be the one to introduce the foldable iPhone, the smart home display, the new MacBook Pro with a touchscreen. Basically, be able to present him as sort of this new product development maestro as Apple rolls out its biggest new product roadmap in its history.

25:59Ed Ludlow:So the thing that will change really overnight in the internal directory at Apple is some job titles. Very quickly, just recap what we've reported on how John Ternus is expected to be different in his leadership, his style. I think there's going to be a lot of continuity there, right? I mean, he has taken every opportunity to praise Cook's leadership style. The big difference is that over the past 15 years, decision-making at Apple has been very collaborative, where Tim Cook has let the executive team make the decisions and maybe he would make the final decision, but it was a consensus-driven leadership process.

26:35At Apple, I expect far more decisiveness and singular decision-making.

26:41Ed Ludlow:Bloomberg's Mark Gurman with two must-reads from the weekend, the Power On newsletter and this piece in Business Week about the transition. Thank you very much. Generally speaking, technology stocks are lower. There's been a flare-up in the Middle East, which is having the sort of risk-off sentiment driven into the market. I would say that chip stocks are outperforming and modestly higher in large part because NVIDIA has held on to the gains that it had after that announcement of investing in MediaTek. NVIDIA is making another multi-billion dollar bet on the AI ecosystem, investing$3.5 billion in what are convertible bonds in Taiwan's MediaTek.

27:16Ed Ludlow:I asked Jensen Wang about the proportion of the content of a server design that NVIDIA is now responsible for, pushing on its latest generation Verirubin platform and how custom silicon would impact that. Listen to this. Inside a gigawatt, if we were just to extrapolate, in a gigawatt, Hopper was about$18 billion of economics for NVIDIA per gigawatt. Grace Blackwell is about 25. And the Verirubin generation, It's about$40-plus billion per gigawatt of total NVIDIA compute economics. And the reason for that is because we've been expanding beyond the GPU to one type of switch after another type of networking.

28:03And now the amount of chips, there's like seven different types of NVIDIA chips that has to go into an AI factory. An XPU would be a replacement of one of those. You know, my sense is that we'll continue to be in every single cloud, we'll continue to be in every single AI factory. Some of those racks might be augmented with XPUs. And so in those cases, NVIDIA's networking, NVIDIA switching, NVIDIA CPUs would also have an opportunity to participate. And we make it easier for the customers to be able to extend out their AI factory using exactly the same system architecture and networking architecture.

28:45In the case of MediaTek, it allows MediaTek to offer an end-to-end solution that is complete. that XPU otherwise would have to go and solve NVLink switching problems, a whole bunch of NICS problems, a whole bunch of Ethernet switching, scale-outs, scale-up, scale-across, scale-in. There's so many different types of networking technologies that have to come to bear, and we offer the entire thing. So between us and MediaTek, if a customer wants to build XPUs, they now have an entire solution as well. And so our opportunity will continue to grow in these AI factories.

29:25Ed Ludlow:That was NVIDIA CEO Genson Wong earlier this morning. Anthropics IPO plans are casting a bit of a shadow over other companies' U.S. listing plans. The clawed developers preparing to file publicly for an offering that is expected to be as large or larger than SpaceX's record$86.2 billion debut. Bloomberg Equities reporter, May Lee Lipscholz is here. ECM, IPO czar. It's getting kind of real. And I find this so interesting that in our reporting, it's not just us waiting for Anthropic to flip the S1 public. It's everyone else waiting for them to do it as well, to decide what it is they want to do with their own public market futures.

30:05Ed Ludlow:Well, Ed, the thing you have to keep in mind, too, is we have a very much live Fed meeting in the middle of September. And then we have an election that gets remarkably close. The longer Anthropic drags its heels. So when I'm talking to bankers, private equity sponsors, and pretty much everyone in the investing world, they're kind of wondering, okay, when will this flip public? Once we have an idea on when they're going public, we either have to A, hurry up and get out the door, prepare to get out the door. We're looking at potentially seeing filings from the likes of an N-Scale, an Aura, and an SB Energy.

30:35Or you got to sit on your hands, let Anthropic do its thing, and hope the market is still there in the aftermath.

30:41Ed Ludlow:And that was a debate we saw with SpaceX. So there's been, even just in the last five days, like lots of anthropic headlines. And it's dominated the news cycle even before it's happened, the S1 flipping public. And in your report, you detail that like a lot of those other IPOs or potential IPOs that you're mentioning are finding it hard to get the attention of some investors. Well, and the thing you have to keep in mind, too, is the investors that a lot of these companies want one-on-ones with, sovereign wealth funds and long-only investors, they're not big teams. So if you were trying to court a Wellington or a Norges, you kind of have to deal with, okay, well, what happens if Dario comes calling?

31:19Because they're going to cancel any meeting that they have to meet with the team from Anthropic, even if they are invested in them privately. So when we talk to some of these sponsors, when we talk to some of these bankers and these VCs, the question comes back to we don't want to be on the road when Anthropic is on the road. We don't want Anthropic to have an investor day that all of a sudden now, Wednesday, pick a date in September is now off the table because everyone is going to be tuned into that call. And then immediately after that, having internal discussions and talking to everyone they know on the street.

31:48So it is potentially going to impact what this race to kind of the end of October looks like from an IPO perspective.

31:55Ed Ludlow:So, Bailey, let's quickly end by recapping the reporting. Where do we stand with what we think will happen with Anthropic's IPO? We're expecting Anthropic to finalize its revolving credit facility. That's the first big step, the next big step. After that, banks come in under the tent. They're able to kind of get their hands on the numbers. They'll have some form of an analyst day. Then the public flip. So public flip may be the end of this week. On the other side of Labor Day, that kickstarts the two-week waiting period. Then you have the launch deal pricing as early as the end of September. But if you look at the calendar right now, things really have to get in motion for it to not slip into October.

32:30work.

32:31Ed Ludlow:Vimbo's Bailey Lipschelz, thank you very much. We have some interesting data here. Madrona has released its sixth annual IA40 list of the top private companies building and enabling AI. Madrona is an investor in the likes of OpenAI and Anthropic, but it sources nomination and votes for this list from 72 outside investors across 54 firms and evaluates the list against pitchbook data. Matt McElwain, Managing Director, Madrona is with us. I find this different. It's a different approach to trying to find out real world what's happening right now. Just explain the methodology and the list itself. Yeah, so this is the sixth year of the list, and it's the fifth year of the summit that we host every year.

33:13And what we have done is we have surveyed the entire market of venture capitalists and technology leaders at companies like Microsoft and Amazon and others to give us their best nominations. And then we have a voting mechanism, and then we combine it with pitch book data, and we come up with this list. Here's what's amazing about the list. The first year we launched it, in 2021, pre-Chat GPT, the entire group of winners raised$15 billion. This year, the list will have raised over$300 billion. And in aggregate, over the last five years, those companies have raised over$500 billion. So as you're talking about companies like Anthropic and its IPO, and Anthropic is on the list, as is OpenAI, those are companies that have already raised approaching$200 billion in the private market before they ever went public.

34:05So these are really the leading indicators of companies, much more in the applied space and in the agentic space relative to the infrastructure space that are going to be the next generation of

34:16Ed Ludlow:leaders. There are several conclusions you can draw, and it's several different ways to slice the data, right? So, you know, I think the 45 winners,$410 billion between them raised since founding. Correct. But of that$410 billion, just three names account for most of it. The three names, no surprise, OpenAI, Anthropic, and Databricks. And Databricks. That's$375 billion, roughly, of all of that capital. But what's interesting is that still leaves$800 million per company for all those other companies that have ever been on the list. So a tremendous amount of money. And remember that we restrict the early and mid-stage to companies that have raised less than$200 million.

34:57And so over time, many companies have migrated from being in the early and mid-stage of size of capital that they've raised up into the larger amounts. What's also interesting is the turnover on this list. Roughly 50 % a year, some years even higher, of companies that were on the list the year before fall off the list in the following year. which goes to this question of durability versus disruption. We all see tons of disruption going on, but what companies can be durable? And how can they be durable?

35:27Ed Ludlow:When I look at the list, I want to dwell just a little bit on those three, but principally OpenAI and Anthropic. It also is a reminder of where the capital is coming from. And in the case of the Frontier Labs, it's what people call strategics. But the hyperscalers and NVIDIA themselves, You heard some of our interview with Jensen Wang and NVIDIA making a pretty sizable investment in media tech. What do we infer from that? That at this scale, that's where the money is coming from as opposed to venture capital. Well, it is more from strategics now. And what I say to both the big tech companies and to the people funding the model companies, thank you very much.

36:06We really appreciate, you know, the five largest companies, you know, in technology, you know, in the world, the Amazons, Microsoft, Googles will combine, invest an estimated$750 billion in CapEx this year. On top of that, you've got hundreds of billions of dollars going into open AI and Anthropic. But what is the ROI of AI? That is the core question that's increasingly being asked by enterprises, by the businesses that leverage all these tokens. You've heard about token maxing. You've heard of all these different dynamics of am I getting an ROI as an end user customer as a applied business? Is there an answer to that question yet?

36:45Oh, I think that you look through this list and you're going to see companies that are not only generating good gross margins. A lot of the companies like the Servals and Gradials and Vercels that are on this list. but they are delivering a ton of value to their end-user customers, whether those customers are in a specific functional area like marketing or IT or sales, or they're more cross-cutting in serving builders, as in the case of Aversell and others.

37:12Ed Ludlow:The other domain that comes to the forefront here is coding. You talked a little bit about some of the players in that field, but it's interesting to see so many of those involved in coding make the list. that is on a daily basis on this program where the news flow is at the moment. Well, I think coding is very important. And Cursor, who was on the list for many years, they've now been acquired, of course, by SpaceX, led the way, but you really should think of it more broadly as builders. How are we enabling builders in so many different areas? A company like Lovable, that is in the Vibe coding area of building, or again, a company like Gradial that's helping marketers and the enterprises, the global 500, be able to build amazing solutions in applied marketing areas.

37:58So I think it's about builders. I think it's about collaboration, not just between humans, but humans and agents. And then finally, I think it's about the continuous learning loop. You're not hearing as much about this yet. We are hearing about it as early stage venture capitalists, that continuous learning from taking the engagement I have with builders, running that data through models and systems, and coming out with continuously improving systems like Cursor did, which is why Cursor has been so successful.

38:27Ed Ludlow:Matt, you heard Bailey's reporting there very quickly. What does an Anthropic IPO do to this list and this ecosystem? Well, I think it actually clears out the way for future IPOs. I think there's a number of companies on the list that are going to wait for that Anthropic IPO, not the least of which, of course, OpenAI. Right. But, you know, you look down the list and you see some really great companies like a ClickHouse, like a Databricks and others. And I think if the Antropic IPO is successful, we can all debate what the definition of a successful IPO is. But let's say it's successful and it's viewed as successful in the capital markets.

39:03I think you'll see that be an opening of the gates in 2027.

39:06Ed Ludlow:Matt McElwain, Managing Director for Madrona. I've really enjoyed having you here. Thank you very much. Coming up, the U.S. Army tapped five nuclear reactor suppliers in an effort to boost America's nuclear industry. Radiant President and Chief Operating Officer, Tori Shavander, joins us next. This is Bloomberg Tech.

39:32The people who seem to get more done than everyone else. They're not working longer hours or running on more caffeine. They've just stopped wasting time on the stuff that doesn't move work forward. Switching apps, re-explaining context, hunting for files. Those aren't small inefficiencies. They're hours wasted every week. Superhuman Go gives you those hours back. From the makers of Grammarly, Go is an AI assistant that sits inside every tab and tool you already use, always available and already aware of what you're working on. Ask it to draft something, summarize a long thread, pull up a file, or prep you for a meeting.

40:12Go handles it without you ever leaving the page you're on. This is what it looks like when AI actually fits into your work, instead of adding to it. It's like having a teammate whose only job is to help you be better at yours. Go keeps up so you can move forward. With Go working with you, you can show off what you do best. Superhuman Go. Find out more at superhuman.com. That's superhuman.com. Some people treat ChatGPT like some kind of smart search engine, and some use it to get work done. ChatGPT Work is a new way of working in ChatGPT that can take action across your apps and files, stay with a project for hours if needed, and turn a goal into finished work.

40:55It's designed to help you move from a chaotic starting point to a reviewable first version. So all the source materials, briefs, and scattered information that you have to grind through to turn into something useful can just become something useful. Put ChatGPT to work on your most ambitious ideas and projects. Get started at ChatGPT.com by selecting Work Mode, available on Plus and Pro plans. You already know how AI is changing how everyday work gets done, how much ground you can cover, and how fast a team can scale. To stay ahead, you need the tools that give you a competitive advantage, built for this new era.

41:33Welcome to Agentic Revenue. Adio is the CRM for this world. It meets you where you work, compounds every customer signal into context, then axon it across your pipeline to let you move it on match speed and scale. With agents and automations for every job in revenue, Adio orchestrates your work around the clock. Built to handle the scale of your workloads, extensible with API and MCP, and with the infrastructure to keep up with your most ambitious agents. Loved by high-growth startups like Granola, Modal, and Etched, Adio runs the work behind every win. That's Adio, the agentic CRM, the intelligent system that never sleeps.

42:08Picks up leads at 2 a.m., catches renewals before they slip, hands you the answer before you ask. Try Adio free at adio.com slash iHeart. That's adio.com slash iHeart.

42:25Ed Ludlow:Okay, some breaking news on the Bloomberg Terminal. Tim Cook says he has enormous comfort with John Ternus taking the helm as CEO and said that he will miss this work as he has his final day as CEO of Apple. This was in an email that he sent to all Apple employees this Monday. Again, his final day as CEO. He writes, As you know, I am not leaving Apple, but I'm stepping away from a role that I have loved deeply. He continues in the email to say, I will miss this work in ways I can only begin to imagine, even as I remain completely at peace with my decision. Cook said that a few people understand what it takes to build products that change the world the way that John Ternus does.

43:08Ed Ludlow:We earlier in the show talking to Bloomberg's Mark Gurman, who detailed the transition in Bloomberg Businessweek over the weekend. Tomorrow, John Ternus becomes CEO of Apple and Tim Cook moves to the executive chair role. If we hear any more, we'll bring it to you. Okay, the U.S. military recently tapped five suppliers investing$2.2 billion to boost the nation's nuclear industry. Radiant was among the five firms tapped, earning a$750 million contract to provide 15 nuclear micro-reactors. Radiant president and chief operating officer, Tori Shibandan, joins us now. Really interesting. Last week, this was a big story, a big story about America taking action for the technology in the field, but also to see the work of the companies doing it.

43:58Ed Ludlow:Could we just start by explaining, Tori, what is Radiant? What is its point of difference in the field of those working on the technology? Yeah, absolutely. Thank you very much for having me. So Radiant builds one megawatt micro reactors that are transportable by land, air, and sea. One megawatt is about enough power for a thousand homes. Our big differentiator is that we are portable. And last week we were awarded$750 million for 15 of our portable reactors for the Army Janus program. That's about one third of the overall Janus program budget, but also about 75 % of the overall Janus program volume.

44:36So 15 of 20 reactors going portable. really showing that a vote of confidence in the maturity of our tech, as well as our ability to scale and manufacture. I'd like to thank Dr. Jeff Waxman and DIU director Owen West for their bold leadership and understanding that portable nuclear is the big solution that the Army needs.

44:56Ed Ludlow:Tori, what did you have to demonstrate in this process to win the award? Yeah, so it's a pretty grueling process through multiple down selects. We've competed against over 40 companies in this. And the U.S. military is an incredibly savvy customer. They have many experts where we've been through hours of presentations as well as grilling on our understanding of what our product is capable to do, as well as them coming on site to see the technology we've developed. We shipped recently our test reactor up to the Idaho National Lab Dome, where it will begin our overall testing over the next few months because the next reactor that we will deliver to a military base in 2028, a customer really needs us to understand our product and the military's selection of us in this award is a vote of confidence.

45:47Ed Ludlow:Tori, when your industry comes on the program, almost universally they talk about China and how China is either ahead on the regulation side or on the technology side. What is Radiant's house view on that, please? Yeah, so when I look at what Radiant's solving, we're solving the entire problem of the American energy grid. And when you compare it to China, we're solving different problems because our baseline solution is different. The American energy grid was built in the 60s. It hasn't really changed. We have generation, transmission, and utilization. And what Radiant is solving is being able to solve generation and transmission on the customer site.

46:25You can either just build large CapEx projects fast without what China is approaching, or you can be innovative. This is what America and the American military are best at and come up with a new solution to make a more resilient grid to build the future America needs.

46:41Ed Ludlow:Tori, very quickly, what do you need to do next to scale? Will you raise money? Will you hire competitively? No, we will do all of the above, but most importantly, we'll put our heads down and build for America. What folks need to understand is that our grid needs a new solution and portable nuclear is that new solution. Radiant President and Chief Operating Officer, Tori Shivanandan, great to have you on the show. Thank you very much. Let's get back to our top story. NVIDIA investing$3.5 billion in Taiwan's Mediatek. Bloomberg Intelligence saying this gives NVIDIA, quote, another path to benefit as custom silicon units growth accelerates, While MediaTek gains a stronger platform to challenge Broadcom and Marvell for hyperscaler A6 programs.

47:25Ed Ludlow:Here with more is Kunjan Sabani of Bloomberg Intelligence who wrote that note. So interesting. Custom silicon in conjunction with NVIDIA's GPU-based architecture. Just take us deeper into the React, Kunjan. Yeah, so most of the times when we discuss what's the biggest challenge to NVIDIA GPUs, it's these custom silicon, right? And this is a great hedge that NVIDIA is playing. So even in a data center that is built up, not using NVIDIA's GPUs, but using custom silicon, like whether it's a TPU, whether it's an Amazon or some silicon, these kind of partnerships will allow NVIDIA to stay within the stack, both from a hardware and a software perspective, and hence not lose out on the entire CapEx dollar spend.

48:06We have already seen a glimpse of this play out. NVIDIA had a similar arrangement with Marvell, which is the key ASIC provider for AWS. And we are seeing this NvLink Fusion deployment happening in AWS. And guess what? Recently, AWS has been the biggest NVIDIA GPU customers. So read what you will in there. But this is definitely a great win-win situation.

48:27Ed Ludlow:So the stock's higher for NVIDIA. I don't know that I'm going to draw a direct causal link. The most interesting point in the interview was probably what Rick Zai said, which is, you know, XPUs are difficult, but this actually accelerates speed to deployment by doing it this way with NVIDIA MVLink. Do you buy that argument? Somewhat true. Look, the challenge that XPUs have, especially when deploying it to external parties, it's a very closed ecosystem, both from a software perspective and a network perspective. So this does help them with time to market. But I would not stretch it to a degree and say this really helps them become sort of a chip seller to the whole industry.

49:07Ed Ludlow:Let's do the kind of finance bit. NVIDIA buys$3.5 billion of convertible bonds of Mediatek. It's their biggest non-US direct investment. What does BI think about that mechanism? I think, you know, like I mentioned in my prior interview, this is sort of the bank of Jensen or the bank of NVIDIA really flexing its muscle. We see these as great use of cash. I mean, look, the amount of cash they are generating, what can they go ahead and do, right? Last time when they announced an increase in dividend, the street didn't really like it. It took it as a signal that the growth is slowing. Even if they do hundreds of billions of buyback, that's basically single digit percentage buyback for the stock.

49:48They are already investing heavily in R &D and OPEX. So we think this use of cash is definitely going to pay a lot more ROI versus the other use of cash.

49:59Ed Ludlow:congenital Bloomberg Intelligence with a react that if you are a terminal subscriber or client, go and read the research, super important, and watch the interview back. It was very interesting. That does it for this edition of Bloomberg Tech. This is what markets look like, some anxiety in the Middle East with Iran, probably the biggest factor, some outperformance in chips because of Nvidia. Recap that interview, or at least some of it on the podcast, you know where to find it. It was a 20-minute conversation, so you can also check out the full version wherever you get your interviews. This is Bloomberg.

50:55Thank you. Learn more at business.optum.com. You already know how AI is changing how everyday work gets done, how much ground you can cover, and how fast a team can scale. To stay ahead, you need the tools that give you a competitive advantage built for this new era. Welcome to Agentic Revenue. Adio is the CRM for this world. It meets you where you work, compounds every customer signal into context, then acts on it across your pipeline to let you move in on match speed and scale. With agents and automations for every job in revenue, Adio orchestrates your work around the clock. Built to handle the scale of your workloads, extensible with API and MCP, and with the infrastructure to keep up with your most ambitious agents.

51:38Loved by high-growth startups like Granola, Modal, and Etched, Adio runs the work behind every win. That's Adio, the agentic CRM, the intelligent system that never sleeps. Picks up leads at 2 a.m., catches renewals before they slip, hands you the answer before you ask. Try Adio free at adio.com slash iHeart. That's adio.com slash iHeart.

52:26Visit TheFoodsTeens.com to learn more and where to buy. These statements have not been evaluated by the Food and Drug Administration. This product is not intended to diagnose, treat, cure, or prevent any disease.

From the publisher

Bloomberg’s Ed Ludlow breaks down Nvidia's $3.5 billion investment in MediaTek, deepening its collaboration with the Taiwanese chipmaker. Plus, today is Tim Cook's last day as CEO of Apple before John Ternus takes the reins tomorrow, and all eyes are on Anthropic as it prepares to file publicly in the coming weeks.

See omnystudio.com/listener for privacy information.

More from Bloomberg Tech

All 342 episodes
Nvidia Makes $3.5 Billion Bet on MediaTekBloomberg Tech · 43 min
Listen in VO