OpenAI, Oracle Expand Stargate Project

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. This is Alexis Christophorus for Bloomberg Surveillance. Most companies have tried AI. Most aren't seeing results. Not because AI doesn't work. It's because AI hasn't reached the workflows yet. That's the gap Asana is built to close.

1:24Asana is the operating system for human agent teams. Ready-to-go AI teammates pre-built for marketing, ops, and IT. No prompt engineering. No setup. They show up where the work is happening, already onboarded in your workflows, ready to deliver. Asana, where humans and agents work together. Try it at asana.com. That's A-S-A-N-A dot com. Bloomberg Audio Studios. Podcasts, radio, news.

1:55Bloomberg Tech is live from coast to coast with Caroline Hyde in New York and Ed Lovelow in San Francisco.

2:06This is Bloomberg Tech coming up. OpenAI ramps up with more computing capacity from Oracle, expanding the Stargate initiative. Plus, the US House of Representatives is on track to vote on Donald Trump's tax and spending bill. What does that mean for tech? And the Trump administration lifts some export license controls, requirements for chip design software sales, specifically for China. Let's get right to financial markets. The Nasdaq 100 is pushing once again to fresh record highs. Strong jobs data for the month of June. Halfway through the show, we're going to go really deep on jobs, and then we're going to look at the labor market when it comes to the technology sector.

2:44But there are other things happening that are pushing the tech sector and equity markets higher. A lot of it is related to compute capacity. Bloomberg broke a really big story last night, and I want to get right to it. The story is Oracle. So on a two-day basis, we saw a 5 % jump in Oracle yesterday. We're pushing a little bit higher again this Thursday. But the deal is OpenAI expanding the Stargate project, getting more compute capacity in new states and new sites beyond what they've already got in Texas. The person that broke that story, Bloomberg's Brody Ford, and he's with us now. I think we need to get to the details of what you reported because there was a lot of detail in it, Brody, specifically in terms of gigawatts, the scale of this new capacity.

3:25Yeah, we saw earlier in the year, you know, Larry Ellison standing there in the White House talking about we're going to deploy$500 billion with OpenAI for data centers. A lot of people doubted that scale. Well, yesterday we broke that they have inked a pretty unprecedented deal, 4.5 gigawatts. The usual rule of thumb is that a gigawatt is about a nuclear reactor. And so we're talking about a scale of four and a half worth of nuclear reactors or powering about three and a half million homes. I mean, this is likely the largest cloud deal of all time, and it's going to be tens of billions of dollars of chips, of power.

4:05I mean, it's going to be hard to wrap our head around what this looks like in the build out. Okay, what do we know? What are sources telling us about where these new sites are going to be? What's under consideration? And kind of like how near term this is? Right. So they've narrowed it on a handful of data center sites, right? I mean, to 10 states like Wisconsin or Pennsylvania or Texas. They'll also be expanding the campus they've already built out in Abilene that a lot of us have seen, you know, images of. How near term? I mean, a lot of this stuff is tough to get off the ground. Oracle earlier this week said that the full ramp would likely not hit until 2028.

4:45And so it's fairly far out, but I would expect to see some of these sites come on in a pretty rapid clip. We don't know exactly which one's going to move first. But, you know, I'm sure we're going to be starting to get more details trickling out in the coming weeks and months. Let's talk about the Oracle side of this. This is a stock that's now, again, pushing fresh record highs almost every day at the moment. It's up 40 percent so far in 2025. This is a big win for Oracle. They are really pushing and changing their position in the landscape of infrastructure and AI cloud computing. Three years ago, if you said Oracle, major cloud player, people would make fun of you, definitely.

5:27But today, that's not the case, right? I mean, they're making, I think, around$10 billion per year in cloud infrastructure. On Monday, they said the deal that we connected is this OpenAI one will be$30 billion per year. I mean, that is triple the size of their entire current cloud business. And so it's huge for revenue. The big question is what it means for margins. And that's kind of the classic with a lot of these AI businesses that they're going to be buying a lot of chips and power and construction and that's going to constrain their margins. What does that mean for cash flow? That's what investors are asking this morning.

6:03Bloomberg's Brody Ford there with the reporting. Let's get to the analysis. This expanded deal between Oracle and OpenAI, it's also good news for some chip makers. That's according to Bloomberg Intelligence who say the likes of NVIDIA and AMD are going to benefit here. Kunjan Sabani offered that note and he joins us now. Inside all of this infrastructure, at the heart of it is the compute. And the compute right now in this market is largely coming from NVIDIA, but also to a greater extent increasingly from AMD. What's the thesis here that you've got, Kunjan? Well, two points. As Brady mentioned, Oracle has been really beefing up their spending versus what we thought a year or two ago.

6:42If you look at the estimates now, even for 25, that is going to be the fastest CSP in terms of CapEx increases. So definitely increasing the wallet share and becoming a serious buyer of these chips. Second point, it is one of the only hyperscalers or cloud providers among the top five who does not have a custom silicon or an ASIC program. What does that mean? is majority of the capex that it is going to spend to bring up this 4.5 gigawatts of capacity is going to merchant GPU providers like NVIDIA, which it has been buying most of GPUs from in the past, but also AMD, because as they announced, they're ordering about 130k M355 for the second half, which equates to somewhere between three to four billion dollars of revenue for AMD.

7:29Just real quick, Kunjan. Brody was giving us the reporting on the dollar value of this deal to Oracle, right? It's that$30 billion that we were talking about earlier in the week per year starting in fiscal 28. Are you able to model or do the math on what these kinds of deals and the value of them are to AMD and NVIDIA in terms of like number of chips that they will send that way or kind of the pipeline of business that they're securing through Oracle? To some extent, because there are a lot of assumptions, but what we can map is the gigawatts to the chip spending. So rather than the revenue that Oracle will collect, the 4.5 gigawatts, assuming that's all brand new data center build out, one gigawatt approximately in today's term equates to$60 billion of CapEx, which includes everything from chips and hardware.

8:21So even if you had to take conservative stance and cut it half for chips, that still leaves you for every gigabond at least$30 billion. Bloomberg Intelligence's Kunjan Zabani there with the analysis. We had the reporting. We got the analysis. Now let's get the market reaction. Janet Mui, head of market analysis at RBC Brewing Dolphin, found it really interesting. These big capital projects in infrastructure just keep coming. What does that signal to you about what the market should model for in the coming years ahead? Hi, Ash. Thanks for having me. I think the significance of this deal is that the demand for computing power is just insatiable.

9:05And remember when DeepSeek came out, a lot of investors were worried that actually there may not be the need for so much computing power if the modeling is so efficient, but actually that's not true. We see that there is high visibility of the scale of the investment by sovereigns, by hyperscalers, and many corporates. So I guess this is really the big conclusion is that the runway is still very long and the incentive to invest is still very high. So I think the theme of investment in AI, particularly in semiconductors, And cloud, which are the AI infrastructure, is particularly relevant. Janet, in Brody's reporting, he's saying that this expansion between Oracle and OpenAI is 4.5 gigawatts worth of capacity.

9:57And in the context of energy, you could power this country at that scale. How do you think about the energy sector and the energy infrastructure requirements that markets are going to have to fund to support these types of data center expansion? And again, I think the utility side, energy side of things goes hand in hand with the need for computing power. And that's why we have seen hyperscaler like Meta investing in nuclear energy, for example. So I think the quest for cheaper, more efficient energy would be there. And I think that is why I think the U.S. will remain basically the region that would lead the AI race because the U.S.

10:41is basically self-sufficient in energy production, relatively cheaper energy compared to, say, Europe and, for example, the U.K. So I guess that's a very important conclusion is that there is incentive for companies to invest and base their production in the U.S. to access its cheap energy base. Janet, the story this Thursday is treasuries are falling, the dollar is stronger, and equity markets, particularly the technology sector, pushing fresh record highs on the Nasdaq 100. Give me your reaction to the jobs data and the eco print that we got this morning. Well, I have to say the data is basically Goldilocks because on one hand you get better than expected jobs data.

11:25In general, a softening trend, but it's still OK. It's still reflective of a solid labor market. Unemployment rate is lower. At the same time, wage growth is actually slower despite a shrinking labor force. So I would say it's a Goldilocks, which the market clearly likes. And I think the conclusion is that the Federal Reserve really is in no urgency to cut rates. But that's fine, because I think the markets would prefer the evidence of a strong labor market rather than fretting over the exact timing of the rate cut. I do see some outperformance in technology. Why is that? Why is that? Is it just a waiting issue right now?

12:02And technology is what this market is? Or is there something in this economy and the path forward for the Fed that makes this sector more attractive right now? I guess there are lots of reasons, right? First of all, in terms of earnings growth profile, technology is still the brightest spot out there. As I mentioned, clear visibility on AI investment. I think, secondly, I think in terms of the productivity gains, I think the tech sector is likely to see or have a lot of those productivity surplus. And, of course, I think in terms of the U.S. market, I think the tech sector is where the attraction is to access the AI theme, et cetera.

12:49And if we see, for example, rate cuts really coming by, I think lower bond yields, lower interest rate also particularly benefit those growth areas as well. So really a combination of factors. If you're a technology investor, that was the what you needed to know. Janet Mui, head of market analysis at RBC Brewing Dolphin. Thank you very much. There is a lot more coming up. The House pulls an all-nighter as Republicans race to get a tax bill to President Trump's desk before their July 4th deadline. We're going to go live to Washington, D.C. next, and we will have the latest on where things stand.

13:23Don't go anywhere. This is Bloomberg Tech. Live the good life, good paying job, good houses, good houses. The 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.

14:05Ask it to draft something, summarize a long thread, pull up a file, or prep you for a meeting. Go 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.

15:01their own voices or rely on ones provided by Suno. While most of the songs generated on the service go unheard of by the masses, a few have taken off, including a recent TikTok trend that involves users uploading text messages to Suno and turning them into musical performances, like taking the text from a crazy night out and making it a gospel song. Suno's pitch to investors is that it'll democratize music creation, leading to a more than$5 billion valuation for the company. The music industry, it's watching, as its biggest concern about AI is whether record labels, rights holders, and artists will be compensated for the work they claim built the training models for companies like Suno.

15:41Another question? Will these AI bangers have any staying power or just be one-hit wonders? That's the Bloomberg Tech Minute brought to you by ChatGPT. Put ChatGPT to work on your most ambitious ideas and projects. Get started at ChachiPT.com today by selecting Work Mode, available on Plus and Pro plans. At Edward Jones, we believe rich is more than caring about the latest and greatest. It's also taking care of what gives your life meaning. That's why your dedicated financial advisor meets you where you are, with personalized financial strategies that help protect what matters, so you can preserve your progress while creating a path forward.

16:23The key to being rich is knowing what counts. Let's find your rich together. Edward Jones, member SIPC.

16:38Welcome back to Bloomberg Tech. You're looking at live pictures from Capitol Hill, where House Minority Leader Hakeem Jeffries is delivering a blistering rebuttal of the tax bill. That's now going into its sixth hour. here with the very latest Bloomberg's Tyler Candle out in Washington, D.C. What is the latest? What do we expect to happen? What do we need to know? Yeah. Hey, Ed. So the House Democratic leader, Hocking Jeffries, has something known as the magic minute, which means that he can speak for however long as he wants. And as you mentioned, he started at 4.53 a.m. Eastern. So we could be in this for the long haul, but it really does feel like they're sort of delaying the inevitable because Republicans appear poised to pass President Trump's signature legislative achievement, which would be the one big, beautiful bill.

17:23Once Leader Jeffries is done speaking, House Speaker Mike Johnson might take the mic, or they could try to get this done as soon as possible, and this would proceed immediately to a vote on finalized passage. Once that happens, we're looking for that magic number of 217, which means House Speaker Mike Johnson can only afford to lose three votes, but he appears to have it. It seems like Republican leadership was able to flip some of those staunchest critics of this bill, those members of the House Freedom Caucus, those Republicans that were pitching themselves as fiscal hawks. And our reporting on the Hill this morning indicates that they did get some sort of concessions and assurances from this White House in order to flip their votes, including one that I know that you pay close attention to, which is when it comes to the phase-out of those clean energy tax credits.

18:07Ralph Norman told our reporter on the Hill today that one of the assurances that the White House gave them was that there would be strict enforcement when it does come to the phase out, particularly around solar and wind tax subsidies. So that's something to look forward and look out to as we try to glean more details on exactly what it took to get these Republicans to flip their vote when it came to the procedural vote earlier this morning. Tyler, technology markets in real time, Nasdaq 100 already pushing fresh record highs, now extending its gain in the session to one percent. It is a short week because it is a July 4th holiday this Friday in the United States.

18:43There's a lot being made of, like, the symbolism of the president signing this bill July 4th. I heard you speaking with Bloomberg's Matt Miller about that earlier. Why is that symbolism there? Well, it really was this self-imposed deadline from the administration that they wanted to get this done by July 4th. We are expecting a pretty big signing ceremony either later today or tomorrow so that President Trump can tout this legislative achievement that they're really hoping will pair with some of these potential trade deals, frameworks that we could get next week ahead of that July 9th deadline that they have put into effect.

19:18Because the administration is really hoping for a few different things to happen here. They want to see a confluence, really, of economic factors going into the summer so that they can help push ahead when it comes to President Trump's economic agenda. Now, of course, they're dealing with criticism. We have the House Speaker, for example, railing against some of the changes when it comes to Medicaid. But this administration is really trying to focus people on this idea of stabilizing the debt to GDP. Yes, this bill is going to add significantly to the national debt, but it is the other policies that they're working to put into place that ultimately they say will help grow the economy.

19:50The national debt pile, the big focus of Elon Musk, as we've discussed all week. Bloomberg to Tyler Kendall, terrific work out in Washington, D.C. Thank you very much. Let's get to another big story. The U.S. and China are beginning to implement their recent trade deals. And as part of that, there's some easing of restrictions on critical technologies, including export license requirements for chip design software sales, particularly in China. Bloomberg's Peter Elstrom, who leads the team in covering Asia tech and European tech, is with us. So what we're talking about is EDA. And as of last night, we're hearing from the key players, the key names in the EDA space, that they're now free to do business in China.

20:28What do we need to know? Yeah, that's right. So this is a strange one. It was just in May when we found out from the companies that they were going to have these export restrictions on China. Their EDA software is necessary to be able to design semiconductors. It was also strange the way that it got announced. It was the companies that actually revealed these new restrictions. It was not the Commerce Department, which is the agency behind that. Again, this time it's kind of similar. We had Siemens come out first. Cadence design systems and Synopsis are also affected here. They came out and they said, in fact, those restrictions are now being rolled back.

21:01They're not going to have the same kind of restrictions on servicing and supporting selling into China to support their customers. So what's going on behind the scenes? As you say, the U.S. and China are now in the middle of these trade negotiations. Back in May, China was kind of playing tough. They were not exporting some of the rare earths, the minerals that American companies needed. Now that the negotiations are proceeding, China struck this tentative agreement to be able to ship those again. and the U.S. and China are coming back to the table. And as a result, these companies are going to have a little bit more breathing space.

21:32But that's very unusual. You've got export controls, which are typically national security issues, now being traded as part of the trade negotiations. We're seeing some of the chip makers push higher. But really, the stories around the software names, cadence, synopsis, the two that you mentioned also pushing higher. We'll show those in just a second. The modern-day cutting-edge chip has billions of transistors, right? No human brain can do the blueprint or design for that. But there is some concern here that Huawei in particular, if they are having free access to EDA, that they can use that technology in that software platform to close the gap a little bit, you know, design their own cutting edge chips.

22:12Why is America not more concerned about that, Peter? Yeah, that's a key issue. In the technology race between the two countries, you have the U.S. ahead in a few different key areas. In particular, EDA is one of the most important ones. Of course, semiconductor equipment is another one. The Netherlands, ASML, is by far the leading player. Huawei, however, has been able to break through some of these barriers. It's become a national champion for China. It's helped try to focus on some of these key bottlenecks that they have. EDA is one of the key areas where if they're able to use these tools, they can step ahead and design some of these chips.

22:51They have surprised the world, I'd say, with some of their advancements in semiconductors, particularly this 7-nanometer chip that they used in one of their smartphones in the past. But now they have not been able to make some of the progress, have not been able to move forward quite as quickly. So these tools are going to help them make some of those steps going forward. Again, that's why it's so surprising that they're trading this away in the negotiations between the U.S. and China. Bloomberg's executive editor for Global Technology, Peter Elstrom, thank you so much.

23:26It's time for Talking Tech, and first up, DeepSeek ramps up hiring. The Chinese AI startup posted 10 positions on LinkedIn, indicating it may be looking to lure talent from outside of its homeland. The listings include three roles focused on general intelligence with the positions based in Beijing and Hangzhou. Plus, ASML and other European semiconductor stocks extended losses today. There was a report from Nikkei Asia that said Samsung would be delaying its completion of a chip factory in Texas. The report citing sources says the delay is due to Samsung struggling to find customers for the plant's output.

24:03And some news that broke this hour. CoreWeave's the first company to receive the latest AI system based on NVIDIA's newest chip. Dell delivered the first GB300 MVL72 rack of servers to CoreWeave and will deploy them in the US, with CoreWeave aiming to bring more of the tech online throughout the year. OpenAI is one of CoreWeave's customers, and the infrastructure player is saying the new systems will work well for larger, more complex models. That's the reporting. That's the detail. I want to get some more analysis on this one. Mandeep Singh of Bloomberg Intelligence is with us. There's this, like, constant tracking of the launch and ramp of NVIDIA's latest generation server design that has the latest generation chip and chip combination of it.

24:49CoreWeave goes first. What do you make of that? Well, when I look at NVIDIA's release cycle, they are on a one-year rhythm, and in between that year, Now they are launching the Blackwell Ultra before the Rubin series comes online. And look, they're almost giving a 50 % performance upgrade with the Blackwell Ultra in terms of token processing. So from that perspective, CoreWeave having that chip first really gives them a leg up over the hyperscalers, which will also get their NVIDIA allocation. But for CoreWeave, they have a backlog of about$26 billion. To convert that backlog to revenue, you need the most impressive NVIDIA chips first.

25:35And that helps with the faster backlog conversion. So really good news for CoreWeave, especially on the training front. Because a lot of these latest Blackwell Ultra chips, I believe, will be used for training. And then the older series will be used for inferencing over time. So CoreWeave's stock really pushed higher after news came out. I'm also looking at Dell. Help us understand the Dell component in this ecosystem. Yeah, well, Dell is the one who is making that server. So NVIDIA is providing the chip, and CoreWeave is really bringing those racks and the servers online to be consumed in a cloud consumption model.