AI's Power Demand: Fusion Energy, Palantir’s Record Earnings & Investment Banking M&A | Aug 5, 2025

0:13Welcome, everyone, to the Informations TI TV. My name is Akash Basfritz. It is Tuesday, August 5th, and we have got some new earnings from Palantir we are going to get to today. We are also looking ahead to Snap and AMD earnings tonight. We got a big show of guests lined up for you. The CEO of Fusion Company Helion is coming on the show. We are also talking AI code review with the CEO of Graphite. We got a great story about how advertisers and ad publishers have this love-hate relationship with Google going on right now. And we're going to close out the show by talking about how AI is already affecting investment banking.

0:50Now, before we get started, I want to highlight two big stories that we published over the past 24 hours. They are two of the biggest topics that are really dominating the news cycle right now. One is AI funding, and the other is M &A happening in the AI sector. Now, the first story is that Reflection AI is in talk to raise more than$1 billion. The startup was literally only founded last year, but the founders are Google DeepMind, former Google DeepMind researchers, I should say. So that very much does the trick. If you're wondering what Reflection actually does, the company is looking to develop open source models in the US to compete with China's DeepSeek and Francis Mistral.

1:29It is, of course, going up against meta platforms in that space. And so it has its work cut out for them. But$1 billion should get them started and on their feet. The other news that we were first to report is the latest on the Winsurf saga. You might recall that after Google snapped up most of Winsurf's leaders, Cognition, another AI company, bought what was left of Winsurf. Well, now Cognition has offered buyouts to around 200 employees from Winsurf. The employees have until August 10th to decide if they're going to leave. And in a memo, Cognition CEO Scott Wu said, quote, We don't believe in work-life balance.

2:07Building the future of software engineering is a mission we all care so deeply about that we couldn't possibly separate the two. One more quick headline before we get to our first guest. Linda Yaccarino, the former CEO of X, has taken a job as CEO of a telehealth company called E-Med Population Health. The company is focused on the popular GLP-1 drugs for diabetes and obesity. that certainly promises to be a change of pace. And you know what, Linda, this is an open call to you. We want you to come on the show and we want you to talk about your new gig. That is my open message to Linda Iaccarino.

2:43All right, let's get to our first guest, folks. One of the biggest topics in AI has been powering all of the data centers that companies need to house all of their chip and servers. And anytime that you talk about power, you have to talk about carbon footprint. And so that is what our next guest is trying to tackle head on. Helion has become a well-known startup in the arena of fusion power. The company generates energy by fusing two atoms together under intense heat and pressure. The company has raised more than$1 billion to do this from the likes of Lightspeed, SoftBank, and even Sam Altman.

3:18And I want to bring on Helion's founder and CEO, David Kirtley, to talk to us about when we can expect fusion to go mainstream. David, thanks for being here. It's good to have you. Thank you, Akash. Thanks for having me here. Yeah, today we're talking about Orion. That's our first fusion power plant. So earlier this year, we actually began construction of that power plant to get fusion power on the grid. So I want to get to Orion in a second here. Let's just backtrack for a minute. I mean, you're in the business of clean energy. You could have decided on anything, really. You could have done geothermal.

3:56You could have done winded solar. You could have even done biofuels. I mean, why did you decide to focus on fusion? Yeah, fusion is what powers the universe. It's what powers stars. It's the physical process that most energy that we use that generates that energy. But we don't really use that on Earth to make electrons yet anyway. And so when we got in, when the founders got into fusion and founded Helion Energy, our goal was not to just demonstrate that you can do fusion. Lots of folks are doing that, but actually do fusion to build the commercial product, to build generators that make electricity.

4:33And it changes the whole way that you do fusion if your goal is to get electrons on the grid. Absolutely, it's possible. And why is fusion better than some of those other clean energy alternatives that I mentioned? Yeah, so fundamentally in fusion, you're taking the most common elements in the universe, hydrogens and heliums. And then under intense pressure, our sixth generation machine got over 100 million degrees. You fuse those atoms together. They form heavier elements and release a tremendous amount of energy. But they do it without carbon. They can't melt down like nuclear power can. And if you can do it right and you have the design from the forefront, we believe you can do it very low cost, lower cost than any other power source.

5:19So it's lower cost. Tell us about when you will actually start delivering this power to customers. Yeah, we've now built seven machines, including our latest generation, which we call Polaris, that do fusion. And they're all in service of that power on the grid. We signed a deal a couple of years ago now with Microsoft, a PPA, a power purchase agreement, to deliver 50 megawatts to the grid in Washington state to power one of their data centers. And that's the work we're doing right now, actually pouring concrete of that facility today. And that's the facility you just broke ground on a couple days ago, right?

5:56Yeah, that's the Orion facility, what we believe will be the world's first fusion power plant. And so we're now pouring concrete. Actually, the second building is going in right now and beginning the construction of that facility. That will be a 50-megawatt scale facility, the first prototype power plant for that Microsoft data center. You know, I want to get your help in sort of putting 50 megawatts into context. And we hear about these massive data centers that all these big tech companies are building. Where does 50 megawatts sit compared to some of the other facilities that we've seen being constructed?

6:32Yeah, it'll be the first fusion power plant and the first fusion power on the grid, but it's still small compared to where we need to go. And so when we founded Helion, our goal was to build fusion in a way that is scalable, globally scalable to massive scale. I think it's pretty interesting that now when we talk about data centers, we don't talk about the amount of compute they have. We talk about the power use they have. A new four to five gigawatt scale facilities are coming online and they need power. And so our goal at Helion, in fact, where I am right now in Everett, Washington, is to have factories that build fusion generators so they can be deployed anywhere in the world.

7:12And so, I mean, this kind of leads me to my next question. I mean, 50 megawatts, it's big, but as you mentioned, it's still relatively small in the grand scheme of things. Still the first fusion power plant, like you said. But how long is it going to be before fusion can really start to make a dent in these massive energy demands that AI is requiring right now? Yeah, so 50 megawatts is that first step. We also have a development agreement with Nucor, a big steel company, for 500 megawatts. And so we're now beginning the work on investing in manufacturing to be able to build that scale of power.

7:48But still, we're talking about now gigawatt scale data centers. And so the only way to get that is investing in manufacturing so that you can mass produce these systems. So today, right now, we have conveyor belts that are building the subcomponents for fusion, capacitors, electronic assemblies, big high power magnets. And so we build those right now here in the United States. And we're investing now. Earlier this year, we raised a new fund to invest in manufacturing to turn up our capacitor development, one of the key electronic components, by 50x over the next year. So that we can rise to meet this goal.

8:26Eventually, where we want to be as soon as possible is not building one Orion power plant, but in fact, building generators, a generator a day off the assembly line is what I've tasked the company with building the manufacturing to support. A generator a day. And so, again, like, you know, a company like Microsoft, I mean, how many generators would you be producing for a company like Microsoft, just to put that into context? Yeah, we are, our goal now is thousands of these generators. So a generator a day is now hundreds of generators a year per facility. And so it's a massive undertaking to scale up the manufacturing to be able to meet those needs.

9:05The customers are there, clearly, not just AI data centers, but we started this journey even before the new increase in pressure on the grid and those power needs. And so clean, baseload, carbon-free power has enormous customer base. And so now it's a question of how fast can we build to meet that and build now having a truck of generators coming out of the assembly line. Well, and that was one of my questions for you was, you know, AI or not, you know, you've talked about this power purchasing agreement with Microsoft. It's slated to start in 2028. It'll ramp up over the course of that year and the year after.

9:45But I mean, three, four years is a long time away. And so you could totally see, you know, the AI hype fading over that time. But, you know, you've been at this since, I think, 2013 is when you founded the company. And so, you know, what other customers are you talking to right now other than big tech companies, other than those who are looking to power their AI operations? Yeah, so we started working with Microsoft in 2015 before the AI power boom happened. And the world needed Fusion then just like it needs it now. Clean, safe, always on, base load power. And so the customer base is pretty large, it's huge.

10:23So 50 megawatts with Microsoft so far, we have a 500 megawatt development agreement with Nucor, large steel manufacturer, which has clean, low carbon steel manufacturing and recycling, but they need power to be able to power that. And so those are our first two customers and that scales from the 50 to the 500 and our goal is to invest in that manufacturing to go meet the rest of that need. But honestly, you can look at the power needs throughout the world And we're just talking about large industry in the United States, but there's other places all over the world that need this clean, always-on baseload power.

10:56Do you think your next customer, you know, what sector do you think it'll be? And who are you talking to other than Microsoft and Nucor? I mean, our focus now is how do we invest in that manufacturing to meet that customer need? And so I don't have any new customers to announce today, but we certainly are pushing to be able to meet those needs. And so we point to what's really exciting about Orion, this first fusion power plant, is that it's actually plugged into the grid. So we're working really closely with the public utilities there to, in fact, releasing the land that we're building, the generator, from the utility themselves.

11:33And so that close working relationship means that we can go into, is proving, paving the way for us to go into all utilities. Last question for you before I let you go. You know, you talked about 2028 as being the start of that power purchasing agreement with Microsoft and when Orion will finally come online. Just very quickly walk us through what needs to get done, you know, year one, year two, year three. Between now and 2028, is there licensing involved from the government? You know, are you trying to have all the machines up and working so it can test by year two? Like, what is the Gantt chart, you know, timeline for the next two and a half years?

12:09Yeah, getting a power plant online in three years is really tough and takes a lot of work, so much so that we've already started. So the licensing and the permitting work of this started with the federal government many years ago, but also even working with the state regulatory agencies and permitting agencies to be able to even go and break ground and begin construction. And so some of the key things that we're working on over the next three years, I mean, the big one was the environmental permit. So this is the first time a fusion power plant has ever been licensed. We're really proud of that, that we paved the way to very quickly get what's called a SEPA permit.

12:50That is the main environmental permit for these facilities. And do it in a way that's cleaner and safer than many, many other power technologies. So that's a big one. But on the way, we continue to work with, as we've done with all of our seven previous fusion systems, we continue to work with the licensing and the Department of Health and Department of Ecology in Washington State to license these. And that's on that permit side. And so that work started years ago and will continue over the next three years working with the local community, making sure the community wants fusion. That's also just as critical.

13:23And the hardware. And how long will it take to get this hardware set up for Microsoft to start testing for them before 2028? Yeah, we've already started work on that hardware right now. We're prototyping it in our facilities, on the conveyor belts and the mass assembly lines that we have here in Everett, and building that hardware today, building the subcomponents, the capacitors, the power electronics. And so our goal now is to be ramping up over the next year, year and a half, that manufacturing, so then we can turn on the full production, build the subassemblies here at our manufacturing in Everett, Washington, right outside of the big Boeing manufacturing plants, and then put them on trucks, deliver them to Malaga, Washington, where the Orion facility is, and then install them.

14:11And continue the manufacturing at that point to go on to support Nucor and all the other customers that are coming online. Wow, look, it's a long journey, and I know you've been at it for a long time. So thank you, David, for coming on and giving us a glimpse into what is going on at Helion. That is David Kirtley, the CEO of Helion Energy. Palantir reported earnings last night. It might be the single hottest software company in the market right now. Shares have grown more than six-fold this year. This time last year, its market cap was around$70 billion. It is now worth more than$400 billion. And I want to bring on Jeffrey's analyst, Brent Phil, to break down the numbers for us.

14:51Brent, welcome to the show. Thanks for being here on TITV. Thanks for having me. So what stood out to you about Palantir's earnings last night? Let's go there. We've never seen anything like this. I've covered software for 30 years. I mean, to have a company put up 45 % growth and a mid 40 % margin is unheard of. So a rule of 90, we talk about a rule of 40, right? In the industry. And this is just the numbers are incredible. And then this morning, Dr. Karp on the CEO of Palantir mentioned that he anticipates that they actually grew a headcount this quarter, but they will shrink the headcount from 4 ,100 to 3 ,600 and grow revenues 10x.

15:31So if anyone thinks AI is not coming for your job, you better look at what's happening. Microsoft flat headcount, Meta put up insane numbers off of decelerating heads from three years ago. I mean, what's happening in the industry right now is pretty remarkable. And so I'd just say that if he can hold to that forecast, I mean, that just tells you what is happening with AI. And I think it's a real wake-up call for every software CEO that's out there. So the numbers were phenomenal. It's just really hard to think about now the combined market cap of Adobe and Salesforce is less than Palantir. Right, right, right.

16:11Let's talk about the forecast. I mean, what did the company forecast in terms of sales and profitability going forward? Yeah, I mean, they've had four straight quarters of accelerating growth. They do expect growth to decelerate. But again, albeit we still think that they're going to put up, you know, 40 to 50 % growth going forward. So I think their forecasts have obviously been too conservative. They've been blown through numbers. I think the real question is, can the sustainability be a 50 % growth company for the next four years? That's going to be required because at that end of four years, right now it trades at 75 times revenue.

16:48We've never seen a revenue multiple sustain like this in software. Snowflake went to 50, fell back to single digits. Datadog went there as well, fell back. We saw this in the client server area with Siebel. We've seen these movies before, and they never last in terms of the multiples. but even if they grew at 50 % for the next four years in a row, they would trade at 25 times revenue, which would still make it the most expensive name in software, you know, four years out. So just tells you, you know, kind of we're in, we're in a different stratosphere in terms of growth. I was going to say the 75 times, is it, I mean, is this meme stock territory?

17:28I mean, like, do you think that multiple lasts really? So if you think about like, Just compare the next two closest multiples. So Cloudflare is at 25 times revenue, and CrowdStrike is around 18 times revenue. And this is 75. So there's no cities between, there's no pit stop between those names. And so if you think about the best operating company in security, CrowdStrike, everyone's like, George Kurtz, best leader in cybersecurity, hands down. They'll pay 18 times for that. And that company's growing 25 to 30 % with a big margin. Cloud Flare, obviously, in a really good spot, protecting the internet, improving the speed of the network.

18:14Obviously, with AI, protecting all the agents and improving the speed of the agents on the network. That's 25 times. So, I mean, you're talking about a 15 multiple point differential to get to Palantir. And the growth rates are not that different, right? We're talking 30%, 30 % to 45%, 50%. The gap isn't that big. Again, I think kudos to Palantir. To answer your question, yes, it's in a completely different stratosphere. We've never seen it. We can't compare it to anything. I think the hard thing for any of us that are fundamental analysts is we've all been bitten by this. Investors were bitten on Snowflake at 50 times revenue thinking that it was the new norm.

19:00and it went back to like seven times revenue. 50 to seven is a big compression in the stock. So we just don't have any, there's no comparison. No, and I think that's sort of what I'm trying to, you know, sort of triangulate for myself is how defensible is this moat that they've built? You know, what is the moat even at all? The moat for Palantir is really the technology works. It's a combination of infrastructure and applications blended together. And as Dr. Karp has said, they don't even have a sales team really this is funnicked led and customer led that when xyz oil and gas company or financial services company has an roi they talk to other customers the customers come to palantiris say we need it it's not you know he doesn't believe in the steak dinners and golf that's not his his thing he said and i think we've all like said what if he had a really good head of sales remember oracle with ray lane remember um uh you know the best companies in industry, you know, even look at what Cloudflare did with their new head of sales coming out of Palo Alto.

20:04I mean, you look at what they could do if they really had a great go-to-market team, they could even grow faster. He's choosing not to do that. He's choosing to, I don't want to deal with any reservation at any high-end steakhouse or any golf club. We are delivering this in our product. So as far as we know, it's actually, it's inbound interest for the company right now. It's inbound. It's led by other companies having the ROI. And we spoke to a lot of these customers that are seeing that ROI. So this isn't hocus pocus. It's not like slideware. This stuff's real. And what they've done is they've really truly blended the application role and the infrastructure role together.

20:45And not many in AI have done that. Right. So. Last question before you before we let you go, you know, they have the the commercial business and they have the government business. Which of these two businesses are you more focused on right now for the company going forward? Commercial for sure. And so what I'd say is it's growing faster. Second, you can analyze it right. You can't go talk to the CIA or the army or any other department of the government, get the ROI case study. You can't forecast it. And so we said this. When we did the direct list for Palantir taking them public, we said like half their business is unanalyzable because you can't really get to those sources.

21:23The commercial side is growing fastest and it has the biggest opportunity. And what I would say in my limited government conversations and my friends that work in the government, I would say they have an incredible opportunity. I think the government's getting increasingly looked to them to automate. And you look at what's going on with Doge and some of the initiatives inside the government to reduce some of the software waste. And Elon Musk has tweeted repeatedly how bad the software is inside the U.S. government. There's an upgrade that has to happen. So I think commercial, to answer your question, but government is another engine that is in an incredible perspective for where they're at.

22:02They're not alone. There are other companies that are going after this market. You can piece it together with Databricks or Snowflake and other custom-built interfaces. But Palantir has done a really, really great job. And it's kind of shocked, I think, a lot of us that have watched the industry for a while. Well, look, I want to thank you, Brent, for coming on and helping us make sense of it. It was a pretty stunning quarter. And like you said, the question is, how long can this go on? That is Brent Phil from Jefferies. Thank you for coming on and helping us understand all that is going on with Palantir.

22:37AI coding tools have been getting a lot of attention lately, but the other half of generating code is being able to review it. And that has been a space that many different startups have started to focus on, including Graphite. The company has raised money from Excel, Menlo Ventures, and Dresden Horowitz. And I want to bring on the co-founder and CEO of the company, Merrill Lutzky, to talk about his business. Merrill, welcome to TI-TV. Thanks for being here. Thanks for having me. So look, AI code review, I think I know what it is, but tell us what the product does very quickly before we get into the business.

23:12Yeah, so Graphite is building the code review platform for the age of AI. What we mean by that is, as you said, so much of the attention in AI so far in software development has been on how can we use LLMs to generate more code? You have amazing tools like Cursor and Cloud Code Now that help engineers write code and build features faster than ever. But anyone who's worked in the software team, building in a professional software context, knows that that's only the first part of the story. You still need to then get that code reviewed. You need to have it tested. You need to deploy it to production.

23:50And all of that, we call it the outer loop. The inner loop is kind of where you're working locally. You're writing the code. You're building the features. that outer loop is where all the collaboration, iteration, and refinement happens. And even before AI was good at generating code, that often becomes the bottleneck at enterprise scale when you have thousands of engineers trying to work concurrently. And now with AI, that problem has just become 10 times worse because every engineer has 10 cloud code agents open and they're able to generate so much more code. But now you're just seeing teams that are getting stuck with that, you know, with a huge pile of TRs that need to be reviewed.

24:31And they aren't able to get the most out of these tools. And that's what we're solving is we think it's not enough to just apply AI to generating code. You also need agents that help you review faster, that help you merge all this code and deploy it. And that's what Graphite is, is a platform that helps you solve and move code as efficiently as possible through that outer loop with AI. So the thing that I'm so curious about with your business is, you know, we have all these AI coding companies and then we have these code review companies. I mean, isn't code review something that could just easily be tacked on for any of these AI coding companies?

25:09I mean, why do you have to have a business around code review on its own? It's a great question. And I think that it really comes down to the way that software is built and how collaboration works, where what we're seeing now, I think, one, I think a lot of these companies are building, are making some effort here. So you see Cursor building BugBot, you see CloudCode having a review tool now. And I think those are great. And it's very fun to see those companies recognizing that the outer loop is the bottleneck and providing a solution there as well. But I think if you look at the way code is generated, as long as there are many of these different agents within a company that are generating code, it still all needs to come together on one platform.

25:57And that's always been the way that it works. Every engineer has had their own IDE. They've had their own local development setup. And really, code review is the first moment in the development cycle where everyone has to be on the same platform and you have to have like have to support each source of code equally. And that is our belief with code review is that it's not unless unless Cursor or Claude owns 100 percent of the code that's being generated within a company. Right. That's not the right surface for review to happen. It needs to be a third party. The other part of that is like, you probably don't want the fox watching the hen house here.