Inside Kalshi’s $4B+ in Annualized Revenue, Chip Infrastructure Startup Eliyan Hits Unicorn Status

0:13Welcome, everyone, to The Information's TI TV. My name is Akash Pasricha. It is Wednesday, August 12th. Today on the show, the information exclusively reported that Calci hit record revenue boosted by World Cup betting. We'll take a closer look at what the numbers mean for the company's momentum. We'll then talk to a pair of chip infrastructure entrepreneurs about the current state of the AI buildout. Next up, we have a scoop about what former White House crypto czar David Sachs is up to these days. And we'll close out the show with a discussion on how a key U.S. government agency is cutting its Salesforce bills.

0:50It's going to be a great show, so let's get right on into it. The information has exclusive reporting on Calci's latest financials. My colleague Yueqi Yang reported that story. I want to bring her on for the latest. Yueqi, welcome back to the show. It's great to have you here. Hey, Akash. Okay, so what do we know about Calci's latest financials? So Calci's annualized revenue has surpassed$4 billion. And this is largely driven by event contracts on World Cup game that happened in June and July. And this is a really fast growing business because the annualized revenue is already doubled the level that we reported about just two months ago when it reached more than$2 billion annualized revenue in June.

1:40And the financial numbers are coming out because Kaoshi is currently in advanced talks to raise funding at a$40 billion valuation. $40 billion, and so what was the last valuation that it fetched? In May, Kaoshi raised about$1 billion at a$22 billion valuation. So effectively, within just a short span of three months or so, they're nearly doubling its valuation. And so the momentum that it's seen from the World Cup, I mean, the World Cup happens once every four years. So what's the sense among people you talk to? Is the momentum going to last? Was this just a blip? I mean, how is the trading data looking right now?

2:26Yeah, this is a great question. And when we look at the weekly trading volume for both Kaoshi and Polymarket, it has dropped since peak level during World Cup. So we do see a slowdown in activities after World Cup concluded. But in the fall, there are a few big events coming up as well. There's going to be NFL games and the midterm elections is going to come up as well. Now, what about profitability? Do you have any reporting on the costs here? That's a great question. So we don't have a number on the profits for the company, But we do know that Kaoshi is incurring a lot of expenses for the month of June, for example.

3:14They spent a lot of money on marketing expense. And this is because they wanted to really build up a direct relationship with retail customers as they were watching World Cup. If you watch any games of World Cup, you might have noticed that Calci's name was on the stadium, on TV, on online advertisement. And these are the kind of marketing expenses that they had to incur in order to make sure that they become a household name. Now, one of the pieces in your story that intrigued me was the partnership that Calci has with Robinhood. what's the status of that partnership and how much has that affected its revenue growth, either positively or negatively, I guess?

4:02So Robinhood also offers prediction markets. And in the past, Robinhood mostly sent its orders from its own customers to Kalshi Exchange. And just about a year ago, Robinhood made up about 50 % of the trading volume on Kalshi. However, the two companies have become less dependent on each other over time. And this is in part because Robinhood has now directed more of its orders to an exchange that it backed along with SIG called Rosera. And the latest data in the second quarter showed that Robinhood now only makes up less than 20 % of the trading volume on Kaoshi. Now, less than 20%, and the other player here, of course, is Polymarket.

4:54How does the revenue here for CalSheet compare to Polymarket? So Polymarket has been losing market share against CalSheet. In the past, Polymarket was actually a front-runner for a long time. Polymarket was the biggest prediction market. However, they're slower than CalShit in moving into the US market, and PolyMarket's market share is now trailing behind CalShit. Now, you talked a little bit about the legal aspect and the regulatory aspect to the business in your story. How could that affect its revenue profile going forward? So a big legal and regulatory uncertainty for prediction markets, including Kaoshi, is their sports event contracts.

5:42There are a few lawsuits currently ongoing with state regulators that allege that prediction markets such as Kaoshi were offering illegal gambling operations. And it is widely expected that these lawsuits will eventually go to the Supreme Court, which will decide whether the states have the right to supervise these contracts or the federal regulator, the CFTC, which Caoshi and other prediction markets have a license from, will be able to regulate these offerings. But meanwhile, we're also noticing this new trend, which is Caoshi could potentially be paying taxes to some of these states. And that will be an alternative outcome for CalShi, which will allow them to keep operating in the states, but it will eat into their profit because they will have to pay taxes to states.

6:37And in North Carolina, for example, the state last month passed a law charging a 6 % tax on prediction market fee revenue generated from users in the state. And of course, we have the election coming up in November. So, I mean, I guess people must be expecting another bump, a World Cup-sized bump, maybe, if you will, for these platforms? Well, see, I think midterm election is very much a U.S. event, whereas World Cup is a global event. Yeah, but U.S. market is important for Kaohsi. Great. Well, Yueqi, I want to thank you for coming on. That is Yueqi Yang, our crypto reporter here at The Information.

7:23A new chip networking company has hit unicorn status. Elion a few weeks ago raised$145 million in a funding round led by Seligman Ventures at a$1 billion valuation. Networking has become a hotbed for AI funding lately. I want to bring on co-founders Ramin Farjadrad and Patrick Soheili to talk more about the chip sector. Ramin and Patrick, welcome to the show. It's great to have you here. Thanks for having us, dude. So, Ramin, I want to start with you. You describe your product on your website as a die-to-die, chip-to-chip, and rack-to-rack chiplet interconnect technology. What exactly does that mean in simple terms?

8:03So basically, AI is in a transition point that is now becoming a system problem, not a chip problem anymore. And to basically enable a right system solution, connectivity becomes the bottleneck. And Elian is set to solve this connectivity challenge for the next generation AI systems to make it more efficient and maximize the usage of the compute that for years companies have been trying to optimize. The next challenge is the connectivity. And our goal is to think from ground up to solve this connectivity solution with the right approach. So, Patrick, I mean, chiplets, what does that mean in the context here?

8:47These are chips that connect other chips to sort of make up a rack? Or how does this actually work? In the kind of the old way of doing things, we had chips, basically semiconductor devices that's sitting inside of a package. And they were monolithic, meaning one chip or one chiplet, if you will, inside of a package. package. And what has transpired over the last few years is given Moore's Law's limitations or lack of growth, you can't put more transistors in less time and grow the chips any further. So now you've gone to create a parallel architecture by putting more chiplets inside of the package, and each of those are called chiplets.

9:34We have to connect them together inside the package, then you have to connect the package together like you have in the past. And because you're putting them into racks on servers, those racks now need to be connected together. And sometimes those server racks have to get connected together. And we address all of those with our technology. So, Ramin, is this, I mean, are you guys competing with sort of the big networking chip companies like Broadcom then? Pretty much, yes. You know, after all, we're all building connectivity solutions. And our goal is to, as a startup, to always think ahead what the problems that industry will face, what the walls they will hit, and come up with the right solutions ahead of time.

10:21Effectively, as I say, is be where the ball will land rather than chasing the ball. And that's how we can be years ahead of them in, you know, coming up with the solutions for the hyperscalers, for anyone that is building these AI solutions. So what is the advantage then for Elyon against Broadcom? Why do you think you can beat them? Just being ahead in some of the solutions or technologies that we have developed. For example, we started from die-to-die many years ago before industry had any focus on this. Today, we have the fastest die-to-die, high-spanly die-to-die at the lowest power effectively.

10:58And also extending that all the way to rack-to-rack, having the most efficient solution, both in terms of power reliability and effectively as these AI pods start to grow and you have to connect these GPUs together over a longer distance as opposed to within the same tray or within the same rack. People are looking for a replace for copper that has limited distance and we've come up with an optical solutions that provides all the features of a copper interconnect in terms of reliability, power costs, but much longer reach to connect these AI. So AI basically racks all together. Right, right. Patrick, I want to take a step back a bit.

11:43You know, we have talked a little bit about the memory crunch on this show and the memory crunch. I mean, people talk about it as just a single thing. You know, there is a memory crunch and people are dealing with it. Prices are going up. You guys are very close to the space. Are there nuances to this memory crunch that you think people aren't talking about enough? Is this all the same resource for people trying to get after, or are there different dynamics here that we should be paying attention to? It's a good question. I think the world is effectively in a very tight bottleneck, if you will.

12:19There's three suppliers that are at capacity. There's a concentration of very few customers that are buying most of that capacity. And just the way things are going, it's not letting up anytime soon. And, you know, creating a fab that manufactures these devices, very fine geometries and very specific specifications that are just tough to hit in a large volume, scalable manufacturing kind of a way is not easy. It takes years, multiple millions of billions of dollars to build them. So Samsung, SK Hynix, and Micron are working really hard to kind of get there and meet the industry's demand. I can't think of too many nuances other than the fact that everybody's trying to catch up with the insatiable appetite that the industry has for AI systems.

13:10What about the Vera Rubin rollout? You know, one of the things that we have reported on here at The Information is with the Blackwell chips, there were some hiccups with the rollout. that Vera Rubin is actually going a lot better, according to our reporting. Patrick, in terms of what you're hearing on the ground, how is that rollout looking right now for those chips? I think NVIDIA does many, many things right and well, hence the market share and the position they've got in the marketplace. The dynamic is a little bit different from where we sit. We help the customers at NVIDIA who don't necessarily want to be NVIDIA customers as much as NVIDIA would like them to be larger customers of theirs, to have some independence and democratize them.

13:59As you know, all the hyperscalers have their own programs as well as buying NVIDIA chips. So Rearubin does have a significant presence in the marketplace. But on top of that, a lot of these guys are developing their own chips with their own solutions. I mean, TPUs are now in generation eight, going to nine, eventually ten. And then same thing, you know, lots of advances at AWS and Microsoft and Meta as well. And the foreign data center guys also doing the same kind of thing, Alibaba and Tencent and Baidu. So the world is trying to effectively come up with solutions that are somewhat different than what NVIDIA does, yet competes with them in a very effective way.

14:44In order for them to do that, they need technology like ours, which NVIDIA has on the inside, and not necessarily willing to share to open up that marketplace so we'll be so ramin you guys are selling to to uh many of these hyperscalers then who are developing their own chips who are your customers yeah many of them that uh try to you know nvidia is not uh cheap and uh with the lot high markup stuff of course they have to pay and you also see the amount of investment that they are making to build their data centers hundreds of billions of dollars even borrowing to do that. So if they can cut that number by a factor of two even, they would be well ahead.

15:25So they've been working on and thinking about developing their own accelerators to reduce that cost significantly. Of course, a part of that is one part is building the right compute chips, but the other part is connecting them more efficiently because by is the connections that is a bottleneck how much of a utilization you can get from an accelerator chip if the connection interconnect is not good you don't get that benefit out of it right the that technology is rare and that's something that we provide for that yeah you touched on something interesting akash if i can just follow up with that um if you want to get a clue on who our customers are you could look at our strategic investors uh corporate vcs typically don't invest uh for the return on the on the investment as much as it is a component of it.

16:17But their interest in investing us is mostly because they see us as a technology enabler for them. We have all three memory suppliers. We have all three fabs. We have two switch manufacturers. We have a couple of data center folks themselves. And we have three of the largest chip makers that are building CPUs and GPUs for the marketplace. So all of these guys are in the ecosystem that would leverage a technology for their products. so patrick last question for you then you know as we are sort of in this environment where you have amd for example unveiling their full rack scale system nvidia obviously is known for their racks um you guys are sort of catering to this market where customers are trying to put together their own entire system sounds like bit by bit piece by piece these sort of seem like uh two different ends of the chip market.

17:11On one scale, you have companies releasing racks. On the other hand, you have customers saying, hey, we want to sort of piece together our own system and customize it the way you want. Which of those two movements right now is dominating the momentum? It sounds like you guys are really suggesting that customers are going to want to piece together their own racks, and that's going to be the future? Yeah, it's a combination of all three. I think NVIDIA will not go away. I think building their own chips at the data centers will not, or the hyperscalers would not go away, because there's a lot of dependency on how they want to optimize their inference models or training models per their chip development kind of projects.

17:58And lastly, the chip makers, the AMDs, the Intels, the ARMS of the world, they have no intention of walking away from this market either. In order to be more competitive against NVIDIA and to think about how to optimize the entire data path, most of these large semiconductor companies, as well as the data centers themselves, are now in the RAC business. So that's their way of consolidating the path, making sure that it's optimized and delivers the value at the end of the day. Great. Well, Ramin and Patrick, I want to thank you for coming on. That is Ramin and Patrick, two of the co-founders of Elyon here on TITV.

18:37Thanks for having us. Thank you. The information has exclusive reporting on David Sachs returning to the venture capital game in a big way following his stint at the White House as AI and cryptos are. I want to bring on our AI and politics reporter, Leo Schwartz, who reported that story along with our venture Capital reporter Julia Hornstein. Leo, welcome back to the show. It's great to have you here. Thanks for having me. So what is David Sachs up to these days? So David Sachs just left the White House in March from his AI and crypto czar position. As we've reported extensively over the past few months, he is still involved in the policy game in D.C.

19:17But as we reported yesterday, he is back in the venture game as well. He's targeting around$1 billion for a new fund for Kraft Ventures, which was the fund that he started in 2017. And it seems like they're making investments again. And how big is this fund compared to other funds that Kraft Ventures has raised in the past? So what's different about this one is it's a sole fund. The past couple, they'd done a two-fund structure with early stage and more growth separated. rated. This is just one fund. And our reporting showed that it's going to focus later stage and just like later than seed. So basically series A onward.

20:00And it's one. The previous one was around 1.3 billion, all told, between the two funds. So this one's a little smaller, but it's the sole fund structure. But you said that they're going to be focused a little bit later stage compared to what they typically have done. So it seems like they won't be focused on seed anymore. They've made some really notable seed investments in the past, but this one will be more Series A onwards. Why do you think that is? I mean, just given, you know, the run that David Sachs has been on, the job that he's had at the White House, do you think there's a strategic reason here?

20:35I mean, the only reason I can think is maybe just that you have to raise big checks nowadays. You know, like a seed round is no longer what it used to be, Series A, I mean, that could mean anything right now. Yeah, I think this is a better question for Julia, but does seed even exist anymore when a seed round is$100 million? I feel like even what today is seed is, you know, Series C in the past. But yeah, it does seem like there's a lot of interest in, I don't think you could say Series A would be growth, but certainly a bit later stage and backing companies that are further along. And you're seeing that strategy from other firms as well.

21:13So, OK, let's go back to David Sachs here. So remind us then again about what he had 130 days in the White House officially. What were you know, what was the sort of resume that he walked out with in terms of, you know, key things he achieved, I guess, in his in his short term? And how do you think that might translate to what he might invest in now with Kraft in the second stint here? Yeah, so Sachs was really the champion of this hands-off regulatory approach that a bunch of Silicon Valley like Mark and Theresa brought to this second Trump term in office. But obviously Sachs was actually in the White House in this special employee role.

21:58He had this Quinn hat of doing both AI and crypto. Obviously, there was a lot of movement in both. Arguably, crypto had more early on, both in crypto and AI. They released action plans that basically set out both an executive priority and also legislative priorities. In crypto, you had the passage of the Genius Act, which focuses on stable coins. The other big piece of legislation is more stalled. That's clarity that deals more with market structure regulation. You can look at our great reporter, Uechi's pieces on that recently. uh and then i would say more in recent months ai has taken the limelight crypto has receded a bit more as prices have decreased sacks actually left in march right around the launch of anthropics mythos when they're turned into this huge debate around how the white house should deal with these new cyber capable models uh and i think sacks clashed a bit with these new figures coming into the White House debate on AI, who seemed to champion a more safety first approach.

23:05Sachs continued to boost this more hands-off approach. And that really came to a head in this debate around the executive order that created a voluntary framework for AI models to share their models with the government before release. Sachs, even though he was out of the White House at that point, was able to convince Trump not to sign an earlier version of that framework or of that executive order in late May, arguing it was too onerous for companies, even though the executive order was signed about a couple weeks later. Does he still have an impact on Trump? What do we know about that? Yes. So our understanding is that he still has a presence in D.C.

23:44He still certainly has Trump's ear when it comes to AI policy as AI becomes one of the biggest issues in Washington. And then he does still serve in this role on this advisory council called PCAST, which is focused on science and technology. He's the co-chair of that, along with Michael Kratzios, the director of the Office of Science and Technology Policy. And PCAST has a number of very notable tech executives on it, although what it's actually doing in reality is a bit unclear. Great. Well, Leo, I want to thank you for coming on. That is Leo Schwartz, our tech and politics and AI reporter here at The Information.

24:23In our latest Applied AI column, my colleague Laura Bratton reports on how AI is forcing organizations to rethink their software spending and in some cases even reshuffle certain vendor relationships. Laura joins me now to share with us what she learned. Laura, welcome back to the show. It's great to have you here. Good to be here. How's it going, Cash? I am doing well. Let's talk about software spending. So you had the opportunity to talk to somebody from the U.S. Department of Agriculture, which is not an agency or a department, I guess, that we have really talked about on this show. Tell me about the conversation that you had and what you ultimately learned.

25:02I learned that the Department of Agriculture is trimming its spending on Salesforce and investing more heavily and using AI capabilities from C3. And I think it's just a late example of how we're seeing different software budgets from large organizations shift around as they become pickier about which AI tools they're going to use from which providers. So it's tapping C3 and it's reducing its Salesforce expenses? What's going on? Yeah, so that's what the CIO, Sam Berry, told me. He said that specifically for this task correspondence tracking, so AI capabilities that can categorize and organize letters that the USDA receives from voters and different agencies and Congress, it believed that C3's capabilities for this specific task were better than Salesforce's.

25:53It wasn't necessarily a pricing decision because technically they're going to be spending more on C3, about$9 million over a year, which started back in January, the contract did. Whereas they're spending on Salesforce for this task would have been about$4 million for a license and then$2 million in payments to outside consultants. So I think for Barry, what he told me is it was just more about picking which providers had the best AI capabilities for this specific task. But we should say it still has a sizable footprint with Salesforce, right? Absolutely. So the Department of Agriculture is set to spend through its contract up to$277 million between 2025 and 2030.

26:37I will say that's slightly less than the big enterprise Salesforce agreement they had between 2020 and 2025, which was about$294 million. And again, these are the biggest Salesforce enterprise agreements that the USDA has. They have some other smaller contracts to spend money on Salesforce's software, but these are the largest agreements. And we're seeing the value of those agreements come down a bit between the five years between 2025 and 2030, which I think is interesting. So what do you think this says then broadly about the shift that we are seeing play out in enterprise software, or at least the shift that people are scared, software people are scared of playing out?

27:21I mean, people change vendors all the time. It's hard to do, given that these are multi-year contracts, and this sometimes involves shifting tens of thousands of accounts to a new platform. But I mean, is this the SaaSpocalypse at work? Is this an isolated event? I mean, what does this tell you about the current situation? The reason I found this conversation so interesting is because I talked to a lot of small companies that have pulled back spending on Salesforce or ripped out Salesforce altogether to vibe code their own solutions. but I haven't really talked to large organizations or enterprises that we're pulling back, they're spending on Salesforce.

27:56And I think this is the first example we're seeing. And, you know, I've had background conversations with sources where they've told me, oh, I've heard of XYZ large enterprise that wants to pull back its Salesforce spending. But we haven't really seen this reflected in Salesforce's financials yet. It's still growing at a healthy pace, but investors are scared that, you know, What I heard from the USDA could become a broad trend and that could threaten Salesforce's pricing power, which is why I think the stock is down so far this year. Because Salesforce has been this kind of golden child of the software industry and had a decades-long hold on its customers.

28:31And we're starting to see that wane. And what does it mean for C3's business? Tell me about how big a win this would be for them. Yeah, so C3 is not doing so well. in its most recent April quarter results, we saw that their revenue fell by more than half. And CEO Tom Stiebel, who came back on as CEO recently, said that the results were unacceptable. And so, you know, it's not in the best position. And I don't think one contract is enough to totally shift the trajectory of the company. But I think it is a signal at least that, you know, it's not over for C3 and maybe they could turn things around.

29:09But, you know, it's... I don't think we really know what's going to happen with that particular company. I think overall, Salesforce. Well, I think it's going to get bought. I mean, I think that's the, I think, you know, I think, I don't know. I don't know if it's going to be Salesforce, but I mean, the company, C3, has been, it's been, you know, a bit of a tough journey for them. They've struggled to find their footing. As you said, revenue was declining. They had, I think Tom Siebel, he left because of health reasons. and then he came back. Sounds like their technology is doing something for someone, including the USDA.