OpenAI’s Shopping U-Turn Complications, Nvidia’s Groq Chip, Synthesia’s AI Video for Enterprise

30:07On the flip side, companies are constantly repurchasing shares to sort of keep the share count in check. Is that your understanding of the way that this, in the end, will affect the company's cash flow profile? Because we all know that we look at the cash flow statement, stock-based comp, I mean, it's a non-cash item, but then I guess it's the repurchases where it comes back to bite you.

30:32Jackson Ader:Yeah, it's certainly the repurchases that come back to bite you, from a cash flow perspective. But it's kind of twofold. If you asked an employee, hey, how much money do you make? They would not say only the cash component of their salary. They would include both the cash and the RSUs that get granted to them. each and every year. And so on the one hand, you have investors that are, you know, potentially new investors to software, like a value investor, you know, or someone who's been taking a broader look at other sectors and saying, okay, how does software stack up relative to other sectors that I've covered in the past?

31:19Jackson Ader:And they say, I'm looking at the gap income statement. I'm looking at the quote unquote economic value of this company. And if there is a huge swath of the expenses that are just not being accounted for on the cash flow statement or on non-GAAP earnings, that investor is going to say, no, no, no, no. The total expenses here should also include this thing that employees say is important to them and that would certainly be demanded by those employees on an ongoing basis. Then you bring up a great point, which is, okay, well, you give shares. I mean, that is value. That's dilution. That means you have to issue shares in the future.

32:03Jackson Ader:And so that's why in our analysis, we choose to say, all right, we forecast a fully diluted share count and we forecast that fully diluted share count to grow over time, even into perpetuity as part of our discounted cash flow analysis. And so you can make an argument that maybe forecasting future dilution and also including the stock-based compensation in your economic expenses is kind of double burdening. I don't know. I hear good arguments on both sides. I think what is definitely clear is that software companies are saying with regularity, we are repurchasing shares to offset dilution. and in the metrics that we all use to value these companies like unlevered free cash flow or straight free cash flow or free cash flow to equity, those repurchases are just, they lie outside of those reported metrics.

33:01Jackson Ader:And so it ends up inflating the multiple at which these companies trade and you have to start doing some adjustments if you are going to be comparing software companies with their 13 % stock-based comp to other companies with their 1 % stock-based company. Right. So now let's bring it back to where we started the segment. Do you think that it's going to come down? Are we seeing evidence that it's starting to come down? I mean, we talked about the trend we've seen to date. Where is this going? We are. We are starting to see the trends come down. Again, you got to dig through the Ks, which have just been filed to get to the actual grant values.

33:45Jackson Ader:A lot of companies only disclose the restricted stock unit grant values in the Ks filing. But our friend Nate on our team actually just pulled some numbers here just a second ago, if I can double check them, that the median company as of 2025, stock-based compensation as a percent of revenue down to less than 12 % in 2025 versus 13 in 2024. before. And so we are starting to see it come down. We saw year to date in the first three quarters, the actual grant values come down as well on a year over year basis. Some of that is certainly manager preference, right? Doesn't sound like it's employee preference, but some of it is manager preference to shift more toward cash.

34:35Jackson Ader:And we've heard some companies be explicit about that. And then the sadder portion of that really is that layoffs, have started and might be continuing. And so it's not just stock-based compensation that's coming down. It's all types of compensation that are coming down as a percent of revenue as margins float higher. Great. Well, Jackson, when you write part five, come back on the show and tell us more about what you found. I want to thank you for coming on. That is Jackson Ader from KeyBank capital markets here on TITV. Anduril has bold ambitions for its work in the defense sector, and at the center of attaining and fulfilling the big government contracts it is aiming for, the company is building out a giant factory in Ohio.

35:23The Information's Deputy Bureau Chief of Finance Corey Weinberg visited the factory recently and published an in-depth profile about not just what he saw, but what it tells us about Anduril's place in the defense tech race. I want to bring on Corey to share more about what he learned. Corey, welcome back to the show. It's great to have you here. Hey, Akash. So where is this factory that you visited? Well, I journeyed to the great state of Ohio. The factory is about 20 miles south of the city of Columbus in Pickaway County, Ohio, surrounded by soybean fields, cornfields, quite a rural area just near Columbus.

36:05a lot of industrial warehouses and a lot of data centers, as you might expect. Why did Anduril pick that state of all states to build their new factory? Yeah, no, good question. I mean, it's there are a multitude of reasons. One, you know, they got, you know, nice, a nice tax incentive package. That's always, that's always nice um but it's also got an air force base there has an air field there and they're going to be building fighter jets uh as one of the the kind of weapons or vehicles that they're building in the factory so that's useful for for testing and things like that okay so you get to the factory and walk us through the the big questions that you were trying to get answered as you were going into this visit to the factory uh well it was my first defense factory that i had been to so uh what what do these things look like uh was one question on my mind and uh yeah i was there with andrel executives and so what is their messaging and framing going to be about what they're trying to do because i think just taking a step back here akash like this is a huge pivot moment for andrel It is a turning point.

37:24They are going from a company that has not been producing at scale. They have been producing in a few factories around the world, in part that they had purchased. They have an R &D production center in Southern California where they're based. This is their first attempt to really produce and manufacture at scale and actually try to, you know, sort of fulfill large contracts that they would be winning from the U.S. military. So, you know, I was kind of curious about all the, you know, sort of details around how do they actually do that? Like, how do they how do they sort of make that a big jump into the big leagues?

38:08And remind us, what is it that they're intending to make at this factory? yep so a few things first off that they that they flagged um uh probably the most high profile project that they'll be working on there is what they've called the fury fighter jet this is an autonomous unmanned aircraft that they are a finalist for in this competition from the u.s air force to create a collaborative uh combat aircraft it's supposed to be a plane that flies with the manned fighter jets in battle. And this one would be smaller, cheaper, autonomous. And Anduril is a finalist alongside another defense company called General Atomics to actually produce and build that aircraft.

38:57And the Air Force should decide within months who is actually going to win that contract or how they're going to divvy it up. Anduril will also be building, they say, their Barracuda cruise missile in the factory. And intriguingly, as is often the case with sort of defense companies, there is a classified confidential project that they say will be in the factory. So, you know, I want to go back to what you were talking about with the scale that they have to achieve with this factory. I mean, as they compete with these bigger companies that have been around so much longer, I mean, you mentioned General Atomics for this one contract.

39:40You also mentioned in your story Lockheed Martin, Raytheon, these companies that have been around for decades, they have their existing facilities. Did you get a picture into what the actual innovation is that Andrel is banking on that will help it win these contracts and ultimately actually make it a significant player in this game? It's banking on this sort of shift within the U.S. military and militaries around the world. to go from sort of what's termed large, expensive, exquisite systems for war. These would be the Tomahawk missiles. These would be the F-35 fighter jets. These huge, expensive systems that we buy, that the U.S.

40:34government buys for war. and real is saying we're not really trying to build those we're maybe trying to like we're definitely trying to build more expensive things than we were before we're building an unmanned fighter jet for instance but they're saying we can do it more cheaply um uh we can do it uh faster uh and so and with sort of commercial parts that was kind of a big part of the message that i got from andreal executives at the factory is, you know, they are trying to reduce supply chain risk and the risk that some component manufacturer goes out of business or has cost overruns by sourcing a lot of their material from sort of commercial providers of jets or boats or what have you.

41:25So is it fair to say that it's a little bit less product innovation that's more like supply chain innovation or yeah i think it's more i would say it's business model innovation it's taking and that's i think a good way to you're getting at the right question about andrel because i think it's one thing that i've been trying to like get my head wrapped around as well is this company and a fair number of like defense tech companies as well they're trying to not take tech risk um you know they're trying not to say like hey we have this like totally untested technology that we want to sell it to the government.

42:02You know, they want to take stuff that it's been tested commercially, or we know that works. You know, they don't want to necessarily push the boundary too much on how do you actually manufacture them. This factory is very simple. It's essentially an assembly line. And, you know, try to ensure that that's not what's holding them back is the tech. Right. Now, you spoke with some of the executives on the ground there. One of the people that you mentioned the story you spoke with one of the leaders for the manufacturing operations. Keith Flynn was his name. I wonder what you gleaned from the conversations with him.

42:39You know, what's his own background? You know, how is he approaching this innovation in manufacturing that you talk about? Yeah, so they had a couple ex-Tesla guys, for instance, including Mr. Flynn, who are manning that Anderle factory or the company's manufacturing efforts more broadly. I mean, look, I think they're confident. They're saying, you know, and they've been through these weeks of production hell on, you know, sort of the Tesla Model S vehicle and things like that. You know, they say, look, there has been this wave of new manufacturing innovations that we've seen at a Tesla or SpaceX or wherever, you know, where you can really use software to simulate and sort of figure out beforehand how you're actually going to make this thing.

43:31You know, sort of how are you going to actually attach the wings to the body of the airplane? You know, sort of all these things that you can do to kind of really dumb down the actual assembly process and ensure that you have a low error rate in your factory. And he's saying that essentially, you know, they're right at the beginning of that process where they're going to, you know, start actually building and assembling a large number of missiles and aircraft and things of that nature. And so there's sort of this mixture of humility and sort of boastfulness or confidence, I would say, on behalf of Andrell.

44:12It's a bit of like, hey, we've kind of been there before with, you know, car companies or airplane companies or whatever. And we're just going to apply that to defense. You know, we'll see. It hasn't really been done before. But what do industry experts, I mean, people who maybe are people who are not at Anduril or any of the other companies, I mean, what do they tell you about this manufacturing innovation that Anduril is pitching? I imagine that Lockheed and Raytheon and General Atomics, I mean, I imagine everyone's kind of doing this, right? Yeah, definitely to a degree. I mean, like the difference here with Andrel, they're using a, you know, a ton of venture capital that they have raised to essentially build out in front and take a bet that, you know, they're going to win these contracts and that this factory is going to be humming.

45:06That's like a key thing is they have this, you know, risk capital that's willing to, you know, sort of take a bet on the fact that Andrel is going to win these contracts and they're going to keep these assembly lines humming. um a lot of kind of traditional larger prime contractors um they're not building a ton of new factories or and if they are they're tied to a very specific contract that they've won um so it's just a little bit of a different way of doing things and it's you know the difference is the fact that you know andreel sort of they don't have public shareholders that they're trying to uh satisfy they have uh injuries and horowitz and founders fund who's telling them you know go big swing for the fences but i i guess the question i i was hoping to get at was the the industry experts that you spoke to people who have been in this industry for a long time who have seen many companies come and go do do they feel like this approach that anduril's taking with vc money with these newer factories i mean are they confident in anduril's prospects here or do they sort of have the viewpoint that you know what it's harder than you think let's see you know No, it's a good question to get at.

46:20I do think people think it's hard. I think Andreal sits in this position more broadly in the industry where they have achieved the status where people are kind of still this mixture of skepticism and also giving them somewhat of the benefit of the doubt. They kind of sit in this middle ground. And that's why this sort of production kind of push for the company is such a big deal. This is like a bit of you got to prove it and show it to me that you can actually win big contracts and deliver significant, you know, sort of stockpiles of weapons or aircraft to the government because they haven't really done it yet.

47:05And so they haven't also had like huge missteps either, I think. I think they have become a big name in the defense industry, despite being a lot smaller than their competitors. And I think everyone sort of thinks like, hey, these ideas that they have around kind of cheaper, more mass manufactured sort of products for the Pentagon is one that people generally—it isn't that hotly contested necessarily. It's just, can you deliver on it? Right. Great. Well, Corey, it sounds like an exciting trip, and I imagine that there will be more trips like these to come as the company grows. Thank you for coming on.

47:50That is Corey Weinberg, our Deputy Bureau Chief of Finance, here at The Information. Our next segment is supported by Google Gemini. AI video generation is one of the most exciting applications of the technology. And to talk about it, I sat down with executives at Synthesia, a popular company in that category. I spoke with Synthesia's Chief Technology Officer at SVP of Customer Success. Here is that conversation. Peter and Carly, welcome to the show. It's great to have you both here.

48:24Jackson Ader:Thanks for having us. Okay, so Peter, I want to start with you. Remind us, I mean, in the AI video landscape, there's a lot of players out there. What is Synthesia's product? How do you differentiate yourself in the market? What are you guys really good at? I appreciate the question. We are the world's leading AI video platform for business. Right now, there's roughly 65 ,000 companies, including more than 90 % of the Fortune 100. And basically, they're there to turn ideas into professional video without cameras or studios. Yeah. Yes. And Carly, the way I understand it is your product is actually a little more focused on enterprises.

49:02You're not so much going for, you know, if I want to make a video on the weekend myself, it's more for what? For companies that want to create like training videos for their employees? Yeah, exactly. Okay. So Peter, tell me a little bit about the advances in AI video models then that have really made a difference for you. I mean, they've come a long way. At the same time, I figure there's probably technology that you guys want more advances in to help you get to even the next level. Yeah, the models have changed really quickly. It feels like every few weeks, you have a new best-of-breed model.

49:38And for us, we use every one of the best models we can find. We're trying to look for contextual content for the most part, and we think about that as B-roll. Today, we use Vio, Sora, and any other model that's highly effective there for us. It allows the creator to be able to use more creative motions. As we think about the technology on the back end, we use LiveKit in our new skills component. And so that allows for deep interaction with the avatars rather than just having a static video. So for us, it's all over the map, and we can dive into really any part of that because each one of those is super interesting.

50:21And do you guys make your own models? as well? We do. We have our own models. And so for us, we have our own video models for static video. We have our own voice models. We have our own soon upcoming real-time avatar. Right. So I've always wondered this. I mean, when you have an enterprise customer and, you know, I imagine these are like video series that they're putting together, how do you decide then, you know, which model is better for what types of videos? Is one better at creating cartoon-looking graphics versus real-life graphics? Or how does it all work? No, that's a really good question.

51:00So for us, why we have our own models, like the other Frontier models provide some fantastic videos. That said, when you look at why our customers will choose our models for most of the cases, what they really want to see is long-form content. Most of the other Frontiers are really gauged around user use cases. And the other thing is that they're going to have multiple scenes. And so it's the length of time. It's longer than 8, 12 seconds. Some of our content's in minutes, and that's just not possible with any other library. The other thing is between scenes, they want consistency. They want to make sure that their avatar, their voice sounds consistent between those frames.

51:35The other thing is that they want to be able to apply things like a brand kit. They want a logo on a shirt. They want it to be able to be in a setting that makes most sense to them. So if it's a jeweler, they want it in their jeweler, not a generic one. So there's a lot of specifics that cause us to build our own models. That said, when we think about, you know, putting that avatar or some of that detail into context, we want to see that person in motion. And that's where the frontier models are just absolutely spectacular. So for us, it's kind of combining the best of breed in both these spaces to get the best outcome for our customers.

52:11Carly, how do you then reconcile all of this video creation with this fear of AI slop, which is very much the topic of conversation these days? I mean, video, gosh, I don't even know sometimes. I'm on my feed, what's real and what's not. How do you guys approach that? Yeah, of course. So I think generally we're less focused on content, which shows up in consumer feeds. And I think that's what you're referring to is a lot of the slop that's out there. In the enterprise, obviously, the bar is different. So what really matters is reliability, control, security, whether or not our solutions save time or ultimately improve outcomes.

52:47So what we really focus on is helping our customers generate high-quality video that's easy to create, easy to update, and really grounded in what the business actually needs to communicate and teach. And what does ROI look like for your enterprise clients? Is this literally just that, hey, I didn't have to hire the five-person production team to produce that video series, and hence those are the cost savings? Or, you know, can you reach more? I imagine if it's enterprise, I mean, this is really a cost center for them, right? Yeah, I mean, sometimes. If we take the example of global onboarding, so a lot of our customers, you know, will partner with learning and development or HR teams.

53:26And what they're really focused on is potentially creating onboarding materials, management training, compliance-related training. And the content there needs to be frequent, it needs to be clear, and it needs to be localized. So taking that global onboarding example, a team might upload an onboarding deck or a handbook and then turn it into a series of short videos, leveraging an AI presenter, and then ultimately ship that to every new hire. And then when policies change, systems change, maybe there are updates needed to be made to product details. It's just as simple as updating a sentence or two in the script and then regenerating the video.

54:00So as you mentioned, no need for sort of reshooting or leveraging external agencies to manage that entire process. And then obviously there's efficiency through the one-click ability to translate into a multitude of different languages. So a number of our customers benefit from just the shift in what they were previously doing manually that now can be done dramatically faster. Peter, have you guys created your own training videos, your own employee engagement content with your own product? We do. Yeah, one of our core onboarding components is to actually have our tools introduce you to the company.

54:38In my particular department, it's introducing you to the development tools. it's walking you through a day in the life, it's walking you through how to do your basic day-to-day life. So absolutely. And we find it far more effective. So for us, our employees are just overloaded with information. The world's changing faster than you can ever imagine it. So for us, it'd be hypocritical for not to take advantage of the fact that video is one of the best ways to communicate, train, share knowledge, and do the basic training. So of course, our focus is to make it super simple and reuse it. And, you know, this is one of those times, you know, drink our own champagne, eat our own dog food.

55:15Well, and the reason I ask the question is because you're selling to enterprise software companies and we have this debate ongoing in the market right now. To what extent will people just build their own tools with AI? To what extent will they use enterprise software? These are the clients that you're serving to. I mean, as you are thinking about building your own tech stack at Synthesia, are you using only AI to build your own CRM or your own enterprise software functions? Or are you still relying on many of the companies that people are kind of scared might go away? No, definitely. I'd say a really good question.

55:52So for us, the way I think about it, in a Biverse build, we are good at certain things. I came from an Amazon environment as a prior employer and build was often the answer for different reasons. For us, though, we are going to absolutely specialize in the things we do best and that's providing enterprise video. In the case where we need a HRIS system, billing systems, we definitely are using best of breed third-party systems for that and we're building where it makes sense. For us, we are just like any other enterprise vendor. We do one thing, we do it well, and we want others who specialize in their area to be able to do their thing.

56:28We focus on human training. That is what we do best, of course, with many things. And it would be silly for us to enter into all these other spaces so it would distract us from our cooperation. Last question for you, Peter. Chips have been in the news this week with all the latest developments and newer chips that are coming out. Do you think at all in your role about, hey, we should be trying to run our workloads on this specific chip because of its capabilities? I just wonder how you think about that as an application layer company, because there's a lot of companies in the middle where you're working with cloud providers or inference providers, for example, that they probably handle the chips.

57:08But it matters to you, right? Oh, it absolutely matters. We're an application layer company, so we're only one separate move from the hardware. We have to think a lot about that. That said, we do an enormous amount of cutting-edge R &D and voice and video spaces, and that's critically important to our business. But we're going to rely on the best-in-class infrastructure to build, train, and optimize them. That is fundamental to our space. We use NVIDIA infrastructure on public cloud to optimize their performance. We also have strategic partnership, research relationship with NVIDIA, so that certainly makes it easy.

57:41And of course, they're an investor in Synthecia, and so for us, that makes it much easier for us. better work on the video and the voice models and computer vision. It really aligns closely with theirs. And so that allows us to have a collaboration on the software development as well. Great. Well, Peter and Carly, I want to thank you for coming on. That is Peter, the CTO at Synthesia, and Carly, the VP of Customer Success at Synthesia. Thank you both for coming on the show. Thank you. Thank you. That does it for today's show. A reminder, we are on this stream Monday through Friday at 10 a.m. Pacific, 1 p.m.

58:17Eastern. I want to thank you all for tuning in. We really do appreciate your viewership. Make sure to subscribe to the information on YouTube and follow us on X, Instagram, TikTok, and check us out wherever you get your podcasts. I'm already excited for our next show tomorrow. Have a great rest of your Monday. Bye-bye for now.