61. Harnessing Software Supremacy in Defense with Shyam Sankar | All Quiet on the Second Front

The more diversity of approaches within the Department to go exercise this and see do we have what we actually need and how do we leverage that R&D to do this. I'll give you a real example that has nothing to do with defense for a second. We built the Operation Warp Speed supply chain in five weeks from a complete dead start during a national crisis, but I could only do that in five weeks because two years earlier in oil and gas production I had the original R&D stimulus to build this very sophisticated supply chain digital twin modeling infrastructure to optimize hydrocarbon production. So being able to take in these diverse stimulus from 50 different industries plus government means that the government, the whole nation, got to benefit from that. And I think there are countless examples that that is the story of commercial software.

Yeah, just there for 18 years. Lucky number 13 for me.

Yeah, I joined Palantir almost 18 years ago as the 13th employee at a period of time where we really aspired just to solve a handful of problems for a handful of institutions in the Intel community. And of course, the company's become something much bigger than that. Half of what we do now is in the commercial world, really what I call software-defined production, helping Chrysler and BMW and Airbus and great American manufacturing companies reindustrialize. And the other 50% in government, helping people close kill chains, really delivering an AI-enabled operating system for the warfighter.

Well, I think we've come a huge way from those early days where I think the Intel community had a front door through In-Q-Tel, the Department of Defense had no front doors, no side doors. You know, today now everything's open, and I love that, embracing innovation, leaning into it, disrupting itself. On the other hand, I think we may have forgotten one of the quintessential lessons of that period, which is that leadership matters. You know, how did we do this special procurement? Who came up with these ideas? Who stuck their neck out? And kind of lionizing, I think there's an opportunity here to really remember that we're all defense innovators. It has always been a hard road. You've always had to fight the bureaucracy, and some of that is a little bit of self-therapy, understanding the road will be hard. You know, whether it's Rickover and building the nuclear navy, I think Zumwalt said Rickover is one of the top three enemies along with the Soviets of the navy. So you know, I think these disruptors are going to have a hard time, and this disruptive community coming together and understanding what we're after, finding the leaders that we can back, and the leadership within us to drive the department forward.

Well, I think that's one of the things, because you remember when it first launched there were a bunch of folks who were like, oh, it's another just named initiative that's gonna go nowhere. I think as you sort of peel it back and if you just ask how a few times, you'd be like, well, hey, DepSecDef wants to push into low-cost attritable swarm capability as fast as possible. Path one, and just go chop wood in the Pentagon for like three years and get it on a quad chart and get it through some board, and then in eight years we field it. Or path two, she says, f it, sends it, tells the news we're doing it, lights a couple executive officers' asses on fire and sends them out to go figure it out. And now all of a sudden we're reprogramming what was there, 200 million in the NDAA for that last year. I think that's one of those things people forget, is you can't really force through the process, you've got to just come over it or you've got to go around it.

Yeah, and I think the what does winning look like and what are the behaviors, right? And like I spent a bunch of time working at Toffler Associates under Alvin and Heidi Toffler and thinking about, you know, first wave, second wave, third wave. And it's really easy to be like, hey, the building is still in the industrial age. But it was reminded, I was talking with Steve Blank the other day, right, and talking about all sort of these innovation, entrepreneurial mindset. And you take away a bunch of the jargon, whether it's DOD or whether it's from B-school, and it's intellectual sort of agility and the scientific method. And a little bit to your point on leadership, it's leaders creating that, you know, safe space is like an overused term, but that safe space for folks to iterate super fast. And I mean, that's where ID Force became JIDO, right? It was an '05 action officer, Ernie Benner, who went up there and she made a pitch, got like 100x more money than he thought he was going to get, came back to General Mattis and was like, we got a thing. But he had the space and they let him push and iterate on it, and the task force became something, became something, became something.

I guess one of the great American strengths in the present moment is software. You know, 86% of all tech market cap is American market cap. America is so good at software that we almost underestimate how good we are at it. You know, think about the fact that there are zero Indian or Chinese enterprise software companies that are competitive on the global stage. So that tells you that software is not just about IQ, otherwise you'd expect some sort of distribution relative to population. Tells you there's something very cultural about it. And I think that culture comes from Bob Noyce, the co-founder of Intel, the co-inventor of the transistor. And Bob Noyce was in Iowa. So there is this caricature in Silicon Valley where maybe it came from Israel or India, and you know, no, really we did import the culture of Silicon Valley, we imported it from Iowa. It's deeply rooted in Midwestern values, a willingness to play positive-sum games. Bob Noyce coined the term open-door policy. And I think that also explains why it's so hard for other countries to copy people, like their culture, they're committed to their culture. So you can come and be a tourist in Palo Alto and you can see Stanford and think maybe it's the research university, you can see Sand Hill Road and think maybe it's the venture capital, but it's actually something much deeper that you're going to have a very hard time replicating. So when I think about the present moment, we have exceptionalism in America in software. It's not evenly distributed. You know, I think when GE tried to reinvent themselves as a software company with Predix, you can't just hire 5,000 software engineers in the Bay Area and become a software company, there's something more to it. But as we think about the strength of our nation, the idea that our national security is supposed to feed American prosperity, I think it's incumbent on the software-literate community within America to help reinvent the industrial base, not just the defense industrial base, the American industrial base, to meet this present moment when our greatest competitor is now the world's best at mass production. What we were at the dawn of World War II, now we find in our competitors. And you know, as a David in this story, we're gonna need a software slingshot to beat Goliath.

The things that have me optimistic, if we look at something like Army TITAN, that the Tactical Ground Station, they really viewed that as a software-first acquisition. So yes, it is a truck on wheels that communicates with space and beyond the line of sight sensing, but actually much more important to them was the software that was going to enable and close that kill chain. And they put that at the heart of what they were actually focused on. So I see the increasing recognition of the primacy of software and the differentiated capabilities to win. The thing that still has me a little concerned is that we don't quite, I think one of the brutal things about software is that it is extraordinarily expensive to build. You want to have as broad of an R&D base as possible to amortize that expense against, and not only because it's expensive but also because of the varied stimulus you get. You know, you don't want to be the sole customer who has to think of all the good ideas and all the roadmap. You know, you want to be able to spread out some of that pain and maybe have other customers anticipate your needs before you even realize it and just be the recipient of that value. And so that is the tech business model, which is that you gain a huge amount of leverage on your fixed costs and the R&D that you're doing by having many customers to do it with. So it results in higher gross margin at the expense of a huge startup fixed cost and the risk of getting there. The historical government business model is built around the idea that, look, we have exquisite needs, we're probably the only buyer for what we're doing, the rational, why would we want to transfer that risk to industry? Why would people want to absorb that? So maybe we should pay them cost plus, what, sure they get a reasonable profit, but in return we're going to control the roadmap and own the risk. And I think that really does make sense for the Cold War era, the prior Cold War. I think it is an impedance mismatch to really getting America's deep capital markets and tech experts on the problems that you have at hand. And I've been really focused in the context of what I've been calling the first breakfast, I view this as sort of the antidote to Bill Perry's Last Supper. How do we explode the industrial base, so not consolidate it down, but how do we bring in tons of new entrants? Well, we need to find faster ways to getting these companies to revenue, getting them to the starting line. We're doing a lot to improve, we the community at large, doing a lot to improve acquisition, improve the number of front doors, the inclusivity. But then there's also a technical component. How do you actually get people to that starting line? How do you start getting them into reps with operational users, into the experiments where they can learn too, where they can see ideas, and then they can put their private capital at risk to build the features they think warfighters need, so that we can get more of that market effect going on here.

Yeah, admittedly it is a little confusing. The way I kind of see it is that, you know, the opposite of one mistake is another mistake. And so much of the software policy is a reaction to bad historical software experiences, feeling locked into specific decisions that have been made in the past. So trying to avoid lock-in at all costs means that maybe you actually don't get the capability you need. You know, that everything has to be custom, that perfect control comes at the expense of actually the primacy of winning. And so the way I think about it is really the commercial analog. The only place in the US that I find people talking about vendor lock-in is DoD. You know, the 50% of my business that's commercial, people talk about switching costs, people have standards. I mean, the Google-Oracle Supreme Court case says that you can't copyright these API definitions, right? So the API definition becomes how you manage modularity in the modern enterprise, and people are always thinking about in the context of their architecture, how do all my systems communicate through that API boundary. Now the government thinks about that in terms of a government-owned reference architecture or government rights to those interfaces, which I think they have no matter what. There's actually no configuration based on the court ruling that means you wouldn't own that. So I think we need to have more of a focus on demonstrating how these commercial companies can come in into an interoperable ecosystem, deliver value, be swapped out, getting closer to something like continuous competition. I think the reality is the commercial markets are completely brutal. I have never met a commercial company who wakes up and wants to disrupt themselves, but their competitors wake up and want to disrupt them, so they don't get a vote. You know, so they have to respond to getting punched in the face every day, which means they go to their vendors and punch them in the face every day saying, give me more to win. And the cycle time for our competitions in the department are, you know, first of all they're kind of existential. We don't want to wait until we're in a real fight to figure out if we have everything we need here. So we need to create these, the more diversity of approaches within the Department to go exercise this and see do we have what we actually need and how do we leverage that R&D to do this. I'll give you a real example that has nothing to do with defense for a second. We built the Operation Warp Speed supply chain in five weeks from a complete dead start during a national crisis, but I could only do that in five weeks because two years earlier in oil and gas production I had the original R&D stimulus to build this very sophisticated supply chain digital twin modeling infrastructure to optimize hydrocarbon production. So being able to take in these diverse stimulus from 50 different industries plus government means that the government, the whole nation, got to benefit from that. And I think there are countless examples that that is the story of commercial software.

Yeah, I think first breakfast is meeting its implementation moment. The first moment here is really with OpenDAGGER, which is part of the CDAO effort under Maven to create and evolve Maven to being a big tent, multi-vendor ecosystem. So with a defined set of APIs and a software development kit that makes it very easy for new capabilities to plug into the breadth of the Maven kill chain. So if you're a very promising AI company and you want to work on target nomination, if you can emit targets in this format, it's going to show up inside of this integrated kill chain. On the other end, if you're a company working on target-effector pairing, you're going to be able to plug in. And this modularity enables different commands to experiment with, in the context of JIDO or even their own exercises, bringing in their own indigenous capabilities. It enables software factories to plug in. So I think the full spectrum of third-party commercial software, first-party green-suitor developed capabilities, and arriving at a synthesis of how these things are actually going to play together in combat. When I look at our experiences in Ukraine, one of the most interesting, I think, socio-cultural observations is that they mobilized, you know, in the beginning about 100,000 Ukrainians who were by and large software engineers or product managers in their day life, in their civilian life pre-war, and now they're infantrymen. And when you're interacting with them, asking for feature requests, it is at a whole different level of generativity. They know what they're asking for, they know whether this is a one-hour feature request or a one-week feature request, and they know how they would translate that into combat power for them. So I think there's huge returns for the Department to have this sort of software literacy across the force. It doesn't mean that at the front line they're sitting there writing code, maybe they should, maybe they could, you know, their job is not to develop all the code but to actually increase the lethality of the force through that collaboration with the entirety of their first-party industrial base. They certainly have great people in key writing software, but also the vendors that they rely on to deliver this capability.

I think at the most abstract level, you do want the government to be the product manager in the way that we would use that term in a software company. You know, it's like, I know how I want these things to integrate, I know what my workflow is, you know, and I want to be able to have that opinion and express that and drive my various partners towards delivering that integrated experience for me. I think what can get in the way of that is that that's different than being like the head of engineering. And so there are going to be lots of opinions on how you pull that together, and I think that's where you really want to leverage the software supremacy in the US amongst many commercial companies to kind of arrive at that. And so staying focused on the workflow layer, I think, is the most important part where there's going to be the most differentiation. And I even think the most successful enterprise software really views themselves as building an Iron Man suit around a principal. You know, if it was Operation Warp Speed, you're building infrastructure for General Perna to manage the breadth of the operation. If you're trying to work with warfighters right now, like you better be wrapping this around the CCOMs and their leadership in order to deliver and maximize lethality. And I think that's one of the things I've really come to appreciate from the Afghan NEO to the present moment is how much you're actually thinking about every operation as this is, software is a malleable weapon system you want to wrap around the commander. So that's going to vary by commander, it's going to vary by the operation of the moment. And so we should be measuring, this is the OODA loop that matters, we should be measuring how quickly can we adapt our software to the present moment, not what is this software. And I think we lose under the truism saying that software is never done, I think we lose the actual thing is like the most valuable software compounds. There's more to do tomorrow than there was today, there's more to do the next day than there was tomorrow. And so if that's true, if your roadmap can only expand when you're doing it right, then you think about the fact that like, yes, it's probably going to cost you the same amount every year and you're going to be getting exponentially more as a consequence. And I think maybe one of the hard parts with the Department is it doesn't intersect well with how Congress and the Department think about their budgeting, which is I want to have some sort of capex-oriented big upfront investment that goes away and I want to be able to meter out what I'm paying based on what capabilities I think I'm going to get.

Yeah, absolutely. You know, I see this with differential. So if we leave the US for a second, we think about like, if we think about places like Japan and Korea, they actually struggle with software as a service. They also have a capex mindset. They tend to be hardware-based economies, you know, they're not excellent at software yet. And of course, I would bet on America every day on that for reasons we've discussed. And I think the capex mindset really limits you from being able to capture the value of software, where so much of the value is not articulable ex-ante. Like, you only realize you built something valuable after you built it. This is also why I think it's suitable to transfer the risk to private industry. You know, if you think about how much it takes to get a company to success, you also need to factor in the nine companies that cost just as much to fail. So the amount of leverage on your return here, like where do you want this to go? And I always feel like there's probably, we're probably lying to ourselves a little bit because you don't really see acknowledgement. You know, if 90% of commercial software fails as you're building the company, shouldn't 90% of government software also be acknowledged to fail? But we, you know, these programs become too big to fail and they have to keep going, and that seems inauthentic to the commercial experiences with software.

What I'm really worried about, you know, we have this unique moment where we have over a hundred billion dollars of capital that's been deployed in the national interest in defense tech. I view this as a return to an American industrial base. You know, the present moment where you have these pure-play defense contractors is the aberration. Chrysler used to make missiles. General Mills, the cereal company, used to make inertial guidance systems for ICBMs. With Ball Aerospace being sold to Collins so that Ball can make aluminum cans, I view that as kind of the end of an era where we had this diversified base. I think that's important for the reasons we discussed, you learn so much in commercial that drives productivity and innovation that provides national advantage. And I think we are potentially doubling down on this mistake by pushing cost plus as the way of building software. It is the antithesis of the tech business model of getting leverage on your fixed costs, right? And that means we might be turning our back, I think it's a big short on all of these VCs who have invested a hundred billion dollars here. We want that capital flowing in the national interest. That's how we're going to build the software slingshot for the near-peer competitions that we're currently in every single day. So I think recognizing the way in which cost certainty, cost visibility is actually getting in the way of winning is what I'm most concerned about in the present moment on the pessimistic side.

The companies are here. I'm excited about the founders. I'm excited about the energy. Like these folks are committed, they're going to win, they are hard-headed, they're going to push through the bureaucracy. Like we're going to do it as great Americans. Like they're already in the fight. And so I'm very confident that we're going to win. It doesn't mean the journey is going to be easy, and we're going to have as much to fix inside as we do outside. But I would not bet against this.

Man, thank you, brother. Yeah.