So, I want to start with the news that has just been out in the last 24 hours, and that is this unveiling of an AI acceleration strategy. What does that mean?
What that means is that the Department of War has historically not deployed AI to its full potential. And in this environment, with AI really exploding in industry and for consumers over the last two years, it's the time to get AI to the 3 million employees at the Department of War across several domains: enterprise use cases like you'd use in any company for efficiency, intelligence which is specific to the Department of War, and for war fighting which obviously is specific as well, and using the max capabilities that these companies had invested hundreds of billions of dollars in their AI models, bringing them to the Pentagon and to the war fighter.
Yeah. And of course, you're leaning into the commercial capabilities that are already out there in an even bigger, more meaningful way. So, how quickly can you bring that into all of these different areas for the department?
Yeah, that's the right question because speed wins is what Secretary Hegseth said yesterday in his speech at Starbase with Elon. And we've already deployed Gemini, Google's model, to 3 million Department of War employees and we've had almost a million uniques use it, which means that people are clamoring for this stuff. So soon, in the next month or so, we're going to bring on Grok, xAI's model, and then hopefully some of the other frontier models as well. But the idea is we want the newest model in front of everyone across every classified and unclassified network so they could do their work in the most optimal way.
What have the lessons been that have been learned so far as you do bring some of these models in and you do start to deploy them across the department?
The lessons that we are learning and that we will learn, just like when people tried ChatGPT for the first time, they learned there were some imperfections. But the leverage you get on efficiency and how much time it saves you, and what it could do given the context windows and the amount of data it synthesizes, is incredible. So the way I view it is that if we can make 3 million employees be able to do the work of 10 million employees because they are now more efficient, more empowered, can use their time more effectively, then we're really going to increase the fighting force of that same workforce by where they could do more with less time.
Yeah. I mean, you mentioned Grok, and you guys are bringing Grok into the fold, and as you mentioned, some others as well. How do you navigate things like hallucinations and issues around guardrails? I mean, I'm just going to use the example of Grok because in the past what, week or so, there was some outcry publicly about some issues with images that were getting generated, maybe for lack of certain types of guardrails or loopholes there.
Yeah. Well, the guardrails are going to be built into the models and the way the model companies choose to do them. However, for the Department of War, what we're going to do is—and what we have done is—work with the model companies directly and say the Department of War needs to use your model and its capabilities for every lawful use case. So long as it's lawful, we'll use it. And in our country, those things are lawful. Now, on top of that, we have our own Pentagon policies of what you can and can't do on the computer systems, as you can imagine. And some of the things that people might do in the consumer world we may not permit in the Department of War because we're a workplace, right? And we have a mission that we've got to accomplish. It's not there for messing around on things that people have ideas on that they might want to explore in their personal life, but in the professional life, it's going to be used for professional purposes.
As a longtime tech guy, you'll appreciate this. When I've been saying, I'd say for the last couple weeks, that disruption has come to defense. So, how does this speak more broadly to this policy push, this strategy shift that we're seeing across the Pentagon right now to contract differently, to stoke more competition, and to move more quickly?
Yeah, that's exactly what we're trying to do at the Department of War. Under Secretary Hegseth's leadership and President Trump's leadership, they're both disruptors. They both have looked at the system and said the system doesn't work anymore. And the way it didn't work in the Pentagon was that we were relying on a small number of prime contractors to deliver systems—weapon systems, defense systems, and so on. And that number was 50 in the 1980s, and now it's down to five. So when you're down to five, you have fewer options for competition. And the business models were broken, too. We were paying cost-plus, which means that there's an incentive to have projects take longer to deliver and to have them cost more.
So instead, we're moving to much more of a fixed bid kind of world and much more of a they spend the capital, and when they deliver the system, they get paid, which encourages them to be on time. That also gives a window for new entrants—for the new Andurils, the Palantirs, the SpaceXes, and the 10 other new companies who have innovative solutions—to have a real shot at winning business at the Pentagon because if their software works or their hardware works and they could deliver it on time and they have financial backing, we can buy it. And we're trying to remove the barriers that previously existed to make those new entrants have their shot at the Pentagon business.
Just today, an announcement between the Department of War and L3Harris, a new first-of-its-kind deal where the government will take a stake that will ultimately convert to an equity stake upon the IPOing of the solid rocket motor business at L3Harris, which is now expected to spin out in an IPO later this year. A boost in production there, stocks trading higher on this. I look at this and I wonder if we're entering a new era of deconsolidation for the industry.
