管理層發言
Good afternoon. I'm Ana Soro from Palantir's finance team, and I'd like to welcome you to our second quarter 2026 earnings call. We'll be discussing the results announced in our press release issued after the market closed and posted on our Investor Relations website. During the call, we will make statements regarding our business that may be considered forward-looking within applicable securities laws, including statements regarding our third quarter and fiscal 2026 results, management's expectations for our future financial and operational performance and other statements regarding our plans, prospects and expectations. These statements are not promises or guarantees and are subject to risks and uncertainties, which could cause them to differ materially from actual results. Information concerning those risks is available in our earnings press release distributed after the market closed today and in our SEC filings. We undertake no obligation to update forward-looking statements, except as required by law. Further, during the course of today's call, we will refer to certain adjusted financial measures. These non-GAAP financial measures should be considered in addition to, not as a substitute for or in isolation from GAAP measures. Additional information about these non-GAAP measures, including reconciliation of non-GAAP to comparable GAAP measures, is included in our press release and investor presentation provided today. Our press release, investor presentation and other earnings materials are available on our Investor Relations website at investors.palantir.com. Over the course of the call, we will refer to various growth rates when discussing our business. These rates reflect year-over-year comparisons unless otherwise stated. Joining me on today's call are Alex Karp, Chief Executive Officer; Shyam Sankar, Chief Technology Officer; David Glazer, Chief Financial Officer; and Ryan Taylor, Chief Revenue Officer and Chief Legal Officer. I'll now turn it over to Ryan to start the call.
Our Q2 results are unprecedented, but entirely unsurprising as the abrupt market shift in large language models that we've been warning you about for years is now here. We delivered 93% year-over-year revenue growth, our highest ever. The story of this quarter is, once again, our U.S. business. It now comprises over 81% of total revenue and grew 115% year-over-year and 23% sequentially. Our U.S. commercial revenue growth accelerated to 149% year-over-year and 28% sequentially. And our U.S. government revenue grew a remarkable 90% year-over-year and 18% sequentially. These top line results are accompanied by a Rule of 40 score of 155% and $1.22 billion of adjusted free cash flow. We closed 220 deals worth $1 million or more, of which 98 deals were worth $5 million or more and 73 deals were worth $10 million or more, record highs across the board. These results are a clear indication of the profound value we've unlocked both for and with our customers who dared to cross the chasm with us. In contrast, enterprises that are not using Palantir are seeing their token meters spinning endlessly just to get slop without any correlation to value. This token model may be working for the labs, but it is not working for anyone else. It's breaking corporate budgets without results to justify the expense. And worse, companies are paying to give away their most important secrets, the very basis for their competitive advantage, ultimately contributing to the commoditization of their own businesses as their secrets become the training data embedded in the foundations of all future models. On our side of the chasm, what enterprises demand is AI sovereignty: owning the operational definition of the data, logic, actions and security of their enterprise. An organization's data is its treasure; in its richness is the alpha. We are fully aligned with our customers, building a stack that enables the compounding of their alpha. Our deep alignment allows our customers' ambitions and our own to become one. As Kirkland and Ellis highlighted, 'through our work with Palantir, we have built a new operating model for legal services, one that centralizes and compounds the expertise of our most senior lawyers.' What used to take days for a lawyer to analyze, discuss and draft now happens in minutes. This would be impossible without the Ontology. We do not see this as a vendor relationship or a one-off endeavor. We think this is a revolutionary change in how our work gets done. This is what we do with our customers across industries, and the deals we are closing are a testament to the monumental shift in the AI market that's underway as we speak. In our U.S. commercial business, we closed $2.1 billion in total contract value with a 271% year-over-year growth rate on a dollar-weighted duration basis. A multinational technology company began working with us in the fourth quarter of last year at one operating company and expanded on their success with our platform to deliver revolutionary impact across their full portfolio, converting to a 3-year nearly $370 million deal last quarter. Customers are decisive and bold about taking the next step with Palantir. A global asset management firm started working with us in Q1, then converted last quarter to a 3-year $35 million total contract value deal spanning asset management automation and investment life cycle intelligence across four verticals. After an agent camp in May, a global software and services company signed an initial $15 million five-month deal last quarter. And a leading nonprofit health system signed a pilot at the end of 2025 and then last quarter converted to a 3-year partnership at $37 million total contract value. Our U.S. government business remains a source of extraordinary strength with momentum across both defense and civil. We continue to take great pride in our U.S. government work, equipping our nation with the most advanced battle-tested AI capabilities. For Palantir, this is our calling. Our customers are making the decision to go deep with us with greater urgency and conviction than I've ever seen before, choosing AI sovereignty over dependency and compounding their alpha in a way that their competitors and adversaries will forever envy. I'll now turn it over to Shyam.
Thanks, Ryan. Ryan just talked about the incredible customer momentum behind sovereign AI. I want to spend some time on the underlying product investments that positioned us for this moment. AIP succeeded because it's the best, most ergonomic environment for AI in the enterprise. It integrates mixed human and AI teams across heterogeneous interdependent workflows and delivers the fastest implementations that turn tokens into real economic value for our customers in complex high-stakes environments. What makes it work is exquisite and layered: data integration and transformation, ontology and actions, security and audit, workflows, agent SDKs, agent orchestration with telemetry and observability, evals and customer-specific benchmarks, AIP Evolve and our latest investments in post-training, both supervised fine-tuning and reinforcement learning. Every layer builds on our foundational primitives and every layer flows together. This infrastructure captures the rich operational telemetry to feed the compounding loop, an automated model factory that runs inside the customer security boundary and accumulates intelligence in weights they control. Last quarter, I said tokens are the new coal and AIP is the train. Now our customers can build their own locomotives. AIP is where your sovereign AI is built, deployed and compounded. We're excited for a new era, not one of benchmaxing but of benchmaking. The assumption that the frontier is actually the best performing is just not borne out in practice. Within 24 hours of bringing Nemotron Ultra into our stacks, we found five production tasks where a standard Nemotron Ultra model without post-training beat frontier models. This underscores that a handful of common benchmarks can be gamed. That era, benchmaxing, has ended. The new era is benchmaking. A customer-specific benchmark is not just a scorecard. It's a hill to climb. It's the normative orientation that directs your entire operational workflows and post-training pipeline. It defines what better means on your terms for your business with your strategy and then everything optimizes against it continuously. And because the benchmark is yours, the trade-offs become yours, too. AIP gives you the control plane to trade cost, performance and latency against each other and to decide where you run your weights versus theirs, workflow by workflow continuously. Now you can have a continuous improvement cycle that compounds your alpha into your weights. AIP was built for this. We continue to see that our product is winning head-to-head. The others are focused on productivity. We are laser-focused on turning tokens into real economic value for our customers. The reality is that the market has created far more intelligence than it has converted into value. A more powerful model does not solve this problem. The limiting factor is the rate of AIP deployment. A major Silicon Valley tech company recently ran a bake-off, a frontier lab and its deployment team against AIP and our forward deployed engineers. Remember, only Palantir has forward deployed engineers. Everyone else has sparkling sales engineers. The lab picked a ticketing automation problem and failed to deliver anything of value against it. We built agent swarms for each of our customers' customers, proactively recommending marketing, packaging and pricing changes to drive revenue and utilization. This work converted into a $10 million annual contract. The lab was shown the door. Same customer, same timeline, same models. The only difference was AIP and Palantir's unique forward deployed engineer tradecraft, and it was determinative. It was an incredible quarter in U.S. government, measured not just by 90% year-over-year revenue growth, but by mission impact. Maven continues to deliver for the Joint Force from the factory floor to the foxhole. We had our first program in the Maven platform launched this past quarter, where a government program of record chose Maven as the platform that they will operate their program in, taking advantage of our open data standards, ontology, developer tooling, peering, security and other platform primitives to go faster and deliver seamless experiences and capabilities to the department's chosen command and control platform. Maven has also continued to win as the developer and builder platform for the Joint Force with over 25,000 builders. Uniformed service members, civilians, contractors and companies are developing agents and applications in and on the platform at the speed of war. And for all that growth, our Department of Defense trailing 12-month revenue is still less than 25 basis points of the Pentagon's budget. Finally, last week, we held our first American Builder Summit in D.C. to celebrate the Americans who stepped forward to join the American Tech Fellowship and to let them make the case that AI is creating jobs and prosperity by sharing their stories and showing what they built. We created the American Tech Fellowship because the most transformative applications of AI that we saw were being driven by people without traditional tech backgrounds on literal front lines and factory floors. We now have over 1,000 American Tech Fellowship graduates. One of our speakers, Jona, who joined Talbot as a submarine parts manufacturer 13 years ago, straight onto the factory floor. He's a proud blue-collar worker who still turns wrenches for a living. And he built an AI application that took production planning from 30 to 40 days down to less than one. AI alone cannot do that. It takes AI in the hands of the American worker, the tribal knowledge earned through success and failure on the line, the insights only they have. The models are commodities, but the American worker is not. I'll turn it over to David to take us through the numbers.
Thanks, Shyam. We had a phenomenal second quarter, delivering our highest ever reported year-over-year revenue growth rate of 93% and our highest ever adjusted free cash flow of $1.22 billion, representing a 63% margin and 115% growth year-over-year. We surpassed $1 billion milestones in GAAP net income, adjusted free cash flow and adjusted operating income. Revenue in our U.S. business grew 115% year-over-year and 23% sequentially in the second quarter. Our U.S. commercial business accelerated to 149% year-over-year and 28% sequentially, and our U.S. government business grew 90% year-over-year and 18% sequentially. We closed $2.132 billion of U.S. commercial total contract value bookings, representing growth of 153% year-over-year and 81% sequentially, nearly $800 million above our prior highest U.S. commercial bookings quarter. We are seeing the immense demand of enterprises recognizing the need for sovereign AI to retain full control of their alpha. On the back of this exceptional continued strength in the U.S. and accelerating demand for our sovereign AI capabilities, we are raising our full year U.S. commercial revenue guidance to in excess of $3.424 billion, representing a growth rate of at least 134%. We're also raising our full year 2026 revenue guidance midpoint to $8.154 billion, representing 82% growth year-over-year and an 11-point increase over our full year 2026 revenue guidance from last quarter and our largest ever full year revenue guidance raise. Turning to our global top line results. Second quarter revenue grew 93% year-over-year and 19% sequentially to $1.935 billion. Second quarter U.S. revenue grew 115% year-over-year and 23% sequentially to $1.573 billion. Revenue from our largest customers continues to expand. Second quarter trailing 12-month revenue from our top 20 customers increased 67% year-over-year to $124 million per customer. Now moving to our Commercial segment. Second quarter commercial revenue grew 110% year-over-year and 22% sequentially to $945 million. We closed $2.337 billion in commercial total contract value bookings in the second quarter, representing 118% growth year-over-year. Our AI platform continues to dominate the U.S. market as the only real choice for operationalizing large language models, particularly as more customers demand full ownership over the data, logic, actions and security of their enterprise. Second quarter U.S. commercial revenue grew 149% year-over-year and 28% sequentially to $764 million. We closed a record-setting $2.132 billion of U.S. commercial total contract value bookings, representing growth of 153% year-over-year. Over the past 12 months, we closed $5.964 billion of U.S. commercial total contract value bookings, a 117% increase from the prior 12 months, highlighting the accelerating demand for AI that creates real operational value. Total remaining deal value in our U.S. commercial business grew 124% year-over-year and 27% sequentially. Our U.S. commercial customer count grew to 653 customers, reflecting growth of 35% year-over-year and 6% sequentially. Second quarter international commercial revenue grew 26% year-over-year and 2% sequentially to $182 million. Revenue from strategic commercial contracts was approximately $400,000 for the quarter, representing 0.02% of overall revenue. We continue to expect revenue from these contracts to be less than $500,000 in each remaining quarter of this year. Shifting to our Government segment. Second quarter government revenue grew 79% year-over-year and 15% sequentially to $990 million. Second quarter U.S. government revenue grew 90% year-over-year and 18% sequentially to $809 million. This growth was driven by continued execution in existing programs and new awards, reflecting growing demand for our AI platform in government. Second quarter international government revenue grew 42% year-over-year and 5% sequentially to $181 million. We closed $3.4 billion of total contract value bookings, up 49% year-over-year. On a dollar-weighted duration basis, total contract value bookings grew 129% year-over-year. Net dollar retention was 157%, an increase of 700 basis points from last quarter. We ended the second quarter with $13.1 billion in total remaining deal value, an increase of 83% year-over-year and 11% sequentially and $4.9 billion in remaining performance obligations, an increase of 103% year-over-year and 10% sequentially. As a reminder, RPO is primarily comprised of our commercial business as it does not take into account contracts with initial term of less than 12 months and contractual obligations that fall beyond termination for convenience clauses, both of which are common in most of our government business. Turning to margin and expense. Adjusted gross margin, which excludes stock-based compensation expense, was 86% for the quarter and reflects an increase in costs associated with taking on cloud hosting for one of our government customers. While this change led to higher cost of revenue in Q2, going forward, we believe it will power faster time to value, drive greater efficiency, provide greater cost certainty to the customer and enable us to expand their future workflows. Adjusted income from operations, which excludes stock-based compensation expense and related employer payroll taxes, was $1.194 billion in the second quarter, representing an adjusted operating margin of 62%. Q2 adjusted expense was $741 million, up 14% sequentially and 37% year-over-year, primarily driven by the continued investment in our AI platform and technical hiring. As in prior years, we expect a significant ramp in expense in the third quarter due to the seasonality of new hire starts and other product and marketing initiatives. We remain committed to investing in the most elite technical talent as well as R&D for our product pipeline and sovereign AI efforts, all delivering on our goals of sustained GAAP profitability. Second quarter GAAP operating income was $912 million, representing a 47% margin. Second quarter GAAP net income was $1.062 billion, representing a 55% margin. Second quarter stock-based compensation expense was $265 million and equity-related employer payroll tax expense was $17 million. Second quarter GAAP earnings per share was $0.41. Second quarter adjusted earnings per share was $0.41. Unrealized gains from our holdings in SpaceX resulted in a $0.03 tailwind to GAAP EPS and a $0.02 tailwind to adjusted EPS in the quarter. Additionally, our combined revenue growth and adjusted operating margin accelerated to 155% in the second quarter, a 10-point increase to our Rule of 40 score from the prior quarter and our 12th consecutive quarter of an expanding Rule of 40 score. Turning to our cash flow. In the second quarter, we generated $1.216 billion in cash from operations and $1.22 billion in adjusted free cash flow, representing margins of 63%. We ended the quarter with $9.2 billion in cash, cash equivalents and short-term U.S. treasury securities. Now turning to our outlook. For Q3 2026, we expect revenue of between $2.16 billion and $2.164 billion and adjusted income from operations of between $1.292 billion and $1.296 billion. For full year 2026, we are raising our revenue guidance to between $8.15 billion and $8.158 billion. We are raising our U.S. commercial revenue guidance to in excess of $3.424 billion, representing a growth rate of at least 134%. We are raising our adjusted income from operations guidance to between $4.889 billion and $4.897 billion. We are raising our adjusted free cash flow guidance to between $4.5 billion and $4.7 billion, and we continue to expect GAAP operating income and net income in each quarter of this year. With that, I'll turn it over to Alex for a few remarks, and then Ana will kick off the Q&A.
Obviously, we are loving these results and loving what they mean for our customers and, broadly speaking, the West. So reflections on how we got to 93% aggregate growth, just under 150% growth in U.S. commercial and an aggregate growth of 115% in America, which is astonishing, even surpassing the already anomalous results we've posted in the past and at a very significant scale. And the story really does begin at the beginning when we dedicated ourselves to our most important partners in the U.S. government, and we built products to deliver value for them. We delivered that value for them by looking at the world in its naked state. We did not have AI available. We had to work with NLP. So we had to develop forward deployed engineer models to extend the technology and deliver value. The nascent version of ontology was developed. Shyam, Aki, others strapped BlackBerries around their heads and made the code work in sensitive environments. And what we learned and what was built into this company is that there are things, values, structures that are more important than purely extracting value from a client. And we rejected the way in which we were being taught in Silicon Valley to build a software company at the time. Palantir, of course, is now both infrastructure software, forward deployed engineers, orchestration and business know-how. So it's a completely different hybrid. But at the time, we were being told our job was to trick the clients into giving us money for something that made them attached to us, but really added no value. So essentially a parasitic model. And in the rejection of that, we fully aligned with our partners. Now to do that, as we went through the years, we built Gotham, Foundry, Gaia, Maven, Ontology, AIP. Now we're taking the AIP stack and extending it for sovereign AI, which requires us to be able to orchestrate and fine-tune models to provide a completely sovereign stack to our partners. But what is the philosophical importance of that? We are offering a present that augurs to a future that we want to live in. What does that future look like? We have more rights in the productive growth frame. It's safer in the sovereign AI frame, which is arguably by far the most important because all these other things are downstream from GDP growth and GDP health and what if we transform America into the only democracy that actually grows where production is more efficient and manufacturing actually happens. What is that frame? That frame is not you are going to buy into a future where you have no job, where adversaries win and everybody who does win is a small tiny group of people living in a tiny place that somehow believes because they eat vegetables and they don't support war fighters that they deserve to have the total means of production of this country. And the rest of us should just sit back and absorb the cost of that revolution, which we're paying for. How are we paying for it? In the enterprise context, people sign up for token self-pleasuring and that at a real cost like other forms of self-pleasuring, where you are paying for the right for them to migrate your IP, your know-how, your expertise to their model so that they can build a competitive business that doesn't require your business, your people. And why are they doing it? It's actually being done for what they believe are moral reasons. They are superior to you. They deserve to colonize your enterprise. You deserve to be colonized. And Palantir, and it's interesting, it doesn't work as well or as efficient purely on the alpha side as having an application layer owning your compute. We built a partnership with NVIDIA. We're expanding our application layer. We are going to enter the market and already entering it in the classified space, as Shyam alluded to, of fine-tuning models. So the models actually fine-tuned by us in our enterprise on an NVIDIA stack outperform frontier models. And you own the weights, you own the alpha, you own everything. Every single enterprise in this country is going to either look at doing this, find ways of doing it or at least avoid the alternative of unprotected interaction with frontier models. This is very dangerous. You were warned about this in high school. And now you're learning about it in your enterprise. And in Palantir, we are on the front of driving this revolution. I am driving the business to grow at a rate equal or above to what we have in U.S. commercial for the next 18 months, which is a very high goal, but it is one we can actually get to because we are fully aligned with what's right and what's good and what actually works well in an enterprise. And for the first time, people believe us. And if you didn't believe us, you can believe 149% growth in the U.S., a Rule of 40 that's 155%, 93% aggregate growth and 90% growth in U.S. government with 62%, 63% free cash flow margins. People thought we wouldn't be profitable. So this is one of the more exciting times to be at Palantir. It's one of the more exciting times to participate in Palantir. And for everyone on the sidelines, you got to get off the sidelines. This is a revolution that will affect the sovereign revolution, where you stand in it will affect your livelihood, the livelihood of the people you care about and whether America and the West win. We cannot regress to a thin philosophical model where only a small group of people who think very differently than most of us actually absorb all of the revenue and value in our business and transfer all the dangers to us. And that's what this revolution is about, and it's extremely motivating for those of us at Palantir. Thank you.
Thanks, Alex. Our first question is from Dan Ives.
分析師問答
Where is the prompt? I'm good. Can you walk through—at the sovereignty boot camp, what did you learn from talking to customers that maybe you weren't expecting? I mean, obviously, just an overwhelming sort of group of executives. Can you just talk about that, Alex?
Well, first of all, for those of you not in the know, we did a sovereign boot camp after this kind of revolution exploded. And just to give the backdrop here, two years ago, we spent four or five months, Sasha, who runs this, organizing AIPCon. This was much more like, hey, let's invite our buddies over to lunch. And then all of a sudden, we start getting bombarded by people. People we invited, people we didn't invite at all levels of the business. So when you're working in enterprise, it's really important that the operational people are interested. So it was like CEOs, operational leads. And there's a huge educational component, and Shyam talks about this a lot. Likely people understand that they need a way of controlling their alpha. They understand broadly that token maxing is at their own cost, and they certainly understand that token maxing is leading to them transferring their data, their prompts, the way they run their business, their expertise to a third party. But they need education on what they can do about that with us preferably, but also without us. How do you do contracts? How do you work with open weight models? How do you work with closed weight models? How does this work in ontology? Is ontology the protective layer that they've been told? Does it create value the way they expect? How would this look in their own business? And how would they work with the compute stack? And so there's just this massive demand, and we're in the business of educating people, both our customers and others. But it was a super heterodox group of people, both in terms of the kind of people who showed up, the demand for it and also people we've not worked with. By the way, one of the reasons the net dollar retention number is so strong, ridiculously strong—I mean people always write these things about customer adoption—but the net dollar retention number is anomalously strong and is also going to shift, as hard as it is to believe, to become even more positive because some of our older partners we haven't really interacted with, they also showed up. They're like, 'Oh, okay, now we get why we would need you.' And not just Foundry—they're migrating across our stack. So customers that were only using Foundry now want Ontology, now want to be part of the sovereign AI stack. I would say, last not least, the internal version: recruiting, retention, excitement at Palantir. I mean, I'm very excited. I think the people around this table are very excited. The legal department is excited. I mean, that hasn't happened. That doesn't happen—it's more fun. So yes, that's the way this went down. And I think this is the beginning. So the way at least we think about this internally is what portion of the market is available to people who want to create value and keep it. That portion of the market has gone from a small portion of the market where we were doing well to like a large part of the U.S. GDP. And that's why we need partners. Behind the scenes now, we're trying to find partners. Now partners, we need technically exceedingly competent partners. Partners doesn't mean a vassal. We don't have to agree on every issue. We don't even have to agree on every client. They can occasionally compete against us. We've seen this in the defense tech stuff. The approach to defense tech is where we partner with people; partner doesn't mean we agree. Sometimes we compete, but it means we're marching in a similar direction that allows us to scale. So those things are going to be a very important part of the guide. Why am I pushing the company to grow not just to the end of the year, but next year? It's because that also forces us to find ways to scale to meet the demand that's out there.
Thanks Alex. Our next question is from Mariana with Bank of America.
So a follow-up to Dan's question. When all this AI revolution started, it was really clear for enterprises that data and proprietary data and how you train your models was going to be the key. But then we're like three years into that revolution and now everyone started to realize owning my data and learning from my data is important. What happened there? Why do you think that you positioned yourself back then with a different approach to AI that enterprises weren't able to see? And why do these numbers reflect that you're the winner of AI today? It wasn't that clear for any other software application back then.
Well, you can kind of divide it up into two parts. There's the first part which is efficacy, and then the second part is scale. In the zero-to-one phase, it's much more important to focus on the application layer: how do you turn that new thing into economic value. And then as people started to experience the economic value and as time passed, you started to see that some of the people who were partners out there were building things that are competitive to you. I think that took some time to seep into the psyche and mindset of enterprises: wait a second, this is valuable in the right hands on the right platform, but I also need to control the weights, and the alpha that is being generated isn't simply the data that's resident in my enterprise. It's also the metadata, the reasoning traces, the exhaust, the usage of this, which I don't yet have mechanisms to control. And now I understand that's actually probably more valuable than just the data in my enterprise. That's been a clarion call for the market over the last quarter or two quarters.
There's also the question implicit in your question of why did we get this right? And again, I think it's because we are fully aligned with our partners. Sometimes we make decisions that are against our short-term economic interest, like supporting lots of institutions in Europe where growth is slower. Without our products they would have much worse security and governance outcomes. It's not actually in our economic interest anymore to do this, but we still do it. It's because we actually are believers for better or worse. I would say also, this is a company that from inception has valued artistic insights, meaning you can't model something purely on science. You have to have an aesthetic or artistic appreciation for it. We have made huge bets. Everyone sitting at this table and many, many hundreds of people at Palantir have essentially artistic insights. We've always said we're a colony of artists and people assume that means we're just difficult. And that's also true. But we value insights that are way before anyone else would see them. We build major parts of our business on those insights. That's very hard for normal businesses to do because one of the jokes running around Palantir is we can definitely meet our guide next year if we get paid for all the people copying us. So we got a small portion of people who copy the fundamentals, we do very well. Normal businesses are built around a playbook that they execute. We're in a non-playbook world. Executing on a playbook that worked five or ten years ago—essentially build parasitic software and monetize it—doesn't work now. There's hundreds of variants of that, but that is a central advantage we have. We are a colony of believers and artists that are very motivated to drive value. That sets us apart much more than I would have imagined ten years ago. And then luckily for us, this is capitalism: you have to look at the results. If our results—and the other thing I'll tell you that's very special for us—we are outsiders. Outsiders have to have really good results. The same is true for Palantir. People aren't buying our product because of networking or dinners. That outsider status has caused huge problems in the first 18 years, but many benefits in the next 18 years. Other people don't like being outsiders. In fact, I'm struggling with our current popularity.
Thank you. Our next question is from Gil with D.A. Davidson.
The topic of sovereignty: you focused a lot on how dangerous it is to give the keys to the labs because they could choose to compete with you. Is there another aspect of this as well, though, that if you choose a lab and you buy the orchestration and the consulting and the harnesses from them, you're beholden to their models? And if something happens—if it's not the best model anymore, if the model gets pulled—then you as a customer are stuck and you may have a mission-critical system fail. This isn't hypothetical. This happened a couple of times this year where one of the frontier models got pulled by either the government or the company. Well, if you work with Palantir, I would assume that when that happens, you can go to your customer and say, 'Hey, if that model doesn't work, I can plug in another model for you?'
No. I'll let Ryan and Shyam comment here, but we're already doing that across the U.S. government. We have a product that allows you to switch out models. Look, at the end of the day, if you are locked into a product, that's the nature of monopoly capitalism—people want to lock in because then they can raise prices and reduce quality. We're against that because we're on the side of the American worker, the American people and its great institutions and other institutions across the West. People are running big enterprises in this country; they're very sophisticated and aware of these risks. They don't like arrangements that feel like they're being taken advantage of. There's a lot of concern. I spend time explaining that some of the people involved in these things are not the caricatures people assume they are. But because the business setup looks like 'heads I win, tails I win,' American business people don't like that. There are a lot of contractual mechanisms that support switching and sovereignty.
Yes. I would say our whole focus is converting tokens to value. The example Shyam gave—that's happening across the board in conversations with customers. That's why they're looking to expand with us and convert how they position themselves in their industry. We're seeing extreme alignment. It's not about being beholden to one model. It's about bringing the right models to bear for the right purposes, and our contracts and structures are set up to do that to support and compound our customers' alpha in the organization.
I mentioned it earlier, but we've been beholden to a small number of benchmarks that people have been designing models to and then releasing models saying, 'Look how well it does on this benchmark.' But the benchmark has almost nothing to do with your business. So how do you figure out how to make the benchmark that represents your reality—what you're trying to succeed at and what you're trying to get better at—and then see what model makes sense? The natural consequence of doing that, even leaving aside all of your other arguments for sovereignty, is: how do I climb that hill? How do I figure out what it is that I do as a business that I feed back into weights that I can control, which presupposes an open model and sovereignty? Then you're not just going to wait and switch when a model gets pulled; you're going to leverage automation to constantly figure out when you have a next checkpoint and whether there's a new model to adopt. I mentioned the example: I literally almost felt gaslit when within 24 hours of getting Nemotron up with no post-training—vanilla Nemotron Ultra—it did better than the frontier on some production tasks. If you only looked at standard benchmark numbers, that shouldn't even be possible. But, of course, the benchmarks are wrong for what matters to a given business. They are right for what they measure, but that's not the tasks our customers had. Moving this to an empirical basis is how we're going to accelerate the realization of tokens to real economic value.
Thank you. Alex, as always, we have a lot of individual investors on the line. Is there anything you'd like to say before we end?
Well, your support was crucial to getting us this far and crucial to getting us to where we're going to go, which is a much, much, much larger company. This is one of the most exciting times to be involved in the Palantir mission. We are going to help transform especially this country, but allied countries, both in commercial and government. The sovereign frame we're using as our organizational principle is one that is inclusive of everybody who wants to have a better world today and tomorrow. We invite everyone to engage with it in some form. Thank you.
Thank you. That concludes Q&A for today's call.