Alex Karp unleashes on OpenAI and Anthropic as AI industry faces growing enterprise backlash

Palantir's Karp bashes OpenAI, Anthropic token model: 'Something has gone completely wrong'

Palantir CEO Alex Karp says 'something has gone completely wrong' with how AI is sold Palantir CEO Alex Karp on Wednesday criticized the token model used by U.S. artificial intelligence labs Anthropic and OpenAI as costs skyrocket. "I'm not throwing shade at them, but something has gone completely wrong," he told CNBC's "Squawk Box." "The basic view among enterprises in this country is I'm going to chillax and waste my time with tokens." As AI costs surge, and new models prove pricier than previous iterations, enterprises are shifting from a mindset of so-called "tokenmaxxing" in favor of a return on investment. That setup is prompting some enterprises to adopt open weight models, capable of performing similar tasks at a fraction of the price. Chinese models are also accelerating capabilities, raising concerns that the AI rival could soon catch up to U.S. frontier labs. Karp told CNBC that the industry should not underestimate the speed at which China is making progress in building AI models. In this environment, many businesses are also shifting from using far-reaching AI models to building and training their own, more efficient proprietary tools. Earlier this week, Palantir announced an expanded partnership with Nvidia to use the chipmaking giant's AI tools to build custom models for U.S. government agencies. Karp views open weight models as a potential solution for CEOs frustrated by AI labs. "What is happening among the most technical players is they're saying, 'I want something I own. This is my business," he said. -- CNBC's Seema Mody contributed to this story.

Opinion | Maybe Anthropic and OpenAI Are Not the Future of Artificial Intelligence

You're reading the David Wallace-Wells newsletter. The best-selling science writer and essayist explores climate change, technology, the future of the planet and how we live on it. Get it sent to your inbox. Last week, the Palantir chief executive Alex Karp made one of his more remarkable television appearances in what is quickly becoming a notorious run of televised rants. "Something has gone completely wrong," he declared on CNBC, in an appearance so vivid and spastic it was widely described online as a "crash out." He was referring to the whole structure of the A.I. industry, which had been built on top of a value proposition that looked to him like a dead end. The big labs, such as Anthropic and OpenAI, have been overhyping their own closed-source models, he argued, hoarding their value rather than empowering their clients and partners with them. More than that, he seemed to say the labs were exploiting those clients and partners -- private companies and individuals but also militaries and intelligence agencies -- by making use of their research and intellectual property. Open-source or open-weight alternatives, which allow considerably more in-house customization and control, were obviously preferable, he suggested, for almost all users. "The jig is up," he announced. Karp is an irresistibly noxious figure, a would-be philosopher king who can't seem to sit still on a TV set or a conference stage as he holds forth about kill chains or tai chi or the woke left. Perhaps even more than Elon Musk, Karp seems to understand that in an attention-economy economy, the job of an executive is to always be pitching. He is also perhaps the most visible spokesman of Silicon Valley's increasingly significant defense tech sector; Anduril, Palmer Luckey's drone manufacturer, is another of its tent poles. For the last few years, a new Silicon Valley has taken shape in a kind of cluster around defense tech, robotics and artificial intelligence, with the leaders of each sector speaking almost in unison to describe the urgency of an American industrial renaissance and present technological competition with China in terms of a new Cold War. In this view, which has been mostly embraced by the Trump administration, nothing should be allowed to get in the way of technological progress, and the march of that progress could be measured crudely in capital expenditure. This is one reason it was so striking for Karp to be yelling that A.I. was heading in the wrong direction -- a presumptive ally openly bashing the big A.I. labs and the business proposition they represent. Karp had been softly floating his critique for some time, but the CNBC event looked like a proper coming out. Just one day earlier Palantir had published a kind of manifesto devoted to what it described as the all-important principle of "A.I. sovereignty." The central argument: Companies should seek to build their own A.I. tools, not just customize those on offer from the frontier labs. This might mean relying on open-source L.L.M.s rather than the proprietary ones on which the A.I. boom has mostly been built in America, but it would amount to a liberating declaration of independence from Big A.I., which in Karp's estimation was sucking up much more value than it was generating. Karp isn't exactly a disinterested observer here. In recent weeks, out of a mix of concerns about political vulnerabilities, national sovereignty and privacy, France has announced that its intelligence service is cutting ties with Palantir. The future of the firm's partnership with Britain's National Health Service also seems to be in jeopardy. Karp was on TV to promote a new partnership with Nvidia that would allow Palantir to develop and sell a distinct set of products to compete with those on offer from the frontier labs -- which is to say, in railing against the Big A.I. business model, he was undeniably talking his own book. But his rant highlighted several genuine and growing questions about the way those big A.I. labs have sold the rest of us on their paths to global domination. He is not alone in asking them. And he is not wrong to. Even if plenty of indicators suggest that the A.I. economy as a whole is rapidly growing, it looks less inevitable than it used to seem that the big, brand-name labs will be the titanic profit engines of that future. Perhaps they will function more like utilities, providing machine intelligence almost like electricity to other innovators, entrepreneurs and enterprises. As I discussed with Natasha Sarin last month, the hype cycle of the last few years has flourished -- and generated a lot of investment -- on the supposition that artificial intelligence was the kind of arms race it was possible to outright win, and that enormous or even monopolistic profits would flow inevitably to the winner. The basic idea was that at a certain point, competition would somewhat naturally come to an end, when the technology would grow so powerful that it could quickly and dramatically engineer its own successor models, producing an exponential liftoff leading quite quickly to what is often called "artificial superintelligence." Beyond that threshold, the leading L.L.M.s would be so powerful, and would be improving so rapidly, that even small initial advantages would compound quickly into something like a natural monopoly on intelligence, which could then be sold to users at almost any price. These days, as A.I. boosters have cooled their talk of a jobs apocalypse, you also hear a little less about artificial superintelligence, now typically short-handed as "A.S.I." But the ongoing A.I. investment cycle is still built on the same underlying paradigm: that historic levels of capital expenditure are justified because the returns from winning the race would be unthinkably enormous. But can the race even be won? Can any lab open up an enduring advantage over the others, let alone one sufficient to justify a monopolistic claim on A.I. revenue? Over the last year or so, this logic has come to seem a lot more questionable, in part because, though progress has continued, no model has retained a long-lasting advantage, and plenty of those cheaper, open-source alternatives have kept a pretty close pace with the best-in-class versions. When A.I. companies began raising prices on their premium products to more closely match the cost of producing them, many of their clients balked, realizing that frontier models were not generating enough profit to justify the expense. Partly as a result, corporate uptake of frontier models flatlined; much cheaper, open-source models exploded. This helped illustrate a broader pattern, visible at least since the explosive release of a cheap, open-source model from China's DeepSeek in 2025: that even if copycats never quite caught up to the best-in-class standards, they'd also never fall so far behind. The frontier labs have grown so worried about this that they are desperately appealing to Congress to take legislative action against what they say are lesser companies, many of them foreign, effectively stealing their I.P. through a process known as "distillation." This looks almost like an existential threat because for most users -- even most corporations with capital to burn -- it might not make intuitive sense to pay a superpremium for an only slightly better product. For many, it would probably make more sense to pay a lot less for a slightly inferior product -- especially given that, because of the rate of progress, no model is likely to retain its advantage for very long. This is one reason a growing number of A.I. watchers have begun emphasizing that however impressive the models were, the ultimate impact of A.I. will be determined as much by what is sometimes called "diffusion": how quickly, widely and capably those tools will be embedded in a broader social and economic ecosystem still directed by humans and full of many human bottlenecks. If that alternative perspective is right, it will make the leading A.I. labs considerably less central to the A.I. future than they have seemed for so long. A draft internal analysis prepared by Treasury Department analysts has reportedly warned that the size of the big A.I. companies represents a systemic risk to the country's economy and financial system, though higher-ups have publicly criticized the report. A few years ago, it was fashionable to say, as Peter Thiel liked to, that cryptocurrency was a libertarian technology, enabling individuals to conduct even large-scale financial business outside the reach or oversight of any government, and that A.I. was a "communist" technology, by which he meant authoritarian, concentrating and centralizing godlike planning powers into a single machine intelligence whose judgment would presumably supersede even that of the market. But as we move further into that A.I. future, it no longer looks so clear that we are heading toward convergence like we used to read about in science fiction. Instead, what we have is a more unsettled landscape, which some have called decentralized and democratic and others simply more competitive. The meaning of this technology is not limited to its market impact, of course, and the trajectory could change again. But that is just another reminder of how early in this story we are -- that such fundamental propositions about the shape of what's to come might change so profoundly in the space of just a year or two.

Palantir CEO Calls Drugs 'The One Thing I Don't Do' in Wildly Unhinged Interview

Palantir CEO Alex Karp gave an extremely high-energy interview to CNBC's "Squawk Box" on Wednesday, with his arms flailing as he ranted about everything from politics to Israel to AI. He also referred to past public appearances where people have speculated that he was on drugs. Karp started the interview by criticizing the frontier AI companies like Anthropic and their fixation on customers paying for individual tokens, something that the Palantir CEO says frustrates the leaders of other companies who are being sold on the productivity gains of artificial intelligence. "I'm not throwing shade at them, but something has gone completely wrong," Karp said of the AI companies. "And the basic view among enterprises, in this country, is I'm going to chillax and waste my time with tokens. I'm going to get no value, and they're going to get my IP." No, we're not sure how that's the fault of AI companies, either. As we said, Karp had a lot of energy this morning. Karp also raised concerns about the security of sensitive intellectual property and classified information. One of the weirdest moments came when Karp started referring to himself and making nods to old interviews where he was accused of appearing to be on drugs. Karp said that CEOs of other companies didn't want to talk publicly about their frustrations with the AI companies, so he apparently needed to be the one to do it on TV. "Every single enterprise in this country, in private, a lot of them don't want to speak in public because it gets outsourced to the neurodivergent crazy person that apparently is on drugs, the one thing I don't do," said Karp. It's unclear what Karp is implying he does do, if not drugs. Other reporting has tried to claim that Karp may have appeared high because he had sugary drinks. After one infamous appearance at the New York Times Dealbook Summit in late 2025, many people thought Karp looked like he was on cocaine. Semafor tried to claim that it was because Karp had broken his own rule against consuming sugary drinks, insisting it must be because he drank a Mexican Coca-Cola before his interview. "Okay, so that's... that's my role," said Karp this morning, sort of getting back to the topic of AI. "But I'm telling you, in this country, at every single enterprise I deal with, these people are livid. They're like, 'I am paying for tokens that create no value.'" Karp also seemed to rail against the proposals for a kind of tax on AI companies that would help fund jobs programs or pay a dividend directly to citizens, something that Anthropic has floated as an idea if unemployment rises to 25%. "These people are stealing the weights and alpha of my business, and they're creating a wealth tax that does not help the poor. It just punishes... starts with the billionaires... every single person at this table is going to be paying a wealth tax only to punish us," said Karp. When "Squawk Box" host Becky Quick said that Karp sounded "pretty angry," the CEO shot back that "this is the voice of American business that is being channeled through me." Karp said that the reason this tax might come about is that the capabilities of the AI models have been oversold. He also talked about politics, as he often does, complaining that he's been "kicked out" of the Democratic Party for warning against what he calls the "far left" and their worldview. "You would think that the greatest problem in the world, if you were on the far left, is somehow it's like, if things work, it's evil and bad," said Karp. The CEO also started to criticize the right before making sure to say there were problems on both sides politically. "Then you have on the far right, like, it's like on both sides, you think the biggest problem of our society are like warlocks roaming the street, building technology. Of course, they can't understand it. Like, they attack Palantir for the craziest..." Karp said, trailing off. "China doesn't have that problem." Karp had a lot of unfinished sentences on Wednesday, including when he talked about what he believes are the three biggest tech countries in the world: The U.S., China, and Israel. The CEO listed other countries in Europe, including Sweden. "And Sweden has a real tech scene. You have to be... you can't just like... you know..." Well, no, we don't know what he was going to say. Because he didn't finish the sentence. Sweden has a small but influential tech scene as the home of Spotify, but it's not clear what Karp was going to elaborate on when it comes to the Nordic country's influence or barriers. We have established that China doesn't have a warlock problem; maybe he was going to address the situation in Sweden as well. Karp was also asked about his support for Israel, which elicited an initially stammering response. He called himself the "most publicly supportive CEO of Israel," and insisted that the country "is on the side of good." But he also called himself the "most effective critic" of Israel because he said he was "fair" to the country. Karp has not previously aired any criticisms of Israel, at least not publicly, and has been known to ridicule pro-Palestine protesters who accuse him of enabling genocide in Gaza. When CNBC host Andrew Ross Sorkin pressed Karp on the memorandum of understanding between the U.S. and Iran, along with Israel's distaste for it. "Look, Israel, I'm going to leave the Israeli thing for my talking to them in private because I just don't think at this point, it's very hard in public to talk about them because you just have people..." Karp trailed off again before saying there are "legitimate criticisms" of Israel, but many people don't think it should exist. "Do I think Iran's been degraded? Yes. Are there things that I don't think are in the public space that would be comforting people if they are? Yes. And I'll leave it at that," said Karp. Sorkin ended the interview, and Karp squirmed in his chair, appearing eager to get out of it. Sorkin seemed to tease him about that, implying he didn't need to leave. Karp then said, "I feel like I'm going to be kicked out of the room," before the CNBC hosts all comforted him and told him how much they appreciate his time -- no doubt because, if nothing else, Karp's antics certainly make for entertaining TV. And in the world of broadcast media, entertainment is everything. It helps explain why a guy like Jim Cramer can still be employed by CNBC after all these years. You don't need to be right or good or intelligent. All you need is to be entertaining for the CNBC audience. Karp made that subtext more explicit when he kept talking about why he liked the morning show, saying that people on Squawk Box had divergent opinions, which was "fun." He said other shows were "boring," and at one point, he asked if they were off the air, clearly hoping to say something more honest. Karp was warned by more than one host that they were actually still broadcasting as they tried to transition to another segment, but he just kept talking. The voice of American business, folks. It never stops. Especially after one too many Mexican Coca-Colas.

Karp unleashes on frontier AI labs

The big picture: Karp put his finger on a trend: U.S. companies are increasingly turning to cheaper Chinese AI models just as the Trump administration is blocking access to some of the best American AI tools -- a double whammy for domestic AI labs. What they're saying: Karp said CEOs are telling him privately that they're getting "no value" out of enterprise AI tools, which they're "tired of." He added that AI labs are charging too much. * The cost problem has driven a slew of companies to start using Chinese AI models to cut their IT bills. * Microsoft is weighing using the Chinese AI model DeepSeek. Coinbase kept its AI costs flat by using Chinese open-weight models. And U.S. startup Cursor built its latest model on top of Kimi 2.5, a model from Moonshot AI, backed by Chinese e-commerce giant Alibaba. * Chinese AI model usage is skyrocketing, per OpenRouter, as enterprises look for ways to lower their AI bills. Friction point: The switch isn't just about saving money. Karp said the models were "irresponsibly" described as "dangerous for everyone" without clarity on how enterprise IP will be protected if that's the case. * The critique comes amid rising regulatory scrutiny, with the White House asking both Anthropic and OpenAI to limit or delay the release of their most powerful models. * Developers are blaming Anthropic CEO Dario Amodei for Washington's response, saying he was too vocal about the potential risks of his company's technology. * One developer told Axios they are using Anthropic's tools less, in part due to Amodei's warnings. Between the lines: Whether it's about money or regulation, the net result is the same: U.S. AI labs are receiving pushback from some of their biggest customers. Yes, but: Karp also called Amodei a "historic figure" and said he's never seen anything like Anthropic's success as one of the fastest-growing companies in American history. * The AI race has also been defined by constant leapfrogging, so it's hard to imagine any one lab or set of labs being the underdogs for too long before rebounding. * Karp, for his part, is both competing and cooperating with the frontier AI labs he's criticizing. Palantir doesn't train frontier models, but it increasingly competes with OpenAI and Anthropic to become the platform through which enterprises and governments deploy AI. The bottom line: After years of hype and seeming invincibility, U.S. AI labs are no longer viewed as untouchable.

'Something has gone completely wrong': Palantir CEO Alex Karp slams OpenAI, says AI industry is "effing insane"

Palantir CEO Alex Karp has hit out at the state of the AI industry, saying it was "effing insane" that the technology is being used in areas such as military and national security. In a heated interview with CNBC Squawk Box, the controversial billionaire also hit out at top AI firms such as OpenAI, claiming he had spoken to major CEOs outside of the industry who were "livid" at how some companies are doing business. Karp also accused some major AI companies of imposing a "wealth tax" on businesses by charging high fees for their services, all while collecting data which may be used to improve their own AI models and tools. "Completely wrong" Karp's ire was particularly focused on the token model being used by the likes of Anthropic and OpenAI, especially as costs continue to rise, but companies look for a better return on their investment. "I'm not throwing shade at them, but something has gone completely wrong," he said. "The basic view among enterprises in this country is I'm going to chillax and waste my time with tokens." This includes a range of Chinese firms, with Karp warning the US not to underestimate the speed of progress being seen at its great rival. Rising AI prices have led many businesses to pivot towards building and training their own models, rather than relying on outside providers, with so-called "open weight" models able to perform at a fraction of the cost. Karp's frustration was clearly visible, with one CNBC host commenting, "You sound pretty angry," with the CEO responding, "This is the voice of American business that is being channeled through me." To shore up its own support, Palantir recently announced a major partnership with Nvidia which will see the latter's AI services used to create custom models for US government agencies. "What aligns me with Nvidia, and I think is what the technical customers want, which is control over their compute, their models, their data stack and their alpha," Karp told CNBC. "They want to know they own the means of production. It's not being transferred to someone else." This follows recent criticism by the US government of firms such as Anthropic, whose Mythos 5 and Fable 5 AI models were deemed a national security risk and shut down shortly after release. Karp went on to criticize the US government for its reliance on AI companies in creating new technology for the military and national security. "Are we really going to outsource the battlefield of this country to the consensus view in Silicon Valley? That is effing insane," he noted. Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds.

"Televised Nervous Breakdown": CEO of Palantir Suffers a Bit of a Meltdown During Live Interview

Can't-miss innovations from the bleeding edge of science and tech Everyone has had a bad day at work, but most of us are lucky enough that ours weren't broadcast with a chyron live on television. On a live interview with CNBC's infamously churlish segment "Squawk Box," Palantir CEO Alex Karp appeared to suffer a nearly 20-minute meltdown, complete with stuttering, nervous backtracking, and a steady supply of digressions so abstruse that the hosts seemed befuddled and perhaps even concerned for his wellbeing. Though Karp was called up to chat about an ongoing deal between Palantir and the chip maker Nvidia to build AI infrastructure for the US government, he quickly went off the rails, using up minutes of airtime to complain about the financial bubble undergirding the AI boom. While there may be a point buried in Karp's diatribe, it quickly became lost in a wash of unintelligible jargon. "These models have been completely over, irresponsibly over-sale," Karp ranted at one point, "and the sale is, 'it's dangerous for everyone, which is why I can give [AI] to all your adversaries but I can't give it to the Department of War, or I can't safely give it to an enterprise in this country, without being certain that the Alpha of that business could transfer to this model tomorrow, ie I have no business, no job.'" "You sound pretty angry," CNBC's Becky Quick interjected after a nearly three minute-long rant from Karp. "No," the CEO snapped. "This is the voice of American business that is being channeled through me!" Even Karp's more intelligible arguments are quickly trampled over as additional intrusive thoughts took the wheel. At multiple points, Karp got hung up on the idea that elite universities might not welcome him as a professor anytime soon -- an aspiration his parents still have for him, apparently. "American enterprises are run by the shrewdest, most widely intelligent people on the planet," the Palantir CEO started to say, setting up an argument that companies aren't interested in foundation models, but in AI apps that can actually solve problems. That train of thought quickly leaves the station, though, as he pivots to his higher ed ambitions literally mid-sentence. "If you think they're going for that [foundation models], you can go try to sell me -- like my, my parents still want me to get a job as a faculty member at Berkeley," he complained. "Go try to get me a job at Berkeley. It's not happening." By the time the lengthy "interview" -- it's really more of a lecture, since every time one of the hosts tries to get it back on track, Karp launches into a new stream-of-consciousness tirade -- comes to an end, Karp jokes that he feels "like I'm gonna be kicked out of the room." To Karp's credit, his interviewers struck an ameliorative tone. "Never, a wide-ranging conversation, really appreciate your time," one of CNBC's professional journalists -- the camera was pointed elsewhere -- replies. Unfortunately, that prompted Karp to dig in even more, starting off on another winding digression as the CNBC chyron cut to a live shot of Donald Trump's new Air Force One aircraft. "I get kicked out of these rooms -- even if I agree with you I would try to disagree with you, it's more fun," the Palantir CEO blathers as the Squawk Box interviewers try to wrap it up. "Alex thank you, we appreciate it very, very much, thanks" CNBC's Andrew Sorkin says, clearly cueing Karp to leave so they can move on. "And I'll tell you -- we're off camera now?" Karp continues. The hosts reply in a chorus: "no, we're still going."

When the sovereign AI diagnosis goes prime time

Palantir Technologies Inc. Chief Executive Alex Karp went on CNBC this week and delivered what one outlet generously called a "televised nervous breakdown." He called the artificial intelligence industry "effing insane." Karp (pictured) also accused OpenAI Group PBC and Anthropic PBC of running a "wealth tax on American business." When host Becky Quick noted he sounded "pretty angry," he clarified that he was merely "the voice of American business being channeled through me." Great theater. But turn down the volume and Karp said the quiet part of the entire sovereign AI thesis out loud. Here's what actually happened underneath the shouting. Palantir and Nvidia Corp. shipped a Sovereign AI OS reference architecture -- a turnkey stack you run on your own infrastructure -- and just extended it to deploy Nvidia's open-weight Nemotron models in secure, air-gapped environments. Customers keep their data, their models, their weights. The stock jumped 9%. Turns out the market likes "own your stack" more than it likes a meltdown. Now read Karp's own words: Customers want "control over their compute, their models, their data stack and their alpha... they own the means of production. It's not being transferred to someone else." That's not a rant. That's the five pillars of sovereignty, delivered at a higher decibel: * Territorial: Where it physically runs. Table stakes, not the answer. * Operational: You hold the keys, you get paged at 3 a.m. * Technological: Own the stack and the IP. Don't license a black box you can't fork. * Legal: Jurisdiction follows the vendor, not the data center. * Financial: The token meter is the new lock-in. Karp bashing usage-based pricing as "completely wrong" is the financial sovereignty argument, verbatim. Ask Uber Technologies Inc., which torched its annual AI budget in four months. And notice who's arriving at the same conclusion from every direction. The EU stood up a Digital Sovereignty Task Force and adopted a formal sovereignty declaration. Mistral raised $830 million -- zero U.S. banks -- for a graphics processing unit data center outside Paris. Add HUMAIN, G42, India, Canada. Everyone, all at once, is trying to get out of the business of renting intelligence from four U.S. labs on a meter they don't control. Karp's delivery was unhinged. His diagnosis wasn't. When the CEO of the most aggressive enterprise AI company on earth starts quoting the sovereignty playbook on live TV, that's not a breakdown. It's the second wave arriving -- the one about control, not capability. The sovereignty wars have begun. This week, they went prime time. Amit Eyal Govrin is co-founder and CEO of Agentcy Labs. He spoke recently with John Furrier, co-host of theCUBE, SiliconANGLE's studio, about why sovereign AI is often misunderstood.

Tokenomics - the worldview according to Palantir as CEO Alex Karp lays into OpenAI and Anthropic's "effing insane" business models

Something has gone completely wrong! As the debate around tokenomics rumbles on, in a no holds barred diatribe, Palantir CEO Alex Karp has laid into the "effing insane" business model of the likes of Anthropic and OpenAI, which he accuses of ripping off American enterprises - and they're mad as hell and not about to take it anymore. In a televised interview with CNBC, intended to publicise a new tie-up around sovereign AI capabilities in partnership with NVIDIA, Karp said: Our clients just say they're unhappy with the frontier labs...I'm telling you, in this country every single enterprise I deal with, these people are livid. They're like, 'I am paying for tokens that create no value. These people are stealing the weights and alpha of my business, and they're creating a wealth tax that does not help the poor, it just punishes'. Every single person is going to be paying a wealth tax, only to punish us, and the reason for that is because these models have been completely and irresponsibly over-sold. He went on: It's like there's just a level of dis-comfort and loss of trust...We need to re-build trust, and that trust is going to happen where everyone gets to ask and answer basic questions - who owns the data, where is it cached, are the prompts secure? If the business models of the Large Language Model (LLM) providers was really all that was claimed for it, then, he asked rhetorically: If it was so valuable and let's say I can make you a billion dollars, then tomorrow, wouldn't I say, 'I'll make you a billion dollars, and I want 30%'. Why are they charging for tokens if it's so valuable? Enterprises need to ask what the true cost of this is, he urged: [It's] not just what you're paying. The true cost is not just is what you make minus what you lose, meaning the value of your business. But on the face of it enterprises have been sold the idea that there is no alternative to the current approach: The basic view among enterprises in this country is I'm going to chillax and waste my time with tokens, I'm going to get no value, and they're going to get my IP. While enterprises may not be speaking out themselves, they are unhappy and their complaints get "outsourced" to him, with Karp claiming that companies prefer them to come from the "neurodivergent crazy person that apparently is on drugs, the one thing I don't do". In fact, he insisted: This is the voice of American business that is being channeled through me, and I'm telling you, it is absolutely a problem for this country, because we are on the cutting edge of every single AI technology. But if you're going to triply over-sell something, enterprises are just tired of it. So what alternative does Karp propose, other than, inevitably, buy from Palantir? He argued: What technical customers want is control over their compute, their models, their data stack, and their alpha. They want to know they own the means of production, that it's not being transferred to someone else. They're not interested in some fake deploy that somehow is deploying tokens that transfers the alpha to a third party. The jig is up. He pitched: When you're using Large Language Models, everyone technical realizes they're like a critical resource to make them valuable in an enterprise...You have to have what's called an application layer. We have this thing called ontology that now everyone's copying, but de facto it takes a Large Language Model, it makes it safe and useful and precise. [It's] safe because it doesn't touch your unlearned data, safe because it prevents the Large Language Model from caching your data and replicating your business, safe because it doesn't transfer your IP. The solution as proposed by Palantir comes from a triumvirate of elements - the model plus an application layer plus compute, he said: In our jargon, it's the value, the reason...The reality of compute plus ontology plus model change is changing the course of history. Ask the Ukrainians, ask the Israelis, ask our Department of War, ask the enterprises that are working with us. We do not have to over-sell what we have. We do not have to over-hype. This bravado has been put to the test in the real world, he claimed: Enterprises in this country trust and love us, especially ones that are involved in critical infrastructure, both public and private...I am slightly frustrated when they, Open AI or Anthropic especially, are talking about critical infrastructure...The reality is critical infrastructure does not run these models without an application layer in almost all cases. That application layer is our ontology. That's a fact, and there's a reason for it. Some people love me, some people hate me, they're not buying, they're not buying it because of that; they're buying because these things have to be made safe...The whole secret of Palantir is the forward deployed model, the products that have been five years ahead. Everyone, everyone, everyone said forward deployed engineering was for services. They said we don't even know what an ontology is. That's the only thing people talk about nowadays. When it comes to AI models, Palantir is completely agnostic, he argued: We like that world. Why do we like that world? The same reason as when you go online and there's 50 kinds of coats, you get to pick the one you like. But what is happening among the most technical players is they're saying, I want something I own. This is my business, I want to own the GPUs, I want to own my data, I want to own the model, I want to control the alpha. And for avoidance of doubt, Karp wanted it to be known that this isn't just sour grapes to cast shade on the LLM providers, stating that Palantir wants the best for everyone: We're coming up to July 4. I want for us and our friends across the globe to have the very best tech resources...We have to find ways to make these models raise the standard of living for every American. My take I'm going to try to keep this more adult than I usually do. Did Karp succeed in that ambition? Well, judge for yourselves. His 20 minute near-monologue clearly came from the heart, but then no-one, least of all me, would accuse Karp of being anything other than someone who says what he feels. There's no soft soap on show here, even if what he says is all-too-often deeply unpalatable to a lot of people. His near rant on the LLM leaders business models and behaviors may not have been entirely coherent as a linear argument, but there were points where, heaven help me, I did find myself nodding in agreement - the notion that the frontier model layer is in practical terms essentially at risk of becoming a de facto tax on enterprise private knowledge, for example. And his argument of the need to re-build trust, even at this early stage of the AI hype cycle, resonates, even if my overall reaction remains that if Palantir is the answer, then we're absolutely asking the wrong question. And I'm not at all sold on Karp's seeming love of open source which was a spin picked up by a lot of people online following his TV appearance. But then I'm no keener on the idea of OpenAI or Anthropic helping themselves to the contents of the enterprise 'brain' either and setting themselves up as latter day competitors to those who are currently their customers. As David Sacks, tech sector veteran and Trump 2.0's main AI policy advisor, put it on X later: Real AI safety for businesses is the ability to control their own data, model weights, and compute, so a frontier lab can't hoover up their proprietary knowledge and turn it into their next product. As Karp explains, technical customers want 'control over their compute, their models, their data stack, and their alpha. They want to know they own the means of production, and it's not being transferred to someone else.' Don't think that can happen? Just look at Figma. According to The Information, Anthropic 'blindsided' its then-business partner with the launch of Claude Design. Figma's founder said Anthropic had not been 'consistently honest' with them. Anthropic's chief product officer had even served on Figma's board until three days before the launch of Claude Design. Figma's stock has fallen sharply this year while Anthropic's valuation has surged. This all comes back to the current hot topic of enterprise disappointment at the poor visibility of value when it comes to early AI investment. That's something that needs to be tackled by all vendors with a vested interest in making sure that the AI bubble doesn't burst. All told, a colorful and typically controversial contribution to the ongoing tokenomics debate. Not one that settle the matter by any manner of means, but certainly one that will attract a lot of interest and prompt further debate.

Alex Karp, frontier models and the real fight for Enterprise AI

Alex Karp, frontier models and the real fight for Enterprise AI Palantir Technologies Inc. Chief Executive Alex Karp's recent broadside against the frontier model vendors put a knife to the throat of the central enterprise artificial intelligence debate. Karp's argument is that frontier model vendors (he didn't mention Anthropic and OpenAI by name) intend to suck the knowledge out of enterprises and destroy the "alpha" companies enjoy through their proprietary data, processes and underlying business advantage. In our last Breaking Analysis we called this approach "data communism," where every firm gains access to the same intelligence. Our counter to data communism is data capitalism, where proprietary advantage remains exclusive to an organization and its broader ecosystem. The familiar graphic below describes how we see the AI software stack evolving. The most important pieces of this stack in our view are the System of Intelligence or SoI and the System of Engagement, the new user/client surface. We believe competitive advantage accrues to firms that can make these two puzzle pieces interact and learn from the reasoning traces of humans. Critical is how they capture tacit enterprise knowledge to take governed and trusted actions both with and without humans. The key issue we explore in this Breaking Analysis is: Who will own the operating intelligence of the enterprise? On the surface, the argument is about closed models versus open models, OpenAI and Anthropic versus Nvidia Corp.'s Nemotron (as an example), and whether enterprises should trust frontier labs with their most sensitive data and workflows. But as Nvidia CEO Jensen Huang said at the GTC 2026 conference, "proprietary versus open is not a thing. It's proprietary and open." The deeper issue is not model choice. It is control. This question goes to the heart of the debate inside theCUBE Research and across the industry. Here are the two ends of the spectrum with two scenarios from theCUBE Research: * Frontier model leadership - One camp believes the frontier model vendors will dominate the stack because their utility, cost curves, research velocity, volume and compute access will outpace everything else. * Dispersed intelligence - The alternative argument is the frontier players lack the mindset, DNA, process knowledge and trust. This camp agrees that the highest-value layer is not the model itself, but the system of intelligence - the enterprise-specific context layer that captures business rules, policies, processes, state and tacit knowledge as governed assets. And the argument is that other players (for example, Palantir, Databricks Inc., Microsoft Corp., Google LLC, Celonis SE and new startups) will cement a more critical position in the AI stack than the frontier model players. Both views can be true. And both acknowledge that: 1) the system of intelligence and the client surface must be part of the winning vendor's stack; and 2) trust and proprietary knowledge must be the exclusive property of the customer. But the Karp conversation unveils the tension between them and we believe will define the future enterprise AI industry power structure. In either case, it's highly likely that silos of intelligence will emerge over the next decade and the industry will remain heterogenous and fragmented. To put a finer point on the topic, Karp's statements imply that OpenAI and Anthropic are stealing data from and overcharging their customers. He positions his company, Palantir, as a critical "application layer" to protect business and governments from those newbies and their bad intentions. He is on a mission to convince businesses they need a more trusted partner than Anthropic or OpenAI. Rather than dealing directly with the two AI firms, he's arguing Palantir (by proxy) should be that intermediary. The frontier model dominance case We've not published extensively on this point of view and it's worth taking a moment to do so here. Much of the following is based on scenario and cost modeling work done by theCUBE Research Analyst Emeritus David Floyer. The methodology and framework is detailed below and draws extensively on Wright's Law, as applied to software. Wright's Law says that as cumulative production rises, costs fall in a predictable way. In AI, the equivalent is not just cumulative production but cumulative usage, cumulative tokens, cumulative feedback, cumulative training experience, cumulative inference optimization and cumulative compute deployment. While some of the AI data forecasts in the methodology are need updating, the principles of the forecasting approach still apply. Here's the argument. Think of the frontier model as a "cognitive surface," the heart of intelligence and the sole source of tokens. It performs reasoning, planning, synthesis and learning. It is built and runs on the most advanced hardware available and continues to improve rapidly. It is capital-intensive, power-dense and scarce by design. Only a small number of organizations can develop and operate frontier models at scale, and enterprises should not attempt to replicate this function. Directly coupled to the frontier model is a new layer that does not exist in traditional enterprise architectures: the System of Intelligence. This layer manages all inputs to and outputs from the large language models. It shapes intent, context, constraints and semantic grounding before intelligence is invoked. It expresses intelligence into actions, system interactions and multimodal outputs after tokens are produced. It hosts security, policy enforcement, compliance, auditability, latency control and integration with enterprise systems. It evolves in lockstep with the frontier model while remaining external to it, preserving both control and adaptability. Though frontier model developers must own their cognitive surface architecturally and evolutionarily, they will almost certainly want to own the SoI and also allow the controlled distribution of both. A fundamental assumption in this scenario must be made explicit: Large enterprises will be permitted by the LLM license terms to operate instances of the cognitive surface locally or within sovereign environments to meet latency, security and regulatory requirements. However, this distribution will occur under strict contractual and technical control. Enterprises will not be able to modify the LLM itself; they will only be able to configure, utilize and integrate it within defined boundaries. But those configurations, and associated data, process logic and underlying enterprise knowledge remain the sole property of the customer. Semantic grounding, safety constraints, interface definitions and evolutionary alignment will remain under frontier model governance. This arrangement preserves enterprise control over data, latency and policy while preventing intelligence fragmentation or semantic drift. The LLM is also the place where a small number of early and strategically positioned software-as-a-service vendors will negotiate with frontier model providers to license access to semantic definitions, interfaces and selected process and execution logic. Likely candidates are SAP SE, Oracle Corp., Salesforce Inc., ServiceNow Inc., and other leading SaaS providers. Vendors of platform services and middleware may also seek certified integration points within the cognitive surface. For example, a platform provider such as Oracle is well-positioned to operate a minimal, authoritative retrieval-augmented generation capability within the cognitive surface, in conjunction with a frontier model provider such as OpenAI. Open-source applications may also be integrated, subject to the same semantic, security and governance constraints. Frontier leaders as this era's disruptors This scenario recognizes frontier players have the world's top AI researchers. They have access to the largest pools of compute. They have massive user volume. They have consumer products and/or consumer-like volumes that act as high-frequency learning loops. They have brand affinity, developer adoption and enterprise pull. And they have the capital to run far ahead of companies that are trying to compete from narrower positions, all while being rewarded and still losing money. As such, this point of view assumes model players will have the lowest cost, highest volume and most functional product. The outlook projects that companies will scale with less labor and achieve 10X productivity relative to current best practice metrics. This economic advantage, the argument says, will overwhelm Karp's concerns about trust because a winner-take-most dynamic and software-like marginal economics will accrue to firms that lead in AI adoption. If the frontier models keep improving at a faster rate than alternatives, and if inference costs keep falling, enterprises will continue to route more work to them. Smaller models and open models will absolutely have a role. They will be used for cost-sensitive tasks, domain-specific workloads, sovereignty requirements, edge deployments and latency-sensitive use cases. But in many workflows, they may be most effective as part of a broader ensemble led by frontier-class models. The most aggressive version of this thesis says that token costs will fall so dramatically that today's budget concerns will look temporary. Enterprises may complain about LLM bills now, but the real comparison is not token cost versus token cost. It is token cost versus headcount. If a company can grow revenue 10x without scaling labor proportionally, tokens become relatively cheap. If agents allow enterprises to compress support, engineering, finance, operations, compliance, analytics and field work into software-driven workflows, the marginal cost of tokens becomes much lower than the marginal cost of people. In that world, the winning providers are the ones that deliver the highest utility per unit of work - not necessarily the lowest nominal token price. That is the frontier model bull case. OpenAI, Anthropic and Google may become the lowest-cost providers of high-utility intelligence because they operate at the largest scale. Their compute purchasing power, model optimization, inference infrastructure and usage volume may give them better marginal economics than alternatives. If that happens, model routing becomes less about avoiding frontier models and more about using them intelligently where their incremental utility justifies the cost. Karp's argument is really about enterprise sovereignty Karp's comments are in a large part a fear campaign against OpenAI and Anthropic. His language was purposefully fiery, and the reported suggestion that frontier model providers could "take the alpha" of a customer's business and transfer it into their weights is attention-getting. Notably, there is no public evidence cited that Anthropic or OpenAI trains on customer data in violation of their terms; OpenAI has publicly said it does not train on customer data. But whether the literal accusation is proven is not the main point. The enterprise fear is real. Karp nailed the sentiment. Customers are asking: If our most sensitive workflows, data, decisions, policies and proprietary operating knowledge flow through a frontier model vendor, are we building our future on a supplier that could one day intermediate us, compete with us, or extract too much margin from us? Palantir's answer is self-serving but thought-provoking: Don't let the model vendor own the enterprise brain. Let Palantir sit between the customer and the model, govern the interaction, route workloads to the best model, preserve sovereignty and make the model interchangeable. That is a classic platform move and the one we've been putting forth based on George Gilbert's work. Palantir wants to own the system of intelligence and treat models as pluggable engines. Its message is the model is important, but the enterprise operating layer is strategic. This aligns with our Enterprise AGI thesis. We have argued that the real enterprise prize is not generalized intelligence in the abstract, but intelligence that is unique to and owned by each enterprise - data, processes, policies, business logic and tacit knowledge turned into governed assets that agents can reason over and act through. The system-of-intelligence counterargument The counterargument is that model utility is necessary, but not sufficient for enterprise AI excellence. Enterprises do not run on intelligence alone. They run on rules, policies, exceptions, permissions, workflows, systems of record, regulatory constraints, domain knowledge, organizational structures and tacit human judgment. A frontier model can reason brilliantly and still not know what the enterprise is allowed to do. It may not know which revenue definition is authoritative. It may not know which customer data can be used in which jurisdiction. It may not know which approval is required before changing payment terms. It may not know whether a support escalation should trigger a field dispatch, a credit memo, a legal review or an executive notification. It may not know the difference between what employees usually do and what policy requires them to do. That is a system-of-intelligence problem. The system of intelligence is the live enterprise map - ontology, digital twin, semantic layer, business-process model, policy fabric and operational context layer. It is where enterprise data, process logic, business rules, skills and tacit knowledge become assets. It is what allows agents to act safely and confidently. This is where Palantir has a strong argument. Palantir has spent years building ontology-driven operational systems for complex institutions. It understands that the application layer is not just a user interface layer. It is where data, decisions, workflow and governance meet. Databricks is attacking a related problem from the data platform side. Genie, Genie Ontology, Agent Bricks, Unity Catalog, Unity AI Gateway and Omnigent represent an effort to move from governed data infrastructure into engagement, intelligence and agency. In our research, we described this as Databricks moving up the enterprise intelligence stack - from data platform toward system of intelligence. But there is a difference. Palantir is closer to an executable ontology and operational decision layer (see graphic below). Databricks is building from governed data, semantic context and agent governance and is currently between levels five and six in the diagram below based on our current assessment. Both approaches are important. Palantir has a far more mature ontology story and Karp took a swipe at firms such as Databricks, alluding to the fact that everyone is now talking about ontology. Regardless, the leaders are converging on the same control point - the SoI. Open models are a wedge, not the whole story Karp's emphasis on Nvidia's Nemotron and U.S.-made open models is clever. For government, defense and regulated industries, model sovereignty is vital. Customers may prefer open or domestic models where security, auditability, deployment control or geopolitical concerns outweigh absolute benchmark performance. But open versus closed is not the real dichotomy. Jensen Huang's positioning -- proprietary and open, not proprietary versus open - is probably closer to how the market develops. Enterprises will use a mix of products. They will route workloads across frontier models, open models, specialized models, small models and internal models. The model router becomes an important economic and governance control point. Palantir's Evolve and Databricks' Unity AI Gateway both point in this direction. Each is trying to help customers decide which model to use for which workload based on cost, performance, security, governance and policy. That is the most pragmatic enterprise answer. The ideological open-source debate is important, but the operational reality is model optionality. Still, Karp's argument has special force in government. If an American open model becomes "good enough" for a classified battlefield workflow, and if Palantir can tune, govern and deploy it securely through its platform, then the closed frontier model may not be necessary for that use case. Good enough plus sovereign plus governed can beat best model plus external dependency. But that will not apply universally. For the hardest reasoning tasks, frontier models may continue to lead. The market will not standardize on one model class. Can the frontier labs build the system of intelligence? This is the crux of the debate. One side says the frontier vendors will inevitably move into the system of intelligence because they must. Consumer AI gives them volume, learning and brand, but enterprise productivity is the obvious total available market. If they stop at model application programming interfaces, they risk becoming suppliers to the platforms that actually capture enterprise value. That is not a sustainable end-state for companies spending tens or hundreds of billions on compute and research. From this perspective, the frontier labs will build, partner and acquire their way up the stack. They will create enterprise memory, shared skills, workflow engines, governance layers, agent orchestration, model routers, data connectors, domain packages and forward-deployed implementation capabilities. Consumer volumes are a feature, not a drawback. Personal memory becomes workgroup memory. Skills become business logic. Agent traces become process intelligence. Over time, they develop or buy the missing system of intelligence. The other side says this is not in their DNA. The frontier labs are research and model-scaling organizations. Enterprise software is a different muscle. It requires long-cycle implementation, governance, domain modeling, integration, change management, procurement patience, compliance, auditability and trust. Building a durable system of intelligence means encoding how enterprises actually operate - and that is deeply messy. This camp believes Palantir, Microsoft, SAP, Salesforce, Databricks, Snowflake, ServiceNow, Celonis, Oracle and others have better starting positions because they already sit closer to enterprise data, workflow, identity, systems of record, governance or business process. But even skeptics of the frontier labs must concede that if OpenAI and Anthropic decide the system of intelligence is existential, they have the capital, distribution, model utility and talent magnetism to pursue it aggressively through M&A and partnerships. The question is whether they can absorb enterprise complexity without losing the qualities that made them successful. The likely market structure Our view is that the industry will not resolve cleanly. History suggests the market will fragment across several control points: The leaders will capture multiple layers. The most valuable companies will combine model intelligence, enterprise context, governance and permission to act. That is why the Karp-versus-frontier-lab debate is so interesting. Karp is arguing that Palantir owns the enterprise context and can commoditize the model. The frontier model bull case argues that model utility and economics will become so overwhelming that frontier vendors will eventually own or absorb the context layer. The truth may be that both sides need each other until they no longer do. Action item For chief information officers, chief technology officers and business technology executives, the takeaway is not to choose a religious position on open versus closed models. It is to avoid architectural dependency. Enterprises should assume a multimodel future. They should use model routers. They should preserve optionality across frontier, open, domestic, specialized and small models. They should separate model choice from enterprise context. Most importantly, they should begin building their own system of intelligence. That means capturing: * Authoritative metrics * Business definitions * Policies and permissions * Process logic * Decision rights * Workflow state * Human skills and tacit knowledge * Agent traces and feedback loops The model market will change quickly. The enterprise operating model should not be trapped inside any one model provider. Bottom line Karp is right that enterprises need an intermediary layer between raw models and mission-critical operations. The model alone is not the enterprise brain. The frontier model advocates are right that scale, volume, compute and learning curves are powerful. The best models may become dramatically more capable and cheaper faster than many expect. Tokens may prove far cheaper than labor, and frontier vendors may enable companies to scale revenue without scaling headcount. The unresolved question is who captures the resulting value. If the system of intelligence remains outside the frontier labs, companies such as Palantir, Databricks, Microsoft, SAP, Salesforce and ServiceNow will have the opportunity to make models interchangeable and capture the enterprise control point. If the frontier labs use their scale to build, buy or partner into the system of intelligence, they could dominate far more of the enterprise stack than today's skeptics expect. This is the strategic debate the industry will wrestle with: Will enterprise AI value accrue to the providers of the most powerful intelligence? Or to the platforms that understand how each enterprise actually works? Our answer, for now, is that Enterprise AGI requires both. But the vendor that combines frontier utility with governed enterprise context will win the biggest prize.

Open and shut source case: There's method in Palantir CEO Alex Karp's 'mad' outburst against Big AI overselling corporates

Palantir CEO Alex Karp has fiercely criticized the US AI industry, particularly OpenAI and Anthropic, for their token-based pricing and 'oversold' AI capabilities. This sparks a debate between closed and open-source AI models. While closed models cite high development costs, proponents of open source argue for greater efficiency and cost savings, potentially saving the global AI economy billions. On Wednesday, Palantir CEO Alex Karp came down heavily on the US AI industry led by OpenAI and Anthropic for using token-based pricing and having 'completely oversold irresponsibly' capabilities and benefits of their AI products. His searing outburst signals the latest rift between creators of foundational software, and firms integrating them to meet enterprise requirements. On the one hand, while the industry has found a way to package and bill AI, Karp's displeasure echoes corporate resentment over how companies are being charged - or overcharged. When the body corporate complains, it's often about their compulsion to ride the current wave, and how it's hurting their P&L. When an AI insider states his view, as vehemently as Karp did, it makes the world wonder whether the creators of AI's foundational models have actually done enough in terms of optimising costs to benefit customers. As a result, divergent industry camps are emerging. Closed-model AI heavyweights, like OpenAI's Sam Altman and Anthropic's Dario Amodei, believe that the pricing reflects the huge costs of deploying GPU clusters, training frontier models and their massive energy spends. Here, the infra belongs to the provider, weights are hidden and payment is by tokens consumed, alongside ubiquitous price wars, in the middle of a battle for market share as some prominent players race toward IPOs. Proposed 'sweeteners', like OpenAI reportedly making plans to offer the US government 5% stake in the company, as AI firms face governmental scrutiny over possible misuse of advanced models, add to the hostility within AI ranks. On the other side, there's Meta's chief scientist Yann LeCun and others, unrelenting proponents of open source, who believe closed systems are inefficient, their metering systems flawed and their architectures unsustainable. Meta's Llama series echoes this philosophy, allowing enterprises to host models, and eliminate token billing. Closed frontier models are ahead when the reasoning is complex and context is long. They are also best for modest volumes. So, there's no compulsion to run your own GPUs round the clock. Owned or rented GPUs are best when data privacy is critical, and when volumes run into millions of tokens a day. Open models like China's DeepSeek are altering the maths entirely, supposedly delivering 90% of the benchmark performance at about a sixth of the cost. Daniel Yue at Georgia Institute of Technology has discovered that optimal reallocation of demand from closed to open models could save the global AI economy $25 bn annually. The fact that GitHub Copilot has shifted from monthly subscriptions to token tracking from June 1, after its CFO called out a gross margin decline in an April earnings call, corroborates where the margins lie. But the line between closed and open providers is blurring, as closed providers offer private deployments and open models carry newer usage restrictions. Days before Karp's outburst, he had announced an expanded partnership between Palantir and Nvidia, integrating the latter's open-weight Nemotron into the former's AI platform. Target customers are government agencies and enterprises, who can both run custom models in onsite environments. Karp had also called out the US government's reliance on firms like OpenAI and Anthropic for the development of military and national security applications - which he deemed 'effing insane' - asking if the battlefield was now being outsourced to the consensus view in Silicon Valley. Concerns about national security were accompanied by dire warnings about China's rapidly advancing capability in AI development. Karp also spoke about how data retention was so important for organisations. The sequence of events - collaboration with Nvidia on June 29, making public Palantir's AI sovereignty manifesto on June 30, and Karp's headline-grabbing interview on July 1 - suggests that Karp and Palantir have been working to a plan. Palantir's shares rose 8-9% on the day of his outburst. This appears to be a case of clear narrative framing, impeccably executed in three moves, in a chronological sequence that shifted market sentiment: Announcement of Palantir's partnership with Nvidia to deploy their open-source model was certainly not about any ordinary tie-up. It was a collaboration with the most valuable tech company on the planet, conveying sound foundations subtly, yet overtly. The 9-point AI sovereignty manifesto on X, urging institutions to protect their data, about how data sovereignty 'dictates your institution's future', and how 'tokenmaxxing' encourages profligacy in IT spending, instead of charging for value. This was aimed to trigger fear of exploitation. Karp's 'catharsis', mocking competitors, getting agitated - the interviewer interjecting with 'You sound pretty angry' after his near-3-min rant - made him look the archetypal soothsayer, which the market possibly interpreted as the unfiltered conviction of an insider. From being seen as a niche data analytics firm, almost overnight, Palantir was repositioned as a 'protector of institutions', the secure defence shield of the corporate world against exploitation by frontier AI models. Karp harped on pre-existing anxieties, and offered everyone a way out by switching to sovereign data stacks. Was it a master class in corporate theatre, insightful strategy or psychological engineering? Was it marketing or manipulation? Or was it all at once? The jury's out. We'll know over the weeks, months and years.

Palantir CEO Alex Karp doesn't hold back in interview as he rails against AI industry

Palantir CEO Alex Karp doesn't hold back in interview as he rails against AI industry Palantir Technologies Inc.'s Chief Executive appeared to go into meltdown mode during an interview with CNBC today where for 20-odd minutes he went off script after being asked to discuss his company's ongoing deal with chip maker Nvidia Corp. The conversation turned to the AI industry where he lambasted the U.S. government's relationship with leading AI firms and railed against AI tokenization that he claimed has left enterprise customers feeling frustrated. He called the high prices being charged a "wealth tax" on businesses. Speaking about the government's reliance on OpenAI Group PBC and Anthropic PBC concerning the development of AI for military and national security applications, he asked, "Are we really going to outsource the battlefield of this country to the consensus view in Silicon Valley? That is effing insane." "And by the way," he went on, explaining that "every single enterprise in this country" despite their apparent reticence on the subject, are upset with the "nerds" of Silicon Valley. His meltdown then hit a high point when he alleged that some of those nerds were "on drugs", adding that he, of course, was abstinent from such substances. "It starts with the billionaires...every single person at this table is going to be paying a wealth tax." "Livid," was the word he used to describe enterprise customers who are "paying for tokens that create no value," while handing over their data to the major firms and in the end, taxing the public who will end up paying for it in some for or other. The reason, he ventured, was the "models have completely over...irresponsibly been over-sold...and the sell is, it's dangerous for everyone, which is why I can give it to all your adversaries, but I can't give it to the Department of War." "You sound pretty angry," the host interrupted. He countered, his finger wagging, "No, this is the voice of American business that is being channeled through me and I'm telling you it is absolutely a problem for this country." Some media later referred to the interview as a, "Televised nervous breakdown."

Alex Karp Bashes OpenAI, Anthropic Token Model -- 'Something Has Gone Completely Wrong' - NVIDIA (NASDAQ:N

"Something has gone completely wrong," Karp told CNBC's "Squawk Box," adding, "I'm not throwing shade at them." After the anchor said his comments "sound like shade," Karp responded, "No, no, no. This is reporting." The billionaire businessman said the prevailing enterprise mindset has become, "I'm going to chillax and waste my time with tokens," warning, "I'm going to get no value and they're going to get my IP." Enterprises Pivot From Tokenmaxxing to ROI Amid the shift, enterprises are moving away from tokenmaxxing and toward return on investment, increasingly adopting lower-cost open-weight models or building their own tools. It's a familiar theme. Karp made similar remarks on a June podcast, warning that enterprises are "token maxing," or overusing AI without meaningful productivity gains. He said frontier labs are "super charismatic with investors" but "super not charismatic with enterprises." During his appearance, Karp also cautioned the industry shouldn't underestimate China's accelerating AI progress. Nvidia Deal Anchors 'AI Sovereignty' Push Karp, talking about this during the interview, said, "What aligns me with Nvidia... is control over their compute, their models, their data stack and their alpha. They want to know they own the means of production. It's not being transferred to someone else." On Tuesday, ahead of the interview, Palantir also posted a nine-point "AI sovereignty" manifesto on X criticizing tokenmaxxing and urging firms to retain control of their data. Trading Metrics, Technical Analysis Palantir has a market capitalization of $301.41 billion, a 52-week high of $207.52 and a 52-week low of $106.38. The technology stock has fallen 25.10% year to date. Price Action: PLTR closed the regular session on Wednesday up 7.77% at $125.73, according to Benzinga Pro. With a strong Growth score of 97.69, Benzinga's Edge Stock Rankings indicate that PLTR has a negative price trend across all time frames. Photo Courtesy: Meir Chaimowitz on Shutterstock.com Disclaimer: This content was partially produced with the help of AI tools and was reviewed and published by Benzinga editors. Market News and Data brought to you by Benzinga APIs To add Benzinga News as your preferred source on Google, click here.

Palantir CEO slams OpenAI, Anthropic's token model

Palantir CEO Alex Karp has slammed the token-based pricing of AI giants like OpenAI and Anthropic, arguing businesses are wasting money without clear returns. Karp said companies are now moving away from simply buying more tokens and are instead focussing on whether their AI spending delivers a clear return on investment. Palantir chief executive Alex Karp has criticised the token model used by Anthropic and OpenAI, saying businesses are spending too much on artificial intelligence without seeing enough value in return. "I'm not throwing shade at them, but something has gone completely wrong," he told CNBC's 'Squawk Box'. "The basic view among enterprises in this country is I'm going to chillax and waste my time with tokens." Most leading AI companies charge customers based on tokens (small units of text that AI models process when users type prompts or receive responses). As newer AI models become more powerful, they also require more computing power, making them significantly more expensive to run. Karp said companies are now moving away from simply buying more tokens and are instead focussing on whether their AI spending delivers a clear return on investment. That shift is also pushing businesses towards open-weight AI models. Unlike closed models, open-weight models make their trained parameters available, allowing companies to customise them and run them on their own infrastructure at a much lower cost. Karp warned that companies should not underestimate China's pace of progress in AI development. Chinese models are improving rapidly, increasing pressure on leading US AI firms. He added that more businesses are choosing to build and train their own models using their internal data, rather than relying entirely on third-party AI providers. "What aligns me with Nvidia, and I think is what the technical customers want, which is control over their compute, their models, their data stack, and their alpha," Karp told CNBC. "They want to know they own the means of production. It's not being transferred to someone else." In simple terms, Karp argues that companies want full control over the computing infrastructure powering their AI, the data used to train it, and the models themselves, instead of depending on external AI labs. Earlier this week, Palantir expanded its partnership with Nvidia to help US government agencies build custom AI models using Nvidia's computing infrastructure. The arrangement will allow agencies to train AI on their own data while retaining ownership of the models and the knowledge embedded in them. Meanwhile, on Tuesday, Palantir published a nine-point manifesto on X, advocating AI sovereignty -- the idea that organisations and governments should own and control their AI systems and data. The post also criticised "tokenmaxxing" as a business model.

Palantir CEO Alex Karp Says AI Labs Are Chasing 'Tokens' - Palantir Technologies (NASDAQ:PLTR)

Palantir CEO Alex Karp Says AI Labs Are Chasing 'Tokens' While Enterprises Fear for Their IP: 'Something Has Gone Completely Wrong' Alex Karp Says Enterprises Want Value -- Not More 'Tokens' Karp delivered a blistering critique of leading AI companies during an appearance on CNBC's Squawk Box, arguing that the industry has lost sight of what enterprise customers actually want. While joking that he would enjoy debating Anthropic CEO Dario Amodei privately and insisting he was "not throwing shade" at AI leaders, Karp said the broader AI ecosystem has "gone completely wrong." "The basic view among enterprises in this country is 'I'm going to chillax and waste my time with tokens, I'm going to get no value, and they're going to get my IP,'" Karp said. According to Karp, executives have privately told him they are increasingly worried that AI companies could gain access to their proprietary data and their "alpha" -- the unique competitive edge that differentiates a business. "We need to build trust," he said, adding that companies are growing tired of AI promises that fail to translate into meaningful business outcomes. Palantir CEO Raises Concerns Over AI Control And Enterprise Data Karp also questioned whether critical decisions surrounding AI, particularly in defense applications, should be influenced by the policy preferences of Silicon Valley companies. Appearing to reference Anthropic's reported disagreements with the U.S. government over certain military uses of AI, he said, "Are we really going to outsource the battlefield of this country to the consensus view in Silicon Valley? That is effing insane." When CNBC co-anchor Becky Quick remarked that he sounded angry, Karp replied, "No. This is the voice of American business that is being channeled through me." He went on to say that many CEOs privately share his frustration, saying they are "twice as livid" about the current direction of enterprise AI. Investors And Industry Leaders Echo Karp's Concerns The interview quickly sparked reactions across the technology community. Futurum Equities Chief Market Strategist Shay Boloor said Karp's message reflected a growing demand among enterprises for ownership, security and control over their "compute, models, data stack and alpha," rather than simply paying for AI tokens. Investor and All-In Podcast host Jason Calacanis argued that frontier AI companies could use heavily discounted AI access to attract customers before ultimately leveraging that position to compete with them, adding that open-source AI is the best defense. Meanwhile, Moor Insights & Strategy CEO Patrick Moorhead said every CEO should watch the interview, writing that Karp was voicing concerns "few will say" publicly about the risks businesses face when relying on frontier AI models. Price Action: Shares of Palantir rose 7.77% to close at $125.73 on Wednesday, gaining $9.06 and edged up another 0.71% to $126.62 in Thursday's premarket trading, according to Benzinga Pro. According to Benzinga Edge Rankings, Palantir ranks in the 97th percentile for growth, though its stock has posted negative returns over the short, medium and long term. Disclaimer: This content was partially produced with the help of AI tools and was reviewed and published by Benzinga editors. Photo courtesy: Shutterstock Market News and Data brought to you by Benzinga APIs To add Benzinga News as your preferred source on Google, click here.

Something Has Gone Completely Wrong in Enterprise AI

There is a moment in every boom when the music is still playing, the lights are still flashing, and someone finally notices that the bar tab has been left on the table. Takeaways * The AI land grab is shifting from who has the smartest model to who keeps the keys to the factory. * Token usage can look like progress, but rising activity is not the same as compounding proprietary advantage. * Data, weights and workflows are the institutional memory. Hand them away too freely and the moat can become a toll road. * The next leg of the AI race will be decided less by benchmark theatre and more by cost, control and who captures the margin. Something Has Gone Completely Wrong There is a moment in every boom when the music is still playing, the lights are still flashing, and someone finally notices that the bar tab has been left on the table. That is roughly where enterprise AI feels today. Palantir's Alex Karp may have delivered the message in his usual " effing" blunt-force style, but underneath the fireworks sits a serious concern. The frontier labs have built extraordinary machines. No one sensible disputes that. The issue is whether the people renting those machines are actually building a business advantage, or simply feeding coins into someone else's meter. Karp's complaint is not that AI does not work. It is that the economics increasingly look like a casino where the house owns the chips, the tables, the cameras and perhaps a copy of every card you have ever played. Companies are being told to embrace token usage, embed models into every workflow, build agents, automate decisions, scale experimentation and move faster than the competition. The promise is seductive. The more AI you use, the more productive you become. The more productive you become, the more valuable the business becomes. But there is a catch. Every time a company pushes proprietary data, internal process knowledge, customer information and years of accumulated institutional judgement through an external model, it is not merely consuming AI. It may be exporting part of its operating memory. That is why Karp's language around "AI sovereignty" matters more than the headline theatrics. He is asking a question that many boards have not yet properly confronted: when you build on someone else's model, someone else's cloud, someone else's weights and someone else's pricing system, how much of the future business do you really own? The token model sits at the heart of that tension. On paper, it looks elegant. You pay for what you use. A few tokens here, a few million there. It feels like turning on a utility. But utilities normally get cheaper as they scale. AI can feel like the opposite. The more deeply a company embeds it into research, coding, customer service, compliance, trading, legal workflows, logistics and internal decision-making, the more the meter starts spinning. The demo may cost pennies. Production is where the bill arrives. An AI agent running across a corporate system is not one prompt and one answer. It can call multiple models, retrieve documents, scan data, invoke tools, write code, check code, run another model, audit the output and then start the whole process again. Multiply that across an enterprise and the token jar begins to look less like a subscription service and more like a taxi with the meter running in heavy traffic. Karp's point is that companies may be confusing activity with ownership. The dashboards show rising AI usage. Token consumption climbs. Internal teams report more pilots, more prompts, more automation and more experimentation. It looks like progress because everything is moving. But a hamster wheel moves too. The question is whether all that motion compounds into proprietary intelligence, or whether it simply compounds into a larger invoice for the model provider. That is where the data issue becomes central. Data is not just fuel. It is the memory of the institution. It is the record of what worked, what failed, how customers behave, where risk lives, what pricing decisions produced good outcomes and what patterns only become visible after years of repetition. A company's edge is rarely one giant secret sitting in a vault. More often it is thousands of small decisions, tiny operational habits, customer relationships, historic exceptions and accumulated scar tissue. Put enough of that through an external system and the danger is not that someone steals the whole vault overnight. The danger is that the moat slowly turns into a public road. Karp's line that "controlling your weights is controlling your fate" is intentionally dramatic, but not entirely wrong. Weights are where the learning lives. They are the compressed residue of data, training, fine-tuning and repeated interaction. If a company gives up control of the intelligence layer, it may eventually find itself renting back part of the competitive advantage it helped create. That is a difficult proposition for any serious enterprise. It is even more difficult for governments, defence organisations and critical infrastructure operators. You would not outsource the command room of a battleship to whichever vendor has the most polished sales deck that quarter. You would not let a third party own the map, the radar, the radio and the operating manual, then charge you by the message every time a storm appeared on the horizon. Yet that is not far from the question Karp is raising around national security. If AI becomes embedded in intelligence, logistics, battlefield decisions, cyber defence and critical systems, then control over the data, models and deployment architecture is not a procurement detail. It becomes part of national capacity. This is also why the growing interest in Chinese open-weight models is more than a curiosity. The shift is not necessarily a declaration that Chinese models are better across the board. The frontier US labs still lead in many areas, particularly at the cutting edge of reasoning, coding and multimodal capability. But enterprises are beginning to behave like rational buyers. They are comparing performance, cost, reliability, deployment flexibility and the ability to keep the system close to home. For some use cases, the most advanced model in the world is not the most useful model in the building. A cheaper open-weight model that can be hosted internally, tuned around proprietary data and controlled by the enterprise may deliver a better economic outcome than a brilliant frontier model accessed through an expensive metered pipe. That does not mean the premium model loses. It means the market starts asking the question it always asks eventually: what am I getting for the price? That is where the AI boom is beginning to change character. The first phase was awe. Look what these models can do. The second phase was fear. What happens if we fall behind. The next phase is audit. Who owns the system, who owns the data, who owns the weights, who owns the customer relationship and who captures the margin. This is the part of the cycle where the slogans get tested against the spreadsheets. Palantir's response is to push a sovereign deployment model with Nvidia, where the customer retains control over compute, models, data and weights rather than simply renting intelligence through a frontier API. That is not merely a technical architecture. It is a different answer to the value-capture question. The frontier labs want to become the intelligence layer of the global economy. Palantir is arguing that no serious institution should hand over the keys quite so easily. Both sides have a point. The frontier labs have created products with genuine, extraordinary capability. They are not selling smoke. But capability alone does not settle the economics. A model can be brilliant and still be too expensive. It can be powerful and still be too externally controlled. It can save time for a department while quietly transferring long-term value away from the enterprise. That is the uncomfortable part of the story. The real AI race may not be between OpenAI, Anthropic, Google, Meta, DeepSeek and the rest. It may be between companies that use AI to compound their own institutional intelligence and companies that use AI to become more dependent on someone else's. The difference may not show up in the first quarter. It may only become obvious years later, when one company owns the factory and the other is still feeding coins into the machine. Karp had warned against underestimating China's progress; these examples illustrate the trend in real time.

Palantir Accuses OpenAI and Anthropic of Siphoning Companies' DNA

Speaking Wednesday on CNBC, Alex Karp launched a direct attack on the leading artificial intelligence labs. The chief executive of American technology giant Palantir said OpenAI's and Anthropic's business model, largely built on token consumption, is increasingly being challenged by major US companies, which are paying dearly for operational gains that are still difficult to quantify. In his view, the bigger risk, however, is the protection of intellectual property. By entrusting their data, internal methods, and certain critical processes to proprietary models, companies could, over time, expose part of their competitive edge to outside players whose infrastructure and operating rules they do not fully control. For Karp, that dependence amounts to transferring part of companies' 'alpha' to AI labs, even though that alpha is often their most strategic asset. The charge goes straight to the heart of the big AI providers' model: selling access to ever more powerful models, while gradually becoming a central chokepoint for their clients' data and decisions. But the broadside also serves Palantir's strategy, which promotes a more sovereign form of AI, controlled by companies and governments. The group has just strengthened its partnership with Nvidia around open models that can be deployed in secure environments, notably for US government agencies and critical infrastructure.