Palantir CEO Alex Karp has launched one of his strongest attacks yet on the business models of leading AI companies, arguing that OpenAI and Anthropic are charging enterprises for expensive AI usage without delivering proportional value. His remarks come as businesses increasingly question whether soaring AI costs are translating into meaningful returns.
Palantir Chief Executive Officer Alex Karp has criticised the pricing strategies of leading artificial intelligence companies, claiming something has “gone completely wrong” in the AI industry as businesses spend billions of dollars on services that often fail to generate measurable value.
Speaking during an interview with CNBC, Karp singled out OpenAI and Anthropic, arguing that their token-based pricing models encourage customers to pay for AI usage instead of business outcomes. His comments reflect growing frustration among enterprise customers as AI spending continues to rise while questions about return on investment become more common.
«”Something has gone completely wrong,” Karp said, describing the current state of AI pricing as unsustainable.»
According to Karp, many organisations have become obsessed with what he called “tokenmaxxing”—consuming ever-increasing numbers of AI tokens without considering whether those interactions actually improve productivity or create business value. He argued that if AI systems genuinely produced transformational results, providers would price their products according to the value customers receive instead of charging primarily for token consumption. “If these models create enormous value, then charge for the value. Don’t charge for tokens that create no value,” Karp said. Karp also raised concerns about how some AI companies handle customer information.
He suggested that businesses risk exposing valuable intellectual property, proprietary data and competitive insights while using third-party AI platforms, questioning whether organisations fully understand how their information may be used to improve future AI models.
“American businesses are livid,” Karp said, adding that companies are increasingly worried about giving away their “alpha” while paying significant fees for AI services. His remarks arrive at a time when enterprise AI costs are climbing rapidly.
As organisations deploy AI assistants across software development, customer support, research and internal operations, token consumption has become one of the largest recurring expenses associated with generative AI adoption. Many businesses are now looking for ways to reduce those costs without sacrificing performance.
Karp believes the answer lies in AI systems that integrate directly with enterprise operations instead of functioning as standalone chatbots. Palantir has increasingly positioned itself as an AI orchestration platform, allowing organisations to deploy multiple AI models while keeping sensitive business data inside secure environments.
See Also: Yann LeCun Called ChatGPT Hopeless for Robotics. He Has $1 Billion and a Different Plan.
The company’s approach contrasts with AI developers whose primary revenue comes from charging customers for access to large language models. Karp insisted that enterprises should retain ownership of their data and intellectual property rather than allowing external AI providers to benefit from it.
His comments also reflect a broader shift taking place across the AI industry. Over the past two years, much of the competition focused on building larger and more capable foundation models. Today, enterprise customers are becoming more interested in practical outcomes, implementation costs and measurable returns than benchmark scores alone.
Industry analysts say that change is forcing AI companies to rethink how they price and deliver their services. Businesses are no longer asking only which AI model is the smartest. They are increasingly asking which solution produces the greatest business impact at the lowest cost. Karp’s criticism extends beyond pricing.
He warned that Silicon Valley has oversold the capabilities of generative AI, creating unrealistic expectations among corporate leaders while encouraging excessive spending on technologies that may not yet be ready to solve complex enterprise problems. “These models have been completely, irresponsibly, oversold,” Karp said, arguing that trust between AI vendors and enterprise customers must be rebuilt.
Despite his criticism, Karp acknowledged that frontier AI models remain remarkable technological achievements. His concern centres on how those technologies are being commercialised and marketed to businesses. The debate highlights a growing divide within the artificial intelligence industry.
Foundation model developers continue racing to build increasingly powerful AI systems. Enterprise software companies, meanwhile, are focusing on helping organisations deploy those models securely, efficiently and in ways that generate measurable returns. As corporate AI budgets continue expanding, pricing models are likely to come under even greater scrutiny.
Companies may prove increasingly unwilling to pay for raw AI usage alone. They will expect AI investments to deliver tangible business value. Karp’s message reflects that changing reality.
Success in the next phase of artificial intelligence may depend less on building the smartest model and more on proving that AI can consistently solve real business problems at a price customers believe is worth paying.
About the Author
Marcel Chidozie is a tech analyst and writer covering foreign news, fintech, and emerging technologies at TechRegard. Based in Nigeria, He's passionate about translating complex tech developments into compelling, accessible stories for diverse audiences. His work focuses on how technology shapes innovation across Africa and globally.
