Artificial intelligence has become the corporate world’s favourite tool. Companies are pouring billions into AI software, encouraging employees to use chatbots daily and measuring adoption rates almost like performance metrics. But not everyone is convinced that more AI automatically means better results. Palantir CEO Alex Karp has become the latest tech executive to question the industry’s growing obsession with AI usage. Speaking during a live interview at Palantir’s AIP Con 10 event, Karp compared excessive AI use to a porn addiction, arguing that businesses are focusing too much on how often employees use AI and not enough on whether it actually creates value.
“Really, we call it the demastibatory, like get off masturbation thing internally,” Karp said while discussing Palantir’s approach to tracking AI usage. “Sure, it’s like people are just sitting there all day kind of like a porn addiction,” he added.
The Growing Backlash Against ‘Tokenmaxxing’
Karp’s comments arrive at a time when many companies are starting to rethink their AI strategies. During the early AI boom, heavy usage was often viewed as a positive sign. More prompts, more tokens and more chatbot interactions were seen as evidence that employees were embracing new technology.
Now that enthusiasm is colliding with reality. As AI providers increasingly charge based on usage, businesses are discovering that AI bills can rise very quickly. Several major companies have recently acknowledged that higher AI spending isn’t always producing equally impressive business results.
Reports have suggested that Amazon removed an internal AI leaderboard after concerns that some employees were using AI simply to boost their numbers. Uber executives have also publicly discussed the challenge of connecting growing AI expenses with measurable returns.
“More Tokens Means More Slop”
Karp’s remarks echo comments made by Palantir CTO Shyam Sankar during a recent earnings call.
Sankar argued that businesses should stop assuming that more AI activity automatically creates more value.
“More tokens means more slop,” Sankar said.
According to Palantir, the real benefit comes when AI is integrated into actual business operations rather than being treated as a tool employees use for the sake of usage.
Microsoft CEO Satya Nadella Reveals He Uses 100 AI Agents, Says They Need Rules And Oversight
Where AI Helps And Where It Falls Short
Karp acknowledged that large language models are useful for many tasks. Generating summaries, answering questions or creating reports are areas where AI can be highly effective.
However, he argued that more complex challenges involving supply chains, manufacturing, defence operations and industrial systems still depend heavily on human judgement and structured workflows.
“They are enhanced by large language models. They are not replaced by large language models,” Karp said.
