- Palantir CEO Alex Karp warns businesses focus too much on AI usage, not value creation
- Karp compares excessive AI use to a porn addiction, criticising "tokenmaxxing" trends
- Tokens measure AI use but higher consumption doesn’t guarantee real business outcomes
Artificial intelligence is transforming the way businesses operate across industries. However, Palantir CEO Alex Karp believes many companies are becoming overly focused on adopting AI without clearly demonstrating how it creates value.
Speaking during a live interview with TBPN at Palantir's AIPCon 10 event, Karp compared excessive AI use to "porn addiction". He criticised “tokenmaxxing”, the trend of chasing high AI usage and compute consumption without regard for real business outcomes.
“People are just sitting there all day, kind of like a porn addiction,” he said while discussing internal AI usage metrics. Karp argued that organisations have prioritised increasing token consumption over assessing tangible value delivered.
In the AI industry, tokens are the units that large language models use to process and generate text. Most providers bill by token count, making usage a key metric for both vendors and customers.
Karp's critique aligns with comments made by Palantir CTO Shyam Sankar during last month's earnings call. Sankar said Palantir operates “a no slop zone” and warned that cheaper AI and more tokens do not necessarily translate into real business value.
"More tokens means more slop, and the more commodity cognition you consume, the more you need a system that can prevent the economic harm, so you can harness the economic value," Sankar said, as quoted by Business Insider.
Karp said AI models excel at many tasks, such as producing a report on China's GDP growth. However, complex business challenges involving supply chains, industrial operations, military logistics, manufacturing and oil and gas require continuous decision-making and precise workflows that go beyond chatbot responses. In those areas, he argued, large language models can augment human work but cannot replace the underlying processes.
"They are enhanced by large language models. They are not replaced by large language models," he said.
According to Karp, AI capabilities will become widely accessible over time. The harder challenge, he said, is identifying which business problems are worth solving.
