Chamath Palihapitiya said his software company is moving away from Cursor because it was spending too much on tokens

Chamath Palihapitiya, a prominent venture capitalist, has expressed his astonishment regarding the escalating expenses associated with AI at his software startup, 8090. During a recent episode of the "All-In Podcast," he revealed that their costs have surged more than threefold since November of last year. Palihapitiya highlighted the significant financial burden stemming from various sources, including AWS inference fees and their expenses with Cursor and Anthropic. As 8090 aims to revolutionize legacy software, the company is on track to incur annual AI costs nearing $10 million. He voiced concern that while expenses rise sharply, their revenues are not keeping pace. "The problem is that my costs are going up 3X every three months, while my revenues are not," he stated. The tech industry is currently grappling with the disruptive impact of AI on traditional sectors, and Palihapitiya's worries echo a broader apprehension regarding the sustainability of current AI funding models. He noted that the ongoing support from major venture capital firms fuels a consumption-heavy approach to AI, reminiscent of the early days of Uber's pricing strategy. In particular, Palihapitiya pointed to Cursor, a widely used AI coding tool, as a major contributor to 8090's rising costs, suggesting that Anthropic's Claude Code offers a more economical alternative. "We need to migrate off of Cursor; it's just too expensive compared to Claude Code," he remarked, advocating for a shift to more cost-effective solutions. The growing financial burden of AI is prompting CFOs to take notice, with industry figures like OpenCode creator Dax Raad noting the increasing expenses associated with AI in engineering roles. Palihapitiya also mentioned a specific usage technique dubbed "Ralph loops,” which involves repeatedly feeding the same prompt into an AI model, leading to inflated bills. Looking ahead, he emphasized the necessity for greater flexibility in switching between AI models, citing recent tensions between Anthropic and the Pentagon as a prime example of the strategic challenges that arise from rigid systems. "We need to gain more flexibility to swap between models without everything breaking," he concluded, underlining the dual nature of the issue as both a cost and strategic flexibility concern.