A divide has emerged in the startup world regarding artificial intelligence token consumption. While some founders aggressively encourage their teams to spend heavily on AI compute tokens to maximize productivity, others dismiss this "tokenmaxxing" trend as financially reckless. Token spending strategies now range from mandatory minimum quotas to unrestricted budgets, reflecting differing philosophies on innovation and cost management. Kavitta Ghai, cofounder of Nectir, has implemented a strategy requiring her engineers to spend thousands of dollars on AI tokens monthly. Initially setting a quota of $100 and later $200 per week, the requirement has evolved into an expectation of significant monthly expenditure. Ghai reports that this approach has transformed engineer attitudes, with skeptics now viewing AI tools as an "army of coders." However, Ghai distinguishes her company from the broader Silicon Valley trend of tokenmaxxing, emphasizing internal competition over external trends. Conversely, a growing number of startups are racing to burn tokens. Aron Solberg of Risotto describes token spending as a "force multiplier," noting his monthly expenses rose from one-tenth to the current $4,000 to $5,000 range within six months. Solberg argues that spending money is essential to making money. Similarly, Vybe's Quang Hoang maintains an unlimited credit policy and considers minimum quotas, while Y Combinator, the prominent startup accelerator, encourages participants to "let it rip" using free token credits. Lance Yan of Traverse, a student cofounder, credits these credits for his ability to use top-tier models without cost concerns, calling rationing strategies harmful. Boris Skurikhin of Docket attributes a tenfold productivity increase to AI tools, noting that while tokens are expensive, they remain cheaper than hiring human engineers. Ghai adds that high spending fosters essential AI literacy within product teams. Despite these benefits, not all founders embrace the trend. Rishabh Sambare of Gale avoids high-usage pricing models, preferring subsidized subscriptions from OpenAI and Anthropic despite their limitations, because he cannot afford the variable costs of other tools. Hassan Ismail of Argos Research utilizes affordable monthly subscriptions to ensure his team can access AI tools consistently. Weave's Brennan Lupyrypa openly criticizes the concept, labeling tokenmaxxing as "extremely stupid." While Weave monitors spending and sets notifications at $500 per month to prevent stifling engineers, Lupyrypa argues that spending is a poor proxy for progress. He predicts the trend will collapse within three months as Chief Financial Officers prioritize fiscal responsibility. The conflict between these viewpoints often centers on stage of development. Solberg suggests that founders who resist token spending likely have not yet achieved product-market fit. He argues that venture-backed companies, having signed up for rapid growth, should view high token costs as a necessary investment to fuel that acceleration. As the market matures, the debate continues over whether aggressive token consumption is a strategic advantage or a fleeting, unsustainable fad.