Microsoft's decision to overhaul GitHub Copilot pricing has sparked a term among Reddit users: Tokenpocalypse. The changes, which shift from a flat-rate subscription to per-token charges, are seen as a harbinger of broader cost pressures in the AI industry.
On a recent episode of TechCrunch's Equity podcast, hosts Kirsten Korosec, Sean O'Kane, and Anthony Ha dissected what these pricing moves mean for the AI ecosystem. With Anthropic and other major AI labs planning to go public, the trio explored how profitability questions are forcing companies to pass on costs to customers and impose usage limits.
The Tokenpocalypse Discussion
Anthony Ha opened the conversation by noting that Sean O'Kane had coined the term Tokenpocalypse for the segment. The GitHub Copilot price hike, moving from a flat monthly fee to a per-token model, is just one example of a broader shift. Ha pointed out that the entire ecosystem has been heavily subsidized by investor money, making many AI products appear cost-free when they are actually expensive to operate. Now, those costs are being transferred to end users and businesses, a change that will likely cause significant pain.
Sean O'Kane questioned whether AI companies can collapse their costs fast enough to match customer demand. He recalled that when ChatGPT first debuted, a $20 monthly fee seemed arbitrary, with no real strategy behind it. While customers pay more for advanced models, that still does not cover the true cost of running them, leaving a gap that must be closed.
Tokenmaxxxing's Rise and Fall
Kirsten Korosec highlighted how quickly the landscape has shifted. She noted that the term "tokenmaxxxing" emerged, peaked, and is now viewed negatively all within six months. The pricing mechanisms for AI tools were established before business models were properly solidified around these labs. This rapid evolution makes it difficult for companies to plan long term, especially when writing risk factors for IPO filings.
Stay updated
Get the day's AI and automation news in your inbox. No spam, unsubscribe anytime.
O'Kane added that Uber provides a compelling case study. In just a month and a half, Uber went from claiming that they blew through their AI budget faster than expected to imposing caps on employee usage. If a large, AI-heavy company like Uber is struggling to control costs, he argued, that raises questions about whether AI labs can ever bring expenses down to meet customer appetite.
Uber's Arc as a Cautionary Tale
Anthony Ha drew a parallel to Uber's well-known journey from unprofitability to profit. He noted that Uber had to fundamentally transform itself, expanding into new business areas and squeezing both drivers and customers to reach scale. AI companies may need to go through similar transformations if they hope to survive. Sean O'Kane wondered whether AI labs have any "squishy" cost centers to squeeze, like Uber did with drivers, but acknowledged that the costs are more straightforward and harder to reduce.
IPO Risk Factors and Government Regulation
Kirsten Korosec stressed that the rapid pace of change makes it challenging for companies to draft their IPO registration statements. She pointed to President Trump's recent executive order on AI, which gives the government a chance to review powerful AI models. That adds another layer of uncertainty. How do you even write risk factors, she asked, when the risks evolve day by day?
The conversation concluded with a sense of uncertainty about the future. AI companies are moving toward IPOs while grappling with cost overruns, changing business models, and shifting regulatory landscapes. The Tokenpocalypse may be just the beginning of a longer adjustment period.