Recently, a new term has sparked widespread discussion: 'Tokenpocalypse'.
The trigger was Microsoft's pricing restructure of GitHub Copilot. Starting from June 1, Copilot has fully transitioned to a token-based billing model, with a significant multiplier difference in token costs between different models, where the price of a single token for some models is 60 times that of other models.
Ironically, the highly praised 'truly useful' advanced models are the ones that have seen the most drastic price hikes.
As companies like Anthropic, OpenAI, and other top AI companies prepare to go public, AI companies will face even more severe profit pressures, perhaps forcing more vendors to follow suit in raising prices.
The cost of using AI is always an inevitable issue for enterprise productivity expansion. The recent trend of 'tokenmaxxing,' which competed on employee token usage, will come to an end as the Tokenpocalypse approaches.
"The entire trend of tokenmaxxxing, from rise to peak and then to being despised, only lasted six months."
The Dilemma for Enterprises
A developer from a large enterprise described an absurd predicament: the company has long mandated that employees use AI tools, where using too few tokens will result in a reprimand. However, with the new pricing, using too many tokens will also lead to a reprimand.
What's even more disastrous is that the Copilot team has not yet been able to launch the 'employee-level token limit' feature. This means that under the new billing model, an employee could burn through the entire company's monthly token budget in a single day.
"My job is no longer about solving business problems with software," the developer wrote, "my job has turned into solving token usage issues."
The comments in the section were even more entertaining. A user summed it up, saying, "The company policy has become: 'Use AI for everything, but be careful not to use too much because if LLM consumes too many tokens, you will be disabled, and then you will be criticized for not using AI for the rest of the month.'
Overemphasizing AI Productivity in the enterprise may also be a double-edged sword.
An information officer from a large law firm even "bragged" at an AI workshop: after their AI system went down, the lawyers basically stopped working because they were too dependent on AI.
Someone who has received years of professional training actually freely admitted that they couldn't work without an AI chatbot? I would be ashamed enough to start reflecting on my career path."
Uber Overspending Event: Industry Microcosm
Nowadays, most AI models come with a usage package, and the budget control issue for enterprises has become more serious as the industry moves towards consumption-based billing with Tokens.
Uber completed a full cycle in just a month and a half: first, they discovered that "the AI budget burned much faster than expected," and then they urgently set usage limits and employee restrictions.
"Imagine if companies that heavily use AI like Uber hit a wall so quickly," discussed in TechCrunch's podcast, "the question is: can the AI lab align costs with customer willingness to pay?"
A fun fact: When ChatGPT Plus was initially priced at $20 per month, there was no strategic consideration, "they just threw out a number." The entire industry is still paying for this starting point.
"Your job may not be replaced by AI, but your budget might be."
There are even more intriguing details on Reddit. Someone in a company built an AWS Bedrock cost monitoring dashboard, showing the real-time cost of each model and each token (including cache tokens) on CloudWatch, "making developers and finance watch the money burn together." The reaction in the comments section was: "Congratulations, you just provided them with a new KPI metric."
Another major company has already faced a similar tightening: after the AI quota was used up, everyone was downgraded to GPT-4.2, losing even the VSCode integration.
An observer from outside the tech industry voiced what many others were thinking: "The amount of mental energy and actual working hours consumed by this whole thing has become so much that it's affecting the actual work delivery that can help the company make money."
While the entire industry was still immersed in the narrative of "AI will replace everything," a more realistic issue has surfaced: the bill for computational power, which someone will ultimately have to pay. And the "Token Doomsday" may just be the beginning of this reckoning.
Welcome to join the official BlockBeats community:
Telegram Subscription Group: https://t.me/theblockbeats
Telegram Discussion Group: https://t.me/BlockBeats_App
Official Twitter Account: https://twitter.com/BlockBeatsAsia
