A central AI monitoring and spending control system is currently being developed at Meta after the company realized it was spending more internally on AI than expected. The decision shows companies are considering whether the returns from AI justify the costs involved.
The company sent a memo to about 6,000 employees detailing its AI spending cap plans, budgets and symbolic limits. Under the AI portal, teams will have access to an overview of AI usage that will automatically send notifications if there are unusual increases in spending. Token management is expected to be fully implemented by 2027.
The memo noted that Meta is seeing rapid growth in internal AI adoption and that it will likely spend tens of billions on using AI for employees in 2026.
The subsequent effect of tokenmaxxing
Meta’s shift in focus from promoting the use of AI to controlling its use shows a recurring theme within corporate America. The company used to motivate its employees to use AI by making them grow Internal leaderboards (“Claudionomics,” named after the Anthropic AI system). Meta no longer runs this particular leaderboard.
The broader trend has a name: “tokenmaxxing”, which is the practice of using maximum… Artificial intelligence codes Possible for any reason, whether to inflate internal dependency metrics or simply to consume them. The same situation occurred at Amazon after its employees created a leaderboard to track token usage, but the company later removed it in late May over concerns that it was leading to wasteful spending. Business Insider reports.
Uber’s experience shows how quickly costs can rise. The ride-hailing company exhausted its entire planned AI development budget for 2026 by April, just four months into the year. Andrew McDonald, Uber’s chief operating officer, told Rapid Response that the company has struggled to link symbolic spending to measurable production. “This link doesn’t exist yet, does it?” MacDonald said. “It’s very difficult to draw a line between one of these statistics and saying, ‘Okay, now we’re producing 25% more useful consumer features.’”
The cost problem that the industry has not solved
And the budget pressures go far beyond Silicon Valley. According to a survey conducted by KPMG First reported by The Wall Street Journalonly 26% of companies have comprehensive visibility into their AI costs, while 50% have partial visibility and 22% either have no visibility or discover spending only after receiving invoices. As pointed out by Steve Chase, Global AI Leader at KPMG, the company is reportedly helping clients who have already exhausted their annual token or cloud computing budgets in a matter of months.
Microsoft recently pulled out Fortune reported that it had nearly all direct licenses for Claude Code and redirected engineers to its GitHub Copilot CLI, just six months after making the tool humanly available to its employees. The move came after employee utilization increased faster than expected.
Economic considerations suggest that initial expectations of rapid profitability of AI due to labor savings were overly optimistic. NVIDIA Vice President of Applied Deep Learning, Bryan Catanzaro, She revealed to Axios The cost of computing for his group already exceeds the cost of hiring people. Goldman Sachs believes Proxy AI could lead to a 24-fold increase in token consumption by 2030, with monthly consumption rates reaching 120 quadrillion tokens per month, even as per-unit token prices decline.
Furthermore, Gartner predicts that lower token costs will not mean cheaper enterprise AI implementations because agent AI algorithms use a much higher number of tokens per task, while service providers will likely keep the overall savings for their part. “Chief product officers should not confuse the deflation of commodity tokens with the democratization of frontier logic,” said Will Sommer, senior analyst at Gartner. Earlier, Cryptopolitan I mentioned Zuckerberg admitted that Meta made “mistakes” in its shift to artificial intelligence
What can Meta employees expect?
according to According to reports, the memo revealed that Meta will discourage its employees from using external software to write AI code and encourage them to use its own assistant, MetaCode, which was previously called Devmate. These changes will be implemented in the following weeks.
Meanwhile, Meta’s efforts to reduce AI-related costs come with significant regulatory changes. In March of this year, Meta was considering layoffs involving at least 20% of the total number of workers of about 79,000 workers, in part due to investments in artificial intelligence infrastructure worth about $600 billion through 2028.
Sam Altman, CEO of OpenAI, highlighted this industry challenge well. “This is the fairest criticism of AI right now,” he stated, saying, “You hear companies say, ‘I’m spending a lot of money on AI.’ I know some great things are happening, but I know there’s a lot of waste.”
For the global economy, the question is whether corporate AI budgets shrink before the technology delivers on its productivity promises, or whether lower nominal prices and better tools will fill the gap first. The answer will shape hiring, capital spending, and competitive dynamics across industries for many years.
Don’t just read cryptocurrency news. Understand that. Subscribe to our newsletter. It’s free.
