Big Tech Cos Fired Thousands For AI, Now Its Adoption Cost is Weighing On Them

From Microsoft and Uber to Meta and Amazon, the AI race is now creating a new problem for Big Tech, soaring adoption bills. Global technology companies are pouring unprecedented amounts of money into artificial intelligence (AI).

For years, AI was sold as a way to cut costs, improve efficiency and reduce dependence on human workers. But recent developments suggest the economics of AI may be far more complicated than expected.

Executives from Nvidia and Uber have warned that the cost of running AI systems and computing infrastructure is becoming extremely expensive in some cases even exceeding employee costs. That reality is now beginning to show up across the tech industry.

Why AI Is Costing More Than Employees

The biggest reason behind rising AI costs is simple, the more companies use AI, the more they have to pay. Unlike traditional software subscriptions, large language models charge businesses based on “tokens”, the small units of text AI systems process and generate. Every AI query, response, coding task or automated workflow consumes tokens, which increases computing costs. In simple terms, AI becomes more expensive as usage grows.

Several companies have actively encouraged employees to use AI tools more aggressively. Amazon reportedly encouraged employees to “tokenmaxx,” or maximise AI usage, while staff at Meta created internal trackers to monitor which teams were using AI tools the most.

The problem is becoming even bigger with the rise of “agentic AI” — advanced AI systems capable of handling multi-step tasks autonomously instead of responding to a single prompt.

According to estimates cited by Goldman Sachs, agentic AI systems could increase token consumption by 24 times by 2030 as companies deploy AI agents at scale across workplaces.

Research firm Gartner expects the cost of running AI models to fall sharply over time. But the firm has also warned that enterprise AI bills may still continue rising because advanced AI agents consume significantly more computing resources per task. As a result, cheaper AI models may not necessarily translate into cheaper AI adoption.

Microsoft And Uber’s AI Reality Check

Some of the clearest signs of the AI cost problem are now emerging inside major technology companies themselves. According to reports from The Verge, Microsoft recently began pulling back on widespread access to Claude Code after employees adopted the AI coding tool at massive scale. The company is reportedly shifting engineers towards GitHub Copilot CLI instead after internal AI usage costs surged faster than expected.

The reversal reportedly came just months after Microsoft had encouraged thousands of employees to experiment with AI-assisted coding tools.

Uber has faced similar challenges. Uber CTO Praveen Neppalli Naga told The Information that the company had exhausted its annual AI coding budget within just four months. Ironically, this happened after the company actively pushed employees to adopt AI tools more aggressively through internal rankings and incentives.

The developments highlight a growing issue for companies: AI tools may improve productivity, but operating them at enterprise scale is becoming extremely expensive.

AI Layoffs Continues

The rising costs are becoming especially important as companies simultaneously cut jobs and increase AI spending.

Meta recently laid off around 8,000 employees globally as it reshaped its workforce around AI priorities. At the same time, the company committed more than $100 billion towards AI infrastructure and future expansion plans.

Amazon has also continued multiple rounds of layoffs across divisions while increasing spending on AI infrastructure and data centres. Reports suggest the company plans to spend nearly $200 billion in capital expenditure, with a major portion going towards AI-related infrastructure.

Meanwhile, Cloudflare announced layoffs affecting more than 1,100 employees as the company restructures itself for what it calls the “agentic AI era.” Yet despite layoffs and restructuring, AI spending continues to rise.

Bryan Catanzaro, vice-president of applied deep learning at Nvidia, summed up the situation in an interview with Axios, “For my team, the cost of compute is far beyond the costs of the employees.” That may be the biggest contradiction in the AI boom.

Companies are reducing headcount partly to fund AI expansion, but the technology itself is turning into a massive recurring expense tied to chips, cloud infrastructure, data centres and computing power. And as businesses deploy more advanced AI agents, total costs could rise even further despite falling token prices.

The scale of spending shows just how serious the AI race has become. According to Fortune, combined capital expenditure by Alphabet, Amazon, Meta and Microsoft crossed $130 billion in a single quarter, largely driven by investments in AI data centres and computing infrastructure. That spending could reportedly exceed $700 billion in 2026 alone, a sharp jump from roughly $410 billion last year.

For Big Tech companies, AI is no longer just a software product. It is becoming an infrastructure race involving massive investments in chips, electricity, cloud systems and data centres.

AI may still replace many workplace tasks in the future. Yet the biggest question for companies today may no longer be whether AI can replace humans, but whether businesses can actually afford the enormous cost of running AI at scale.