When ChatGPT burst onto the scene in late 2022, corporate boards around the globe saw a gold rush. Artificial intelligence promised to handle everything from writing code to managing customer service for a fraction of the cost of human teams. The immediate result? A wave of sweeping layoffs, team downsizings, and restructuring across the tech industry.
But as we cross into 2026, the financial maths behind the AI revolution is beginning to fall apart. Major enterprises, including giants like Microsoft, Uber, and Meta, are quietly pulling back on their aggressive AI plans.
The reason is simple: AI is proving to be incredibly expensive to scale, and the bills are coming due.
The rise of "token maxing"
How did a technology marketed as cheap and efficient become a budget-killer? It started with corporate pressure.
Starting in 2023, tech firms aggressively incentivised—and sometimes forced—their developers to integrate AI into every aspect of their work. To measure usage, companies tracked “tokens” (the basic units of data, roughly equal to 750 words per 1,000 tokens, used by AI models to process and generate content).
This created a highly counterproductive cultural phenomenon in Silicon Valley: token maxing.
To prove they were driving innovation and meeting KPIs, engineers began consuming tokens excessively for minor, insignificant tasks. In some cases, tech leaders publicly stated they would be “deeply alarmed” if highly paid software engineers weren’t consuming hundreds of thousands of dollars in tokens annually. This artificially inflated demand and created massive, unbudgeted expenses.
A steep rise in token prices
As demand skyrocketed, the infrastructure couldn’t keep up. Severe hardware shortages and delayed data centre construction projects squeezed the supply of raw computing power.
Consequently, token pricing spiked:
- December 2025: The average cost was $1.11 per million tokens.
- May 2026: The average cost surged to $2.12 per million tokens.
While a few dollars per million tokens sounds nominal, the sheer volume of enterprise usage multiplies these figures exponentially. For example, a company with 40,000 employees using just $200 worth of tokens weekly racks up an annual bill of $400 million. Tech giant Meta reportedly consumed 60 trillion tokens in a single month, translating to roughly $900 million in API costs.
The bottom line: Technology budgets are being stretched to their absolute limits. A Gartner survey revealed that three-quarters of executives expect tech budgets to increase, with AI projected to represent over 20% of total enterprise technology spending by 2035.
The comparison: AI vs human labour
We are rapidly approaching a threshold where human labour is simply more cost-effective than paying for AI tokens.
Currently, running highly advanced AI agents is comparable to the cost of human labour, but the scale is tilting. While AI remains cheaper for highly specialised, repetitive coding tasks, human workers are already proving to be more cost-effective in areas like data entry and customer call centres.
Furthermore, AI companies like OpenAI and Anthropic have historically subsidised token costs using venture capital funding. As these platforms prepare for public listings and face mounting investor pressure to turn a profit, they will inevitably have to raise token prices even higher, making humans the financially smarter choice for many business functions.
What this means for women in tech
For women working in software engineering, product management, and data science, this economic correction brings a mix of caution and opportunity:
- The return of the human premium: As companies realise they cannot completely substitute their workforce with costly AI agents, human-centric skills, including system architecture, complex problem-solving, and emotional intelligence, will see a major resurgence in value.
- A shift in technical evaluation: The era of “token maxing” is ending. Engineering leaders will transition from evaluating how much AI tools are used to how efficiently they are used. Writing optimised, clean code that minimises token consumption will become a highly sought-after engineering skill.
- Strategic budgeting roles: As corporate finance departments grapple with massive AI budget overruns, tech leaders who understand how to balance infrastructure costs, cloud workloads, and human capital will be highly valued in executive leadership.
The initial hype of the AI boom is cooling down, replaced by hard financial realities. As tech firms pivot from “AI at all costs” to sustainable, cost-effective growth, the industry’s most valuable asset remains exactly what it has always been: skilled human talent.
