Companies Abandon AI Layoffs as AI Systems Cost More Than Employees

Mass employee replacement with AI, which corporate leaders promised, is running into an unexpected obstacle: maintaining artificial intelligence is becoming more expensive faster than economic benefits are growing. Fortune broke down the math of this scenario.

AI Prices Skyrocketed in the Past Year

The cost of accessing powerful AI models is growing exponentially. The APIs of OpenAI, Google Gemini, Anthropic Claude, and Mistral have all significantly raised prices over the past year. If a company makes hundreds or thousands of API requests per day, monthly bills can grow from hundreds of dollars to tens of thousands. On top of this come licenses for specialized platforms, models trained on proprietary corporate data, and dedicated cloud computing resources. Analysts have noted that startups in the AI race often spend more on APIs than on team salaries. Maintainers of Anthropic and OpenAI models have openly raised prices for computing, correctly calculating that demand will remain high despite the cost.

• The cost of GPT-4 API calls increased by 40–50 percent over the year
• Corporate plans require guaranteed capacity and private model instances
• Maintaining a local LLM requires expensive GPU infrastructure
• Salaries for AI integration specialists are twice as high as for professional developers

Corporate Math No Longer Adds Up

Major corporations (Goldman Sachs, Morgan Stanley, Microsoft) conducted internal calculations and were surprised. The annual cost of deploying one production AI system can reach $50,000–$150,000 per year—this includes APIs, infrastructure, maintenance, model updates, and the person overseeing it. For comparison: the average salary of an office employee (analyst, manager, assistant) in the US is $45,000–$80,000 per year. The result: firing a person and fully replacing them with AI costs more than keeping the person and providing them with quality tools.

The problem is further compounded by work quality. AI excels only at narrow, repetitive tasks. For everything else, it requires human oversight, editing of results, and verification against factual errors and hallucinations. The cost of this oversight often adds 20–30 percent to the total cost of an AI system.

People + Powerful Tools — The New Strategy

The result was a strategic recalculation. Major corporations (Goldman Sachs, McKinsey, Morgan Stanley) abandoned the idea of full replacement and are instead investing in enabling employees to work more effectively with AI assistants. They're hiring prompt engineering trainers, purchasing licenses for user-friendly interfaces (GitHub Copilot, ChatGPT Enterprise, Claude for Teams, specialized internal platforms), and training their teams. According to internal measurements: one analyst with the help of modern AI can do work that previously required two people.

But this is not replacement—it's multiplication of competence. The employee stays, the salary remains, but their actual productivity increases by 20–40 percent. Meanwhile, the cost per hour of employee work decreases.

What This Means

AI has finally ceased to be an existential threat to mass employment and has instead become an ordinary—but expensive—corporate tool, much like computers and the internet once were. For workers, this means instead of a wave of layoffs, there will be a wave of retraining and reskilling. For business, it means a shift from sci-fi fantasies of full automation to sober reality: productivity grows through a cost-effective symbiosis of human and machine, not through replacing one with the other.