Meta announced 8,000 job cuts on the same earnings call where it committed $135 billion to data centre expansion over the next three years. The juxtaposition was not accidental. At roughly $17 million per dismissed employee in fresh infrastructure spending, the company is executing the most explicit trade in corporate history: human capital for compute capital. This is not cost-cutting dressed as transformation. It is transformation funded by cost-cutting. Intel's stock jumped 20% last week after reporting that its data centre revenue had grown 89% year-on-year, driven entirely by AI workloads. Nvidia's H100 chips now command $30,000 each and six-month waiting lists. Every major tech platform is making the same calculation: fire the expensive humans, buy the scarce silicon, and let software do what salaries used to accomplish. The labour-to-compute exchange rate has never been more explicit, or more permanent.
The pattern extends beyond Meta's headline-grabbing cuts. Coinbase shed 20% of its workforce in January while announcing a $2.3 billion AI infrastructure investment. The crypto exchange framed both moves as responses to market conditions, but the numbers tell a different story. Those 1,100 redundancies, at an average fully-loaded cost of $180,000 per employee, freed up $198 million annually. The AI spend is a three-year commitment, meaning Coinbase is trading two years of human wages for a decade of compute capacity. The math is deliberate. Similar trades are happening across the sector, though few companies state them as bluntly. Thomson Reuters shares fell 12% despite strong quarterly results after Anthropic unveiled AI agents designed to handle financial analysis tasks. The market grasped immediately what the company's management couldn't say: software that writes research reports makes research analysts redundant. Sterling Infrastructure, which builds data centres, hit an all-time high the same week. Capital is flowing toward the infrastructure that replaces workers, not the companies that employ them. , - The economics driving this shift are structural, not cyclical. A software engineer costs $200,000 per year in total compensation. A comparable AI model costs $50,000 to train and $2,000 per month to run. Over three years, the human costs $600,000, the model costs $122,000. But the AI works 24/7, doesn't take holidays, and improves with each iteration. The productivity multiple is stark enough that companies are treating it as inevitable, not optional. Briefed Intelligence data shows UK consumer pressure index climbing 8.9 points over four periods to 58.2, indicating sustained household stress. Tech companies are reading this as validation of their strategy: consumers can't afford higher prices, so costs must come from operations, not revenue. Labour is the largest operational expense they control. GPUs are the investment that makes labour optional. , - The financial engineering behind these trades is becoming more sophisticated. Anthropic's surprise compute deal with SpaceX illustrates the point. Musk's rocket company had excess data centre capacity from planned Mars mission simulations. Rather than let it sit idle before a potential IPO, SpaceX is leasing 100% of the compute power to Anthropic for AI training. The deal reportedly values the unused capacity at $4.2 billion, transforming what would have been sunk infrastructure costs into revenue that boosts pre-IPO valuation. This creates a feedback loop: companies that cut staff can afford more compute, which makes them more efficient, which justifies further staff cuts. Adani Group's energy units have become AI infrastructure plays purely because they power the data centres that replace human workers. The group's green energy division trades at 34 times earnings, pricing in a future where every kilowatt-hour feeds a GPU, not an office. , - The human cost of this trade is becoming visible in hiring data. UK tech sector job postings have fallen 31% year-on-year, while data centre construction jobs have risen 89%. The displacement is direct: for every AI researcher hired, three software engineers lose roles to the models they helped create. The sector is not automating inefficiencies, it is automating expertise. DeepSeek's $45 billion valuation from its first funding round exemplifies the inversion. The Chinese AI lab achieved breakthrough performance using a fraction of the compute power that Western rivals required. Rather than hiring more engineers, investors are betting on models that make engineers unnecessary. The premium is for replacement, not augmentation. , - This exchange rate will define the next phase of tech industry consolidation. Companies with the deepest pockets can afford the most expensive GPUs and the largest layoffs, creating competitive moats that smaller rivals cannot match. Meta's $135 billion infrastructure commitment is not just a bet on AI, it is a bet that human-dependent competitors will be priced out of the market. The question is not whether this trade will continue, but whether it will accelerate. Intel's data centre revenue surge suggests the answer. Every GPU sold represents multiple jobs that will not return. Every data centre built houses the replacement for work that humans used to do. The exchange rate between jobs and GPUs is no longer implicit. It is the central mechanism of how technology companies now create value.