In June, Apple raised prices across almost its entire Mac and iPad range, with some of its most popular models costing twenty percent more overnight. The reason the company gave was telling: the rapid expansion of AI data centres has created such a surge in demand for memory and storage that component prices have risen faster than it has ever seen. I read that and stopped, because nobody buying a laptop signed up to fund a data centre. That is when a pattern became impossible to ignore.
The true cost of the AI revolution doesn’t appear on any invoice we sign. It is embedded in our electricity bills, our software renewals, and now the price of the machine on our desk. We are paying an AI tax, and most of us never noticed it being levied.
Let me be clear: this is not an argument against AI.
At Marching Sheep, we use it every day to streamline our internal operations, and we have used it to productise our services so our work reaches more organisations. I am a believer in what this technology can do. But believing in it doesn’t mean looking away from who is quietly carrying its cost.
Where the money is actually going
The scale of investment behind this boom is unlike anything the corporate world has seen. The largest technology companies are expected to spend over six hundred billion dollars on AI infrastructure in 2026 alone, an increase of more than sixty percent over last year’s record. Closer home, cumulative data centre investment commitments in India have crossed 126 billion dollars.
Now hold that against an uncomfortable question: who is earning this money back? Largely, nobody yet. OpenAI is projecting a fourteen billion dollar loss for 2026 and does not expect profitability before the end of the decade. xAI is reported to be burning close to a billion dollars a month, losing many times what it earns. Anthropic, the maker of Claude, has just projected its first quarterly operating profit, and it makes headlines precisely because it is the exception in an industry where analysts expect most AI startups to fail.
When an industry spends this much and earns this little, the gap doesn’t disappear. It moves.
How the bill quietly reaches the rest of us
- The first place it lands is the electricity grid. Data centres consume enormous amounts of power, and utilities are spending billions on new transmission and generation to keep up. In the United States, residential electricity costs have risen by over forty percent in five years, and what should truly give us pause is that large data centres often negotiate preferential rates while households absorb the increases. India’s data centre power demand is projected to grow nearly five times by 2030, and unless we plan deliberately, the same pattern will play out here.
- The second place is our technology itself. Microsoft raised its Microsoft 365 pricing when it bundled in Copilot. Analysts expect consumer AI subscriptions to climb well beyond today’s twenty dollar norm, because current prices were designed to win users, not to sustain a business. Apple’s June increase shows the third route: when a company won’t charge a visible AI fee, the cost simply travels into the hardware price.
When infrastructure spending runs this far ahead of revenue, the difference is recovered from the quietest payer in the room: the consumer who never agreed to fund it.
This is what makes it a tax. It is systemic, it is invisible, and it was never put to a vote.
What leaders owe their people and their customers
In my work with organisations across sectors, I have watched AI budgets get approved with a fervour that would never survive scrutiny in any other capital decision. So here is what responsible leadership looks like right now, in practice.
- First, interrogate AI spend the way you would a factory or an acquisition. If you are buying two thousand Copilot licences, name the workflows they will change, baseline the hours those workflows consume today, and review actual usage data at ninety days before you renew. I have seen organisations where only a fraction of licences are actively used while renewals sail through unquestioned.
- Second, be honest about pass-through. If AI-related costs are raising the price of your product or service, say so plainly, the way Apple was at least candid about its reasons. Customers forgive honest pricing far more readily than they forgive discovering a hidden surcharge.
- Third, invest in your people as seriously as in the technology, because a licence without capability building is pure cost. Before the next tranche of spend, train managers on where the tools fit their team’s actual work, redesign the workflow rather than bolting the tool onto the old one, and make one named leader accountable for converting the spend into outcomes.
The AI revolution may well prove worth its price. But a cost this large, carried this quietly, by people who never chose it, deserves to be seen.