UpTrajectory Review

Five companies—Alphabet, Meta, Microsoft, Amazon and Oracle—have now committed more than $3 trillion in future spending on data centres, equipment, leases and the energy to run them, according to Bloomberg figures cited in this TNW piece. Alphabet alone accounts for over $1T of that total; Meta carries more than $800B. Some of the leases stretch 30 years, which means these firms have locked in obligations that will outlast most of the hardware they will house. The scale dwarfs every previous infrastructure build-out in tech history, and it is happening while AI revenue at most of these companies still runs far behind the capital they have pledged.

For a small-business operator, this build-out cuts two ways. On one side, cheaper and more capable AI tools keep arriving—chatbots, coding assistants, document analysis—because hyperscalers need paying customers to justify the spend. On the other side, the concentration of computing power in five balance sheets means your vendor options narrow. If you build workflows on a tool that one of these giants decides to absorb, deprecate or price into a premium tier, you absorb the disruption. The 30-year leases signal that these companies expect AI demand to persist for decades; they also mean they cannot afford to let the market soften, which raises the odds of aggressive bundling and lock-in.

The genuinely new element here is the lease duration and the energy line item. Earlier coverage of AI capex focused on chips and construction; this report confirms that long-dated real-estate commitments and power procurement now rival silicon as the binding constraint. We are skeptical of one implication the piece leaves unstated: that more infrastructure automatically means lower prices for small-business users. Hyperscalers have raised cloud prices even as they expanded capacity before. The $3T figure is a pledge, not a delivered product, and history suggests delivered cost savings lag promised ones by years.

Second-order effects reach beyond your software bill. Data-centre construction is already squeezing regional power grids and water supplies in places like Northern Virginia and Phoenix, which pushes utility rates higher for every business in those markets. Commercial real-estate markets feel the pull too: long leases convert office parks into server farms, tightening space for growing firms in secondary cities. On the talent side, giants paying premium salaries for AI engineers drain the same pool that small consultancies and SaaS shops draw from, which keeps wages elevated even as layoffs hit other tech sectors.

Watch two numbers over the next four quarters: the ratio of AI revenue to capex at each of these five firms, and the price per unit of compute for the services you actually use. If revenue growth stalls while spending continues, expect the giants to squeeze existing customers harder through bundling, tier changes or API price hikes. Practical steps now: audit which of your tools depend on a single hyperscaler, keep exportable copies of your data and prompts, and test at least one alternative model or provider before a pricing change forces the issue. The build-out will reshape your stack whether you plan for it or not.

“Five tech giants have lined up more than $3T in future spending” — The Next Web

Takeaway: Audit your AI vendors now: map which tools depend on one hyperscaler and keep your data portable before pricing shifts force a costly migration.