Why hyperscaler CapEx is your liability

The cloud story has always been sold as “variable cost, infinite scale.” What gets overlooked is who (actually) carries the risk and how it affects your spend?

In 2025, hyperscalers are spending at a pace that breaks every record. The US alone is tracking $40 billion in annualised data center construction. McKinsey projects nearly $7 trillion in global data center investment by 2030, with over 40% in the US.

These billions don’t just sit on hyperscaler balance sheets.
They have (already) been flowing downstream.
The way hyperscalers de-risk their CapEx is simple: by selling you commitments.

Hyperscalers pour billions into GPUs, networks, and power at the top of the chain. Gravity does the rest. Finance teams carry it at the bottom through Savings Plans, Reserved Instances, enterprise contracts, or PPAs.

It’s simple gravity. Their CapEx = your liability.

Why now: the screws are tightening

The rules are shifting against you.

CapEx is ofc accelerating too. Hyperscalers cannot dig foundations and pour concrete for their new data centres quick enough. Each new build requires new reasons to lock customers into longer, larger contracts. Their certainty, their investors’ certainty, comes by converting your spend into their hedge.

Boards too are feeling this pain, as a result of data centre spend. Cloud costs used to live inside procurement, now they are a fixture within board packs. Investors are now asking: how are commitments laddered? What’s the downside if demand misses? This is not about missing a discount here and there. It’s about exposing a new governance gap.

This governance gap

Commitments are marked as “savings.” In 2025, they behave as long-term liabilities.

Think about the office leases signed during Covid boom. Those weren’t inked by CTOs because engineers needed more desks. They were signed by the CFOs. Because leases are financial obligations. Cloud commitments (now) (must) fall into the same category.

They mature. They expire. They cliff.
Miss your forecast and those supposed “savings” turn into stranded costs.

With commitments now closer to debt than discounts, surely now they should be:

• stress-tested against multiple demand scenarios
• laddered to avoid maturity cliffs
• tracked on your balance sheet (ie not buried inside procurement reports)

(For hyperscalers, CapEx is a capital structure decision) + (For CFOs, commitments framed as expense) = that mismatch is the governance gap.

T’s tl;dr

Cloud commitments are not savings. They are liabilities in sheep’s clothing. Manage these accordingly.

The billions have already been spent at the top of the food chain. The only remaining Q is whether you manage the downstream obligations with the same discipline as you’d apply to debt. That means:

• quarterly reviews of utilisation vs forecast
• laddering maturities to avoid cliffs
• recognising commitments as financial instruments, not as discounts.

Boards and investors are already treating them this way.
The Founders and CFOs who govern accordingly will protect margins (and credibility).

Never forget…hyperscaler CapEx does not stay with them…
…it always rolls downhill.