Cloud capacity is becoming scarce, CIOs need to plan now

23.06.2026

7 min read

63 percent. That’s how fast Google’s cloud business grew year-over-year-outpacing Microsoft (39 percent) and Amazon (28 percent). The order books of the three hyperscalers are filling faster than new data centers can come online. For CIOs, that means the cloud capacity once taken for granted now requires advance planning and safeguards.

Key Takeaways

Backlogs are exploding: Hyperscalers’ contractually committed order backlogs are in the triple-digit billions and rising at double-digit rates each quarter, driven by long-term AI contracts.
Availability becomes the bottleneck: Where demand spikes hardest, peak periods bring longer lead times and narrower pricing flexibility-especially for compute-heavy workloads.
Planning beats spontaneity: Booking capacity on demand now costs more and takes longer. Reservations and forecasts belong in the budget cycle.

When capacity becomes a bargaining chip

A cloud leader is drafting next year’s budget. Until recently, the exercise boiled down to cost optimization: cheaper instance types, a few Reserved Instances, some FinOps discipline. This year, a question that no one would have asked two years ago has been added to the mix: will the capacity I need next quarter even be available?

It’s not a hypothetical. If a hyperscaler has already sold the bulk of its new compute capacity via multi-year AI contracts before the data center is even built, normal enterprise workloads compete with training runs that tie up multiples of the GPUs. The cloud market is shifting from a self-service shelf to a business with lead times.

For an enterprise cloud strategy, this breaks a long-held assumption. Elasticity-the promise of on-demand extra compute-remains central, but now it comes with conditions.

Why order books are exploding right now

The driver is spelled out in the quarterly reports. Hyperscalers list their contractually committed but not yet realized revenues as a key metric. This is the order book of the cloud, and it’s growing at breakneck speed.

What is a Remaining Performance Obligation? RPO refers to revenues already contractually promised but not yet recognized as income. In cloud business, this primarily means long-term service revenue from multi-year contracts. A rising RPO shows how much future revenue is already locked in, even if it hasn’t been invoiced and isn’t one-to-one tied to physical compute capacity.

At all three providers, this line item has recently surged, fueled by multi-year AI commitments from sectors such as finance and pharma. Exact figures vary widely depending on how the metric is defined, but the trend is unmistakable across every report: a growing share of future cloud revenue is contractually secured long before the corresponding compute power is even available.

+63 %

Google Cloud’s year-over-year growth most recently, the highest among the big three providers.

Source: Alphabet, Q1 2026 earnings report

The growth is good news for the market-and uncomfortable for planning. Where demand surges fastest, the risk rises that a spike will hit a ceiling. Which provider hits that ceiling and when is nearly impossible to predict from the outside, because quarterly numbers show demand, not free capacity. If you’re running mission-critical workloads on a single provider, you should bake that risk into your architecture.

What CIOs Should Factor Into Their Planning Now

Those who integrate this early into the budget cycle negotiate from a stronger position. Computing power becomes a resource to manage like a supply chain: with lead times, alternatives, and negotiated terms. Three moves pay off in the short term.

First, forecast your own demand further into the future than before and translate it into early commitments. Locking in capacity a year ahead secures better prices and more reliable availability than booking in the quarter of use. Second, maintain a second provider or an in-house reserve for truly critical workloads so a bottleneck at a hyperscaler doesn’t disrupt daily operations. Third, make availability a contractual issue, with guaranteed lead times and escalation paths. Price is only one lever among many.

The trickiest of these steps is the first-and it’s not just an IT issue. A one-year-ahead commitment ties up budget, so the CFO joins the conversation. That’s where the rubber meets the road: whether insight turns into a plan. Presenting a solid demand forecast alongside a comparison of commitment discounts versus spot-price risk gets the green light. Asking only for more cloud budget invites pushback. For a CIO, this is less a technical task and more a negotiation challenge, and it’s easier as long as reserves haven’t tightened in the market yet.

Capacity Roadmap for the Next Quarters

Now

Model demand for critical workloads through 2027 and check availability with the provider.

Next quarter

Negotiate capacity commitments, lock lead times and escalation paths in the contract.

2027

Keep a second provider or an in-house reserve ready for business-critical workloads.

Bottom line: mindsets are shifting. Cloud once promised never having to worry about compute again. In a market that’s already sold its next few years, proactive planning becomes a competitive edge-especially for the workloads that keep the business running.

Frequently Asked Questions

What does a high cloud backlog mean for my business?

A large order backlog indicates that a significant portion of future capacity is already committed. For sudden, large-scale demands, this can mean longer lead times and less pricing flexibility, particularly for compute-intensive workloads.

Does this mean cloud elasticity is dead?

Not at all, but it now comes with conditions. Scalability remains available for standard workloads. For very large or GPU-heavy demands, it pays to secure capacity in advance rather than relying on on-demand availability.

Which provider is most affected by shortages?

Generally, wherever demand is strongest. However, it’s nearly impossible to pinpoint from the outside, as quarterly figures reflect demand rather than available spare capacity. Companies running critical workloads on a single provider should manage this risk independently of growth rankings.

How can I reliably secure capacity?

By making early commitments with longer lead times, securing contractually guaranteed availability windows, and maintaining a secondary provider or an in-house reserve for business-critical workloads.

Is a multi-cloud strategy worth it just for capacity reasons?

For critical workloads, often yes, since a bottleneck at one provider won’t automatically disrupt daily operations. However, multi-cloud introduces complexity. It makes sense as a targeted approach for key workloads, but rarely as a blanket solution for every application.