A new survey wave shows 73% of organizations now find the cloud more complex than what it replaced, and 31% of cloud spend is wasted. The pattern is not an accident: it is how the dominant providers are built to make money.
The cloud was supposed to make infrastructure disappear. Servers, storage, and networks would become a utility. You would write a credit card number on a form, spin up what you needed, and only pay for what you used.
That was the 2012 pitch. The 2026 reality, according to Flexera's 2026 State of the Cloud survey, is that 73% of organizations now say the cloud has made their operations more complex, not less. And a separate analysis of cloud-spend data from Finout, surfaced in industry coverage, puts the waste at roughly 31% of what companies pay their cloud providers.
That gap between the promise and the bill is not a failure of execution. It is the predictable output of a pricing architecture that rewards the major cloud providers for every additional service a customer turns on, every integration that makes switching harder, and every byte of data a customer pays to move out.
To understand why, it helps to know what kind of company is doing the selling. The dominant public-cloud vendors, Amazon Web Services, Microsoft Azure, and Google Cloud, are commonly called hyperscalers: extremely large operators that run the underlying compute, storage, and networking on which most of the modern internet sits. The hyperscaler business model is built around three revenue primitives. First, consumption: revenue scales with how many resources a customer uses, so every new service a customer adopts is incremental revenue. Second, integration depth: the more of a customer's stack that runs natively inside one provider, the harder and more expensive it becomes to move. Third, egress fees: the charges a customer pays to move their own data back out of a provider's network, a category of cost that has no real parallel in the on-premises world.
Put those three primitives together and you get a system in which the provider's commercial incentive is structurally misaligned with the customer's desire for a clean, simple environment. The hyperscaler does not need you to be happy with a smaller footprint. It needs you to be locked into a larger one. As a TechRadar analysis adapted from a Nexcess CEO column puts it, the major cloud vendors have built a Home Depot of services: you walk in for one thing and leave with a cart, and the cart keeps growing because every aisle makes the next one easier to justify.
The 2012 cloud pitch was not dishonest. At the time it solved real problems. Buying physical servers meant long procurement cycles, sunk capital, and utilization rates that fluctuated between feast and famine. Cloud replaced that with elastic capacity, no upfront capital, and a meter that only ran when a customer used something. For a generation of engineering teams, it felt like trading a fixed-cost mortgage for a pay-as-you-go utility bill. The bill was, in fact, simpler in one important way: there was no basement full of hardware to maintain.
What replaced the basement was something else. Cloud did not just move the work of running infrastructure from the customer's data center to the provider's data center. It moved the management surface upward into pricing tiers, identity systems, networking topologies, and a steadily growing catalog of managed services that each come with their own knobs. A modern cloud account does not look like a utility account. It looks like a sprawling configuration problem that the customer owns. The 2026 Cloud Security Report from Cybersecurity Insiders describes this state as a complexity gap: a widening distance between what cloud platforms offer and what typical security and operations teams can realistically govern.
This is the part that is genuinely new, and it is worth naming. The old complexity was physical. You had to buy hardware, rack it, power it, cool it, and replace it. The new complexity is contractual and architectural. You have to decide which of a thousand services you will use, how they connect, who has permission to touch them, what the data leaves the building for, and what it costs to bring it back. The customer's environment does not shrink with consolidation. It grows because the provider's catalog grows, and the customer's adoption of new services tends to outpace the customer's ability to retire old ones.
That dynamic shows up in the spending data. Survey-based coverage of the 2026 cloud market consistently reports that organizations are moving toward multi-cloud architectures as a way to escape single-vendor dependency, a pattern that a Flexera summary circulated by Tech Channels treats as the industry's current workaround for the very problem hyperscalers helped create. Multi-cloud is often presented as a strategic choice. In practice, the TechRadar analysis argues, it is more often a defensive response: customers spread their workloads because concentrating them inside one provider has become too expensive or too sticky to tolerate.
This is also where the limits of the simple "cloud is a ripoff" story become obvious. The structural incentives of the hyperscaler model do produce sprawl, lock-in, and a complex management surface. They also produce a level of on-demand infrastructure that did not exist before, and that has real economic value for many buyers. A ProsperOps analysis of the Flexera data treats the 31% waste figure not as evidence that cloud is broken but as evidence that FinOps, the practice of managing cloud spending with financial discipline, is now a permanent part of how large organizations buy infrastructure. The honest read of the 2026 data is that cloud did deliver what it promised, and it also delivered a new kind of complexity that the customer now owns.
What does that mean for someone who is not a cloud architect? Three things are worth carrying into the next vendor conversation. First, complexity in a cloud bill is not a sign that the buyer's team did something wrong. It is the default output of the pricing model. Second, the most expensive line items are usually the ones a buyer did not realize were negotiable: egress fees, data transfer between regions, idle managed services left running after a project ends. Third, the question worth asking a provider is not "what does this cost per hour" but "what does it cost to leave." That single number tells a buyer more about the real shape of the deal than any price list.
The cloud was supposed to make infrastructure invisible. In 2026, according to the latest survey wave, the infrastructure is still there. It just moved from the customer's basement to the customer's spreadsheet, and the meter keeps running whether the customer is looking at it or not.