Public Cloud Platforms

One rung, and now an order inside it

The previous edition of this Wave put four vendors on the Leaders rung and treated them as a set: AWS, Alibaba Cloud, Google Cloud and Microsoft, four names sharing a tier. The 2026 edition still has Google Cloud and Microsoft on that rung, and it has something the 2024 edition did not, which is a scoring order inside it.

Google Cloud's own account of the result is the specific one. The company states that it received the highest overall score of any cloud provider, the highest score in the current offering category, and the highest possible score in twenty-three of the report's thirty criteria.

A Leader placement tells a buyer which tier a vendor occupies. Twenty-three out of thirty tells a buyer how far into that tier it sits, and how much of the evaluation it swept. For a market where the top tier has been stable for years, the second number is the one that changed.

The Forrester Wave: Public Cloud Platforms, Q3 2026, ten providers and thirty criteria

The report examined ten public cloud providers against thirty criteria. The 2024 global edition examined nine.

The tenth entrant matters more than the arithmetic suggests. The 2024 field's lower half was Tencent Cloud, Oracle, IBM, Huawei Cloud and OVHcloud, a group whose shared characteristic was geography or sovereignty rather than compute economics. The 2026 field includes Vultr, a privately owned provider with thirty-three regional data centers, which Forrester placed as a Strong Performer.

Forrester's rationale, quoted in Vultr's own announcement, describes a market "challenged by specialist GPU-focused neoclouds" and places Vultr's vision between the two models, betting that a mix of commodity offerings and GPUs can compete with the hyperscalers on both price and performance for AI workloads. Rather than matching the larger providers' proprietary services, Vultr differentiates through the open-source Kubernetes ecosystem and open-source AI projects, with NVIDIA and AMD partnerships underneath.

Read that against the 2024 field and the floor of this market has moved. The non-hyperscaler case in 2024 was sovereignty, which is a regulatory argument. The non-hyperscaler case in 2026 is price and performance for AI compute, which is an economic one.

Twenty-three of thirty is a different kind of Leader placement

The two numbers in Google Cloud's announcement answer different questions, and the gap between them is the useful part.

A highest-overall-score claim is comparative and unfalsifiable from outside, because it depends on a chart nobody outside the report can see. Twenty-three highest-possible scores in thirty named criteria is a different kind of claim. It names the criteria, and it can be checked against the reports Forrester publishes and against what a buyer can test.

The criteria Google Cloud names are vision, innovation, AI development services, database services, analytics services, containers and Kubernetes services, modernization services, and security services. The company also carries two verbatim Forrester observations. On operations: "Operators will find strong offerings in operations management as well as containers and Kubernetes services. Our evaluation did not identify significant capability gaps." On positioning: "Google Cloud's vision is to enable the 'agentic enterprise,' and AI already permeates its platform, positioning the company to push further up the tech stack toward business users who increasingly shape AI adoption in the enterprise."

Both citations describe a platform argument rather than a service list, and the second one is the report's own framing of where the market is going.

Two Leaders, two opposite theories of the same year

Microsoft's account of the same result runs on a different theory, and reading the two announcements side by side is the most instructive thing available about this edition.

Google Cloud's case is vertical co-design. The company traces its position to building silicon, systems and models together, naming TPUs, the Transformer architecture, Kubernetes, Axion and Gemini as products of one engineering culture, and argues that an integrated stack is what the agentic era requires.

Microsoft's case is data gravity. Its announcement states that customers are becoming multi-model, using frontier models where capability matters and smaller or open-weight models where economics and control matter, and concludes that value does not come from any model in isolation. Microsoft attributes the sharpest version of this to Forrester directly: "Models come and go; data has gravity." The architecture that follows is Foundry for model choice, Fabric for analytics, Purview for governance, Azure SQL and Cosmos DB for operational data, and Microsoft IQ as a context layer, with the stated goal that an organization can change models without rebuilding the data and governance around every application.

Those two theories point at different purchases. One says the value is in the integrated stack, so buy the stack. The other says the value is in the data estate, so buy the platform that lets you swap the models underneath it. Microsoft also reports that Forrester describes its direction as a vision of Azure as a single, vertically integrated system, which means both Leaders are, by their own accounts, being scored as integration stories. They just disagree about which layer the integration is for.

The agent replaced the model as the thing being scored

The 2024 edition's third finding was that cloud AI was moving "from scattershot to orchestrated services," and it noted that many generative AI services were not yet generally available when vendors were scored.

The 2026 language has moved past services. Google Cloud's citation from the report is about the agentic enterprise. Microsoft's post is titled around a system built from silicon to agent, and describes customers moving from isolated AI projects to production systems. Vultr's announcement frames its own roadmap around demand for agentic AI infrastructure.

For a buyer, that shift changes what to test. In 2024 the question was which AI services a platform offered and whether they were generally available. In 2026 the question is whether the platform can run agents against real workloads with the security, identity, governance and cost controls that production requires, which is a question about the platform underneath the AI rather than about the AI itself. Google Cloud's operations citation is the clearest signal in the public material that this is how the evaluation was weighted: the criteria it swept include containers, Kubernetes, serverless, and operations management.

What the 2026 Wave leaves unestablished

Three placements in the 2026 report are established here. Google Cloud is a Leader with the highest overall score. Microsoft is a Leader. Vultr is a Strong Performer. Every other placement in a field of ten is not established, and this page does not assert one.

The omissions are the ones a reader will notice. AWS was a Leader in 2024 and holds the largest share of this market by most public measures, and no 2026 placement for it could be reached. Alibaba Cloud was the 2024 edition's first-time Leader, the only Chinese vendor on that rung, with seventeen highest scores; its own analyst recognition page still lists the Q4 2024 result as its most recent entry for this report, so nothing about 2026 is published there. Oracle, Tencent Cloud, IBM, Huawei Cloud and OVHcloud, all charted in 2024, are similarly unestablished for the new edition.

The reason is structural rather than editorial. Forrester's own report page serves a script shell to a direct request, so the document itself is not readable from outside. Placements in a Forrester Wave travel publicly only when a vendor chooses to announce one, and a vendor announcement establishes exactly one placement. The analyst credit for the 2026 edition is not established here either; the 2024 edition was led by Lee Sustar, Charlie Dai and Dario Maisto, and nothing reachable names their successors.

So the honest summary of this refresh is that the rung is confirmed to include Google Cloud and Microsoft, the field is confirmed at ten, and the internal ordering of the rung changed in Google Cloud's favor by the company's own account.

What to establish before the commitment

Ask for the criteria, not the quadrant. A placement is a position. The scoring across thirty named criteria is a capability list, and it is the part a vendor can hand over without disclosing anything about its competitors.

Ask which edition, because this market has three. Forrester publishes a global Wave for this category and separate China and Europe editions, and a vendor citing a Leader placement may be citing a regional one that scored a different field on different criteria. The 2024 cycle ran all three.

Ask what the platform does when the model changes. Both confirmed Leaders build their case on model churn being normal. Microsoft states that customers want to change models without rebuilding the surrounding data and governance, and Google Cloud's position rests on the layers below the model being co-designed. A buyer should test the swap directly, with their own workload, before signing a multi-year commitment.

Ask where the agents will run. The 2026 evaluation language is about production agentic workloads, and the controls that matter there are security, identity, governance and cost, not model access. Those are the questions a platform team can answer and a marketing page cannot.

Forrester does not endorse any vendor named here, and tier placement should not be read as a recommendation to buy.

The natural next comparison is Public Cloud Platforms In Europe: this evaluation exists to draw a line most vendors blur on purpose: operating a cloud region physically inside Europe is not the same as being certified sovereign under European regulatory regimes.

The neighboring coverage here is Strategic Cloud Platform Services. The 2026 quadrant published on 1 September with five Leaders instead of three: Alibaba Cloud enters the top band for the first time as the only Asia-Pacific vendor in it, and AWS's share slips to 39.8 percent.