Cloud Market Share 2026: AWS vs Azure vs Google Cloud
Cloud infrastructure is no longer a steady, mature market. Synergy Research Group estimates that enterprises spent about $143 billion on cloud infrastructure services in the second quarter of 2026 alone, up 43% from a year earlier, and the three leaders are moving in different directions. Amazon Web Services still leads, but its share is eroding while Google Cloud posts gains and Microsoft holds steady. This analysis explains what those percentages measure, why analysts disagree, and how to turn the numbers into a sound platform decision.
TL;DR
Latest data: Synergy’s Q2 2026 estimates put AWS at 28%, Microsoft at 20% and Google at 15% of a $143.4 billion quarter. Q2 2026 is the newest quarter published as of October 4, 2026.
Direction: Against Q2 2025, AWS is down 2 percentage points, Microsoft is unchanged and Google is up 2 points. All three are still growing quickly in dollars.
Definitions matter: Synergy and Gartner measure different markets, so their percentages are not interchangeable and should never be averaged.
AI is the accelerant: Synergy attributes the fastest market growth in eight years primarily to generative AI, and Gartner forecasts AI-optimized IaaS spending of $42.3 billion in 2026.
Choosing: Market share is a popularity signal, not a buying criterion. Existing estate, skills, data location, model needs and contracts decide fit.
Quick answer
According to Synergy Research Group, AWS led worldwide cloud infrastructure services in Q2 2026 with 28% share, followed by Microsoft at 20% and Google at 15%. The market, covering IaaS, PaaS and hosted private cloud, reached about $143.4 billion for the quarter, up roughly 43% year over year. Q2 2026 is the latest quarter available.
Which cloud provider is your organization most likely to prioritize for net-new cloud and AI workloads over the next 12 months?
0%Amazon Web Services (AWS)
0%Microsoft Azure
0%Google Cloud
0%Multicloud — no single primary provider
Table of Contents
Cloud Market Share 2026: The Latest Numbers
As of Q2 2026, the newest quarter Synergy Research Group had published by October 4, 2026, Amazon leads worldwide cloud infrastructure services with a 28% share, Microsoft follows with 20% and Google has 15%, according to Synergy’s July 30, 2026 release. I found no newer comparable release before the research cutoff, and one quarter is not a full-year result.
Provider | Q2 2026 share | Q2 2025 share | Change (percentage points) |
Amazon (AWS) | 28% | 30% | −2 |
Microsoft (Azure and related) | 20% | 20% | 0 |
Google (Google Cloud) | 15% | 13% | +2 |
Data scope: Synergy Research Group estimates of worldwide revenue for quarters ended June 30, 2026 and June 30, 2025 (prior-year shares from Synergy’s July 31, 2025 release). Scope is cloud infrastructure services: IaaS, PaaS and hosted private cloud, excluding SaaS. Shares are rounded estimates, and the change column is a calculated percentage-point difference.
Synergy estimated quarterly cloud infrastructure service revenue at $143.4 billion, with trailing-twelve-month revenue of about $500 billion. Year-on-year growth reached 43%, the highest in eight years and the 11th straight quarterly increase in the growth rate. Over those 11 quarters the market doubled. Synergy says generative AI (GenAI) has been the primary driver, and it reports that GenAI-specific cloud services grew 165% year over year.
Public IaaS and PaaS, the bulk of the market, grew 47%, and the top three providers hold 67% of that public cloud segment. Adding the three rounded shares above gives roughly 63% of the broader market, a calculation rather than a Synergy-published total.
In short, Amazon is the largest provider in this dataset, Microsoft is second, and Google is third and gaining. That describes revenue share under one definition of the market, not quality or fit. For adoption and spending context beyond share, see Articsledge’s cloud computing statistics for 2026.
What Does “Cloud Market Share” Actually Measure?
There is no single cloud market share number. Each analyst defines a different market, so each produces a different denominator and a different percentage. Reading the definition is the first step in reading any chart.
Synergy Research Group: cloud infrastructure services
Synergy’s series covers IaaS (infrastructure as a service: rented compute, storage and networking), PaaS (platform as a service: managed databases, containers, analytics and AI platforms) and hosted private cloud services. SaaS (software as a service) is excluded. Synergy also reports a separate public-cloud slice, in which the top three hold 67% in Q2 2026, which is a different denominator from the 63% sum above.
Gartner: public cloud IaaS
Gartner’s widely cited vendor table measures public cloud IaaS only. For calendar 2024, Gartner reported a $171.8 billion market in which Amazon held 37.7%, Microsoft 23.9% and Google 9.0%. Gartner’s scope, calendar-year period and vendor-revenue methodology all differ from Synergy’s, so the gap cannot be traced to a single cause.
Dataset | Scope and period | AWS | Microsoft | |
Synergy Research Group (estimate) | Cloud infrastructure services (IaaS, PaaS, hosted private cloud), Q2 2026, worldwide | 28% | 20% | 15% |
Gartner (estimate) | Public cloud IaaS, calendar 2024, worldwide | 37.7% | 23.9% | 9.0% |
These rows illustrate non-comparable scopes. Do not read across them or blend them into a “consensus.”
Company-reported segments and other terms
Companies report their own boundaries. Amazon reports an AWS segment. Microsoft reports Azure and other cloud services growth, Intelligent Cloud revenue and Microsoft Cloud revenue. Alphabet reports a Google Cloud segment that includes Google Cloud Platform (GCP) and other offerings. None of these equals a market-share denominator, and none is interchangeable with the others.
AI-optimized IaaS, meaning GPU and accelerator capacity sold as a service, is a growing subset that Gartner tracks separately. Hosted private cloud sits in Synergy’s scope but not Gartner’s IaaS table.
Two arithmetic habits prevent common errors. A move from 30% to 28% is a fall of two percentage points, not two percent; in relative terms it is about 6.7%. And a provider can lose share while its revenue grows if the market or its rivals grow faster. Compare only numbers from the same source, scope and period.
How AWS, Azure, and Google Cloud Market Share Has Changed
Using Synergy’s Q2 figures for 2024, 2025 and 2026 holds scope and seasonality constant. Over two years, AWS lost four percentage points, Microsoft lost three and Google gained three.
Quarter | AWS | Microsoft | Market size (Synergy) | |
Q2 2024 | 32% | 23% | 12% | about $79B |
Q2 2025 | 30% | 20% | 13% | $98.8B |
Q2 2026 | 28% | 20% | 15% | $143.4B |
Change, Q2 2024 to Q2 2026 | −4 points | −3 points | +3 points | about 1.8x |
Sources and method: Q2 2024 figures are Synergy data as reported by TechCrunch on August 2, 2024; Q2 2025 and Q2 2026 come from Synergy’s releases cited above. The change row and the 1.8x multiple are calculations. Synergy revises history: its Q2 2026 release says the market rose by more than $43 billion to $143 billion, which implies a Q2 2025 base near $100 billion, slightly above the $98.8 billion first published.
Revenue growth, market growth and share are different things. Amazon reported AWS segment sales growth of 37% for Q2 2026, strong and its fastest in 18 quarters, but below Synergy’s 43% market growth, so share slips. Microsoft reported 43% growth in Azure and other cloud services, in line with the market and consistent with a flat 20%. Alphabet reported 82% growth in its Google Cloud segment, far above the market and consistent with a two-point gain. These are directional cross-checks between differently defined metrics, not reconciliations.
Absolute dollars matter too. Because the market added more than $43 billion in a single year, a provider can lose share and still add substantial revenue. Losing share in a fast-growing market is not the same as shrinking.
Concentration at the top has also eased. Adding rounded shares, the three leaders held about 67% of the whole market in Q2 2024 and about 63% in both Q2 2025 and Q2 2026. Synergy’s public-cloud measure drifts the same way, from 68% in Q2 2025 to 67% in Q2 2026, while tier-two providers including CoreWeave, Oracle, Crusoe and Nebius posted the fastest growth.
Avoid cherry-picking single quarters. Synergy noted in its Q3 2025 release that Amazon’s share averaged just under 30% over the prior four quarters, down from a little over 32% in 2021. The multi-year picture is gradual erosion for Amazon and steady gains for Google. Microsoft’s Q2 share fell three points between 2024 and 2025 and has been flat since.
AWS in 2026
AWS remains the largest provider in Synergy’s cloud infrastructure services data (28% in Q2 2026) and in Gartner’s latest public IaaS vendor table (37.7% for 2024). Amazon reported that AWS segment sales rose 37% year over year to $42.2 billion in Q2 2026, its fastest growth in 18 quarters, at a $169 billion annualized run rate. AWS segment operating income was $16.6 billion, up from $10.2 billion. These are AWS segment results as Amazon defines them and should not be set directly beside Microsoft’s or Alphabet’s segment figures.
Core services and the AI stack
The building blocks most buyers evaluate are Amazon EC2 for compute, S3 for object storage, Lambda and Fargate for serverless, EKS for managed Kubernetes, and databases from Aurora and RDS to DynamoDB. Analytics centers on Redshift, Athena and the S3 data lake pattern.
On AI, AWS pairs Amazon Bedrock (managed access to foundation models and agent tooling) with SageMaker AI (model building and training). At re:Invent 2025, AWS announced Trainium3 UltraServers, built on its first 3nm AI chip with up to 144 chips per system, plus Bedrock AgentCore for agents and AI Factories that place dedicated AWS AI infrastructure in customer data centers. AWS says some customers cut training and inference costs by up to 50% with Trainium, a vendor claim rather than an independent benchmark. In April 2026 AWS and OpenAI announced OpenAI models on Bedrock in limited preview, and Amazon’s page says a general-availability rollout followed in June.
Footprint
As of early October 2026, AWS’s infrastructure page lists 124 Availability Zones within 39 geographic Regions, with plans for seven more zones and two more Regions in Saudi Arabia and Chile. The page treats edge locations and other zone types separately from Regions, which is why vendor counts are not like for like.
Strengths, trade-offs and common fit
Strengths: the broadest service catalog, a large partner and marketplace ecosystem, and a deep pool of engineers with AWS experience, which many teams treat as lower hiring risk (an industry perception, not a measured figure).
Trade-offs: service breadth brings overlapping options and pricing complexity. Share is eroding relative to faster-growing rivals. Organizations built around Microsoft identity and licensing may need more integration work than on Azure.
Common fit: cloud-native product companies, teams with existing AWS skills, workloads that need wide service and regional coverage, and buyers who want Bedrock’s multi-model approach. Fit still depends on data location, commitments and compliance needs.
Microsoft Azure in 2026
Microsoft Azure is second in Synergy’s Q2 2026 data at 20%, unchanged from a year earlier.
What Microsoft reports, and what it does not
Microsoft’s fiscal Q4 2026 release does not give a standalone quarterly Azure dollar figure. It reports 43% growth in Azure and other cloud services, Intelligent Cloud segment revenue of $39.3 billion (up 32%) and Microsoft Cloud revenue of $59.3 billion (up 27%), a broader measure that spans more than Azure. CEO Satya Nadella said Azure revenue surpassed $100 billion for the first time in fiscal 2026. Commercial remaining performance obligation, contracted revenue not yet recognized, rose 84% to $678 billion. None of these numbers equals Azure revenue for the quarter.
Ecosystem, hybrid and core services
Azure’s distinctive advantage is the Microsoft estate. Microsoft Entra ID, Microsoft 365, Dynamics 365, Power Platform and GitHub connect naturally to Azure, and Nadella said Microsoft 365 Copilot passed 30 million paid seats, which ties AI adoption to existing Microsoft licensing. For hybrid, Azure Arc extends Azure management and policy to servers and Kubernetes clusters running elsewhere, and Azure Local runs Azure-managed infrastructure on premises. Core services include Azure Virtual Machines, Blob Storage, Azure Kubernetes Service (AKS), Azure Functions and Container Apps, Azure SQL Database and Cosmos DB.
AI platform and custom silicon
Microsoft’s AI platform is Microsoft Foundry, the third name for a product line that began as Azure AI Studio in 2023, became Azure AI Foundry in 2024 and is now Microsoft Foundry. It combines a model catalog with Foundry Agent Service and related tools. The catalog spans OpenAI, Anthropic, Meta, Mistral and DeepSeek models, and Anthropic’s Claude models became generally available in Foundry in June 2026 with Azure authentication, billing and governance. Microsoft says its Maia 200 inference accelerator is in production in Iowa and Arizona and that Cobalt 200 Arm virtual machines are in early-access preview, according to Build 2026 coverage. Microsoft’s chip comparisons with rivals are vendor claims.
Footprint, strengths, trade-offs and common fit
Microsoft Learn says Azure provides over 70 regions, and regions that support availability zones have at least three zones. The page does not give a single zone total.
Strengths: Microsoft identity, productivity and developer integration; strong hybrid tooling; a multi-model AI catalog with OpenAI and Anthropic options; and deep enterprise relationships.
Trade-offs: limited disclosure of standalone Azure revenue makes share estimates less transparent. Microsoft notes that sovereign cloud regions often have limited services and features. Licensing and commitment structures can be complex.
Common fit: Microsoft-centric enterprises, organizations with .NET and Windows estates, hybrid environments, and buyers who want OpenAI or Claude models inside Azure governance.
Google Cloud in 2026
Google Cloud is third in Synergy’s Q2 2026 data at 15%, up from 13% a year earlier and 12% in Q2 2024. Alphabet reported Google Cloud segment revenue of $24.8 billion for Q2 2026, up 82%, with segment operating income of $8.8 billion versus $2.8 billion a year earlier. Alphabet attributed the growth to Google Cloud Platform (GCP) across enterprise AI solutions and AI infrastructure, plus core GCP services. The Google Cloud segment is broader than raw GCP infrastructure consumption because it also includes other enterprise offerings such as Google Workspace, so it is not a pure infrastructure revenue figure.
Data, Kubernetes and serverless
Google’s reputation rests on data and containers. BigQuery, which Google describes as an autonomous data-to-AI platform, anchors its analytics stack, alongside databases including Spanner, AlloyDB, Cloud SQL, Firestore and Bigtable. Kubernetes originated at Google, and Google Kubernetes Engine (GKE) and Cloud Run, a serverless container service, are flagship offerings. Core compute and storage are Compute Engine and Cloud Storage.
Gemini Enterprise Agent Platform, TPUs and GPUs
On April 22, 2026, Google announced Gemini Enterprise Agent Platform as the evolution of Vertex AI, and AIwire’s coverage describes a rebrand and expansion into a full agent stack. Google’s release notes map the renames; for example, Vertex AI Agent Engine became Agent Runtime. Google pitches a full-stack approach: custom TPUs (Ironwood is its seventh-generation TPU), Gemini models, the agent platform, and distribution through Workspace and the Gemini Enterprise app, as The Next Web reported from Cloud Next 2026. Google’s Model Garden lists over 200 models from Google and partners, and GPUs remain available alongside TPUs.
Network and footprint
Google’s locations page, updated September 23, 2026, lists 43 regions, 130 zones and more than 200 network edge locations. A region has three or more zones in three or more physical data centers, except Stockholm, Mexico, Osaka and Montreal, which have three zones in one or two data centers and are being expanded.
Strengths, trade-offs and common fit
Strengths: data and analytics services, Kubernetes heritage, a vertically integrated AI stack from silicon to models, and a global network.
Trade-offs: a smaller share means a smaller local pool of engineers and partners in many markets (a general observation, not a measured figure). Gartner’s IaaS-only table showed Google at 9.0% for 2024, well below its Synergy share. One hypothesis is that Google’s revenue leans toward platform, data and AI services, but the datasets differ in many ways, so treat that as a hypothesis, not a finding.
Common fit: data-heavy and analytics-led organizations, Kubernetes-centric platforms, teams building on Gemini models, and companies that already use Google Workspace.
AWS vs Azure vs Google Cloud: Side-by-Side Comparison
This table compares the capabilities that most often shape shortlists. Names are current as of October 4, 2026 and omit hundreds of niche services. It deliberately contains no star ratings.
Capability | AWS | Microsoft Azure | Google Cloud |
Share, Synergy Q2 2026 | 28%, largest | 20%, second | 15%, third and gaining |
Virtual machines | Amazon EC2 | Azure Virtual Machines | Compute Engine |
Object storage | Amazon S3 | Azure Blob Storage | Cloud Storage |
Managed Kubernetes | Amazon EKS | Azure Kubernetes Service | Google Kubernetes Engine |
Serverless | Lambda, Fargate | Functions, Container Apps | Cloud Run, Cloud Functions |
Relational databases | Aurora, RDS | Azure SQL Database | Cloud SQL, AlloyDB |
Distributed databases | DynamoDB | Cosmos DB | Spanner, Bigtable |
Analytics and warehouse | Redshift, Athena | Microsoft Fabric, Power BI | BigQuery, Looker |
AI platform | Bedrock, SageMaker AI, AgentCore | Microsoft Foundry | Gemini Enterprise Agent Platform |
Model access | Bedrock: Amazon Nova, open-weight and partner models, OpenAI | Foundry: OpenAI, Anthropic, Meta, Mistral, DeepSeek | Gemini; Model Garden with 200+ models |
Custom silicon | Trainium3; Graviton5 CPUs | Maia 200 (inference); Cobalt 200 CPUs | TPUs (Ironwood) |
Hybrid and multicloud | Outposts, AI Factories | Azure Arc, Azure Local | Distributed Cloud, Cross-Cloud Network |
Enterprise ecosystem | Broad partner ecosystem, AWS Marketplace | Microsoft 365, Dynamics 365, Entra ID, GitHub | Workspace, Gemini Enterprise |
Footprint, Oct 2026 | 39 Regions, 124 AZs | 70+ regions; zones vary | 43 regions, 130 zones, 200+ edge |
Commitment discounts | Savings Plans, Reserved Instances | Reservations, savings plan | Committed use discounts |
Security building blocks | IAM, KMS, GovCloud | Entra ID, Key Vault, Azure Government | Cloud IAM, Cloud KMS, Assured Workloads |
Read the table at the category level. Each cloud has an answer for nearly every row, so differences lie in depth, integration and price rather than presence. AWS offers the widest menu, Azure integrates most tightly with Microsoft identity and licensing, and Google positions its data and AI stack as the most vertically integrated. Confirm service availability in your target region before relying on any row, because services and AI models do not launch everywhere at once.
AI and Generative AI: How the Cloud Race Is Changing
AI has become the main driver of cloud growth. Synergy attributes the market’s 43% growth in Q2 2026 primarily to GenAI and says AI is also lifting growth across a broader range of cloud services. The competitive question has shifted from who rents the most virtual machines to who can supply scarce accelerator capacity, models and agent tooling together.
Training, inference and capacity
Gartner forecasts worldwide AI-optimized IaaS spending, which covers GPUs, ASICs and other accelerators sold as a service, at $42.3 billion in 2026, up 96%, rising to $66.1 billion in 2027. It expects inference spending to pass training in 2026 ($23.3 billion versus $19 billion), with inference at 55% of AI-optimized IaaS spend. Inference rewards low cost per token and regional proximity, which helps explain why all three vendors promote custom silicon: Trainium3, Maia 200 and TPUs.
Capacity is the constraint that spending reveals. Microsoft’s cash flow statement shows additions to property and equipment of $115.9 billion in fiscal 2026 versus $64.6 billion the year before. That line excludes finance leases, and each company defines capital expenditure differently, so avoid comparing it directly with rivals’ figures. Heavy spending lifts growth but raises the bar for returns, which is why contracted backlog, such as Microsoft’s $678 billion commercial remaining performance obligation, draws attention.
How the three platforms position AI
AWS: Bedrock for model access, SageMaker AI for training and customization, AgentCore for agent runtime, and Trainium3 plus NVIDIA GPUs for compute. OpenAI models joined Bedrock in 2026 alongside Amazon Nova and other providers.
Microsoft: Microsoft Foundry for models, agents and governance, with OpenAI and Claude options, Maia 200 and NVIDIA-based infrastructure, and the Copilot ecosystem, where Microsoft 365 Copilot passed 30 million paid seats.
Google: Gemini Enterprise Agent Platform with Gemini models and Model Garden, TPUs and GPUs, BigQuery for data-to-AI workflows, and Workspace distribution.
Neoclouds and tier-two pressure
Synergy says nine neocloud companies are now among the top 40 cloud providers by cloud infrastructure service revenue, and the fastest-growing tier-two names include CoreWeave, OpenAI, Oracle, Crusoe, Nebius, Anthropic and Nscale. Gartner has described GPU-as-a-service providers as nascent but important for meeting immediate capacity needs. In interpretation, neoclouds compete on GPU availability while hyperscalers compete on integrated data, security and enterprise contracts, and many buyers combine both.
How to compare AI offerings
Separate four layers: infrastructure (GPUs, custom silicon, regional capacity), model access (which models, in which regions), orchestration and agent tooling (runtimes, memory, policy, evaluation) and data integration (where your data already sits and what moving it costs). A provider can lead in one layer and trail in another. No source supports an objective universal AI winner, and this article does not name one.
For background, see Articsledge’s guides to AI infrastructure and AI data centers, and a comparison of Azure AI, AWS SageMaker and Google Vertex AI. Vertex AI is now part of Gemini Enterprise Agent Platform.
Pricing and Total Cost of Ownership
List-price comparisons rarely predict real cost. This article does not publish an instance-price comparison, because a fair one must normalize region, vCPU, memory, CPU architecture, operating system and licensing, commitment term, storage and data transfer, and results change with every price update. No provider is universally cheapest.
What drives the bill
Region, availability-zone design and machine shape, including x86 versus Arm.
Operating system and licensing, including bring-your-own-license options.
Commitment term and coverage, and spot capacity (Spot Instances on AWS, Spot VMs on Azure and Google Cloud).
Storage tier, request volume, and managed-service premiums versus the operations labor they replace.
Data transfer and egress, support plans, utilization and architecture choices.
Negotiated enterprise agreements and marketplace commitments, which list prices never show.
Commitment mechanisms by provider
AWS: Savings Plans come in four types: Compute, EC2 Instance, Database and SageMaker AI. They commit to hourly spend. Reserved Instances remain available for specific configurations.
Azure: Reservations commit to specific resources. Azure savings plan for compute commits to an hourly amount for one or three years, and Microsoft advertises savings of up to 65% against pay-as-you-go prices for select compute services, a vendor claim whose results vary. A savings plan for databases also exists.
Google Cloud: Committed use discounts are resource-based or spend-based. Compute flexible CUDs cover Compute Engine, GKE and Cloud Run, service-specific CUDs cover products such as Cloud SQL and AlloyDB, and Flexible Savings Plans cover some generative AI workloads. Eligible uncommitted Compute Engine usage may receive sustained use discounts.
Egress, exit costs and FinOps
Moving data out of a cloud is a separate cost line. In 2024 Google, AWS and Microsoft each announced programs that waive eligible data-transfer-out charges for customers leaving entirely, with conditions such as notifying the provider and finishing the move within a set window, according to CIO Dive. Routine egress, such as serving users or moving data between clouds, is still billed. The EU Data Act also calls for providers to abolish switching charges within three years, which points to January 2027. Gartner analyst Lydia Leong told CIO Dive that egress fees are a red herring compared with the skills cost of switching.
FinOps (financial operations for cloud) is the practice of making cloud spend visible, accountable and optimized across engineering, finance and procurement. In practice, utilization, rightsizing and commitment coverage often move the bill more than the choice of provider. Model total cost of ownership (TCO) including people, migration, support and exit, using each provider’s pricing calculator and your own usage data.
Global Infrastructure, Reliability, Security, and Compliance
Infrastructure counts change often and use different units. The figures below come from each provider’s own pages in early October 2026.
Provider | Regions | Zones | Other units and notes |
AWS | 39 Regions | 124 Availability Zones | Plans announced for 7 more zones and 2 more Regions; edge locations and other zone types listed separately |
Microsoft Azure | Over 70 regions | Varies; at least 3 where supported | Geographies act as data residency boundaries; many newer regions are not paired |
Google Cloud | 43 regions | 130 zones | 200+ network edge locations; four regions still expanding to 3+ data centers |
Sources: AWS, Microsoft Learn and Google Cloud. Regions, zones, data centers and edge locations are not equivalent units, and announced regions are not operational regions. AWS’s page contrasts its multi-zone Regions with providers that, it says, define a region as a single data center, while Microsoft and Google both document zone-based designs, so check the specific region you plan to use.
Resilience design
Resilience comes from architecture, not counts. Spread workloads across at least two zones, confirm that each service you depend on supports zone redundancy in your region, and decide whether a second region is justified. Microsoft notes that many newer regions rely on availability zones rather than paired regions, and Google distinguishes zonal, regional and multi-region resources. Regional service availability matters more than raw region count, because a service or AI model may not exist in your region.
Data residency and sovereignty
Gartner forecasts worldwide sovereign cloud IaaS spending of $80 billion in 2026, up 35.6%, and expects Europe to overtake North America in that category in 2027. Sovereign and government regions, such as AWS GovCloud and Azure Government, often offer fewer services, so verify the service list. Articsledge covers the trade-offs in its guides to sovereign cloud and hosted private cloud.
Security and compliance
All three publish extensive compliance documentation and offer identity and access management, encryption with customer-managed keys, logging and policy tooling. No provider is categorically more secure. Outcomes depend on configuration, identity hygiene and your operating model under the shared responsibility model, so test each provider’s controls against your regulator’s requirements and your team’s skills.
Hybrid Cloud, Multicloud, and Migration
Hybrid and multicloud are the norm rather than the exception. Flexera’s 2026 State of the Cloud report, based on 753 respondents, found that 73% of organizations use hybrid cloud, and it says multicloud adoption keeps rising, often unintentionally through acquisitions and team choices.
Hybrid and on-premises options
Azure Arc and Azure Local, AWS Outposts and AI Factories, and Google Distributed Cloud extend each provider’s services to your data center or edge. The right choice usually follows your existing management plane, identity system and hardware.
VMware and migration paths
Broadcom’s licensing changes have pushed many VMware customers to evaluate Amazon Elastic VMware Service (EVS), Azure VMware Solution and Google Cloud VMware Engine. AWS introduced EVS so customers can run VMware Cloud Foundation inside an Amazon VPC with portable licenses. Licensing terms change frequently, so confirm current terms with each provider. For database moves, each provider offers a database migration service, and for modernization all three support containers and managed Kubernetes. See Articsledge’s guides to cloud migration, hybrid cloud and Kubernetes.
Multicloud: benefits and costs
Benefits include negotiating leverage, resilience to a provider-wide outage, and access to best-of-breed services. Costs include duplicated skills and tooling, separate identity and security models, cross-cloud network and egress charges, and a tendency to build to the lowest common denominator. Portable layers such as Kubernetes and infrastructure as code help, but they rarely make workloads fully interchangeable. Articsledge’s multi-cloud guide covers the strategy in depth.
Which Cloud Is the Strongest Fit for Different Workloads?
There is rarely a single winner. The table lists plausible strong fits and the reasoning, and every row assumes you have verified region, service and compliance fit.
Scenario | Plausible strong fits | Why, and the main caveat |
Microsoft-centric enterprise | Azure | Entra ID, Microsoft 365 and licensing integration. AWS and Google also work with identity federation. |
Cloud-native startup | AWS, Google Cloud, Azure | Credits, skills and ecosystem usually decide. Avoid choosing on share. |
AI and ML | All three, by model need | Azure for OpenAI and Claude under Azure governance, Google for TPUs and Gemini, AWS for Bedrock and Trainium. Check models and capacity in your region. |
Large-scale analytics | Google Cloud, AWS, Azure | BigQuery is Google’s hallmark, AWS suits S3 data lakes, Azure suits Power BI-centered shops. Data gravity dominates. |
Kubernetes-centric platform | Google Cloud, AWS, Azure | All offer managed Kubernetes. Google has the heritage, so decide on team skills and portability needs. |
Hybrid or on-premises enterprise | Azure, AWS, Google Cloud | Arc and Azure Local lead for Microsoft estates. Outposts and Distributed Cloud serve others. |
Serverless-heavy application | AWS, Google Cloud, Azure | Lambda, Cloud Run and Functions each fit. Watch limits and lock-in. |
Global consumer application | AWS, Google Cloud, Azure | Compare edge presence, regional service availability and latency to your users. Counts are not comparable. |
Regulated workload | Depends on regulator and region | Each offers compliance programs and sovereign or government options. Verify certifications and service lists. |
Multicloud strategy | Google Cloud, Azure, AWS | Azure Arc and Google’s Cross-Cloud Network target multicloud directly. Any approach adds operational cost. |
Cost-optimization focus | Any; model it | Commitments, enterprise agreements, architecture and FinOps maturity outweigh list price. |
Three caveats apply to every row. Skills: a team that knows one cloud well often outperforms a theoretically better platform it does not know. Data gravity: large datasets pull compute toward them. Contracts: existing discounts and marketplace commitments change the economics. Switching cost is part of the decision.
How to Choose Between AWS, Azure, and Google Cloud
Market share describes what others have bought. It is one signal, not a substitute for workload-level evaluation. Use a weighted checklist and test with real workloads.
Existing estate and contracts: Microsoft licensing, current AWS or Google commitments, and marketplace spend that can be burned down.
Team skills and talent availability: hiring and training often cost more than platform price differences. See Articsledge’s cloud operating model guide.
Workload architecture: VMs, containers, serverless, databases and AI, mapped to equivalent services.
Data gravity and location: where data lives, how much moves, and the egress it triggers.
Regulatory constraints and latency: residency, sector rules, certifications and user proximity.
AI and model requirements: required models, GPU or TPU access, and capacity in your regions.
Reliability and security operating model: zone design, identity, key management and incident processes.
Support and migration effort: support plans, partner availability and the cost of moving.
Portability, lock-in and exit strategy: what leaving would cost in dollars and months, and where you accept proprietary services deliberately.
Concentration risk: dependence on one vendor for critical systems.
Proof of concept: benchmark representative workloads and measure cost, performance and operational effort.
The technically strongest platform in one dimension may not be the cheapest choice once switching, retraining and migration are counted. Decide on evidence from your own workloads.
Cloud Market Outlook for the Rest of 2026 and Beyond
Observed facts. The market grew 43% year over year in Q2 2026, and growth has accelerated for 11 straight quarters. GenAI-specific services grew 165%. The U.S. share of the worldwide market has risen in the last two quarters as hyperscalers and neoclouds build U.S. capacity. Amazon, Microsoft and Alphabet each reported strong or accelerating cloud growth in their latest results.
Forecasts, not facts. In July 2025, Synergy forecast that average annual growth would stay above 20% over the next five years. Gartner’s August 10, 2026 forecast sees total IaaS spending of $287.3 billion in 2026 (up 29.3%) and $359.9 billion in 2027, with AI-optimized IaaS at $42.3 billion and $66.1 billion in those years and inference rising to 59% of that spending in 2027. Gartner’s February 2026 forecast puts sovereign cloud IaaS spending at $80.4 billion in 2026 and $110.6 billion in 2027. Gartner’s IaaS scope is narrower than Synergy’s, so the dollar totals are not comparable.
Forces to watch: accelerator supply and power availability, the shift from training to inference, custom silicon from all three leaders, sovereignty requirements that favor local partners and sovereign regions, neocloud competition for AI capacity, and multi-model platforms that reduce model lock-in. This article does not forecast future AWS, Azure or Google Cloud shares, and no source cited here does either. Synergy’s next release, covering Q3 2026, will show whether the Q2 pattern holds.
FAQ
What is AWS’s cloud market share in 2026?
According to Synergy Research Group, AWS held 28% of worldwide cloud infrastructure services revenue in Q2 2026, down from 30% in Q2 2025. On Gartner’s narrower public IaaS measure, Amazon held 37.7% for calendar 2024. The two figures use different definitions and periods.
What is Microsoft Azure’s cloud market share in 2026?
Synergy estimates Microsoft held 20% in Q2 2026, unchanged from Q2 2025 and down from 23% in Q2 2024. Microsoft does not report standalone Azure revenue, so shares are estimates. It reported 43% growth in Azure and other cloud services for the quarter ended June 30, 2026.
What is Google Cloud’s market share in 2026?
Synergy puts Google at 15% in Q2 2026, up from 13% a year earlier and 12% in Q2 2024. Gartner’s IaaS-only table showed Google at 9.0% for 2024, because the two datasets measure different markets.
Which company has the largest cloud market share?
Amazon (AWS) is largest by revenue in both Synergy’s Q2 2026 data and Gartner’s latest vendor table. “Largest” describes revenue share in one dataset. It does not describe quality, price or fit for a particular workload.
Is AWS losing cloud market share?
In share terms, gradually: AWS fell from 32% in Q2 2024 to 28% in Q2 2026 in Synergy’s data. In revenue terms it is growing, and Amazon reported AWS segment sales up 37% to $42.2 billion in Q2 2026. A provider can lose share while growing if the market grows faster.
Is Google Cloud gaining cloud market share?
Yes, in Synergy’s data: 12% in Q2 2024, 13% in Q2 2025 and 15% in Q2 2026. Alphabet reported Google Cloud segment revenue up 82% in Q2 2026. Google remains third and well behind AWS in share.
Why do Gartner and Synergy show different cloud market shares?
Gartner’s table measures public cloud IaaS for a calendar year. Synergy measures IaaS, PaaS and hosted private cloud by quarter. Different scopes, periods and revenue allocations produce different denominators, so the percentages are not interchangeable and should not be averaged.
Is Azure bigger than AWS?
Not by share in Synergy’s Q2 2026 data: AWS 28% versus Microsoft 20%. Microsoft’s Intelligent Cloud and Microsoft Cloud revenue figures include more than Azure, so they are not like for like with AWS segment revenue.
Which cloud is better for AI?
There is no universal winner. AWS offers Bedrock, SageMaker AI and Trainium. Azure offers Microsoft Foundry with OpenAI and Anthropic models. Google offers Gemini Enterprise Agent Platform, Gemini models and TPUs. Fit depends on required models, data location, regional capacity and skills.
Which cloud is cheapest?
None is universally cheapest. Cost depends on region, instance type, commitments, licensing, storage, data transfer and architecture, and negotiated enterprise discounts can outweigh list prices. Model the same workload on each cloud before deciding.
Do most enterprises use more than one cloud?
Often, yes. Flexera’s 2026 State of the Cloud report, with 753 respondents, found 73% of organizations use hybrid cloud, and it says multicloud adoption keeps rising, often unintentionally through acquisitions and team choices.
How big is the cloud infrastructure market in 2026?
Synergy estimated $143.4 billion of cloud infrastructure services revenue in Q2 2026 and about $500 billion over the trailing twelve months. Gartner, using a narrower IaaS definition, forecasts $287.3 billion of IaaS spending for 2026.
Key Takeaways
Cloud market share is a definition before it is a number: Synergy’s 28/20/15 and Gartner’s 37.7/23.9/9.0 describe different markets and periods.
AWS’s share is slipping because the market is growing faster than AWS’s 37%, not because AWS is shrinking.
Microsoft’s share is steady at 20%, but limited Azure disclosure makes estimates harder to verify.
Google Cloud is the share gainer, with 82% segment growth, though it remains well behind in share.
AI is reshaping competition around accelerator capacity, models and agent tooling, with neoclouds adding pressure.
List prices mislead. Commitments, enterprise agreements, egress and architecture drive real cost.
Fit beats rank: existing estate, skills, data gravity, compliance and model needs should shape the shortlist.
Forecasts exist for market size and AI spending, but none supports predicting future provider shares.
Actionable Next Steps
Inventory workloads, dependencies, data volumes and owners.
Classify compliance, data-residency and latency constraints for each workload.
Model TCO for two or three representative workloads on each shortlisted cloud, using provider calculators and your own usage data.
Benchmark those workloads, measuring performance, reliability and operational effort.
Evaluate AI and data requirements: required models, GPU or TPU capacity and where the data lives.
Review commitments and discounts, including enterprise agreements, savings plans, reservations and marketplace commitments.
Test portability with containers, infrastructure as code and data export paths.
Run a time-boxed proof of concept with success criteria agreed in advance.
Plan an exit path, and revisit the decision when Synergy’s Q3 2026 data and your contract renewals arrive.
Glossary
AWS: Amazon Web Services, Amazon’s cloud business.
Azure: Microsoft’s cloud platform.
GCP: Google Cloud Platform, the infrastructure and platform part of Google Cloud.
IaaS: infrastructure as a service; rented compute, storage and networking.
PaaS: platform as a service; managed databases, containers and developer platforms.
SaaS: software as a service; finished applications delivered over the internet.
Hyperscaler: a very large cloud provider operating global data center networks.
Region and Availability Zone: a geographic cloud location, and an isolated zone within it.
Hybrid cloud: public cloud combined with private or on-premises infrastructure.
Multicloud: using more than one public cloud provider.
FinOps: managing cloud spend across engineering, finance and procurement.
Serverless: running code without managing servers, billed by use.
Kubernetes: open-source system for running containers at scale.
GPU and TPU: accelerator chips for AI; TPUs are Google’s custom design.
Inference: running a trained AI model to produce answers.
Foundation model: a large general-purpose AI model.
Data egress: data leaving a provider’s network, often billed.
Committed use discount: a discount for committing to usage or spend.
Percentage point: the unit for differences between percentages.
Neocloud: a newer provider focused on GPU and AI capacity.
Sources & References
Q2 Cloud Market Passes $143 Billion; Highest Growth Rate in Eight Years, Synergy Research Group, July 30, 2026.
Q2 Cloud Market Nears $100 Billion Milestone, Synergy Research Group, July 31, 2025.
Cloud Market Share Trends: Big Three Together Hold 63%, Synergy Research Group, Q3 2025 release, covered by The Register on November 20, 2025.
Cloud infrastructure revenue approached $80 billion this quarter, TechCrunch, August 2, 2024.
Gartner Says Worldwide IaaS Public Cloud Services Market Grew 22.5% in 2024, Gartner, August 6, 2025.
Gartner Forecasts Worldwide AI-Optimized IaaS Spending to Grow 96% in 2026, Gartner, August 10, 2026.
Gartner Says Worldwide Sovereign Cloud IaaS Spending Will Total $80 Billion in 2026, Gartner, February 9, 2026.
Amazon.com Announces Second Quarter Results, Amazon.com, Inc. (SEC Form 8-K exhibit), July 30, 2026.
Microsoft Cloud and AI Strength Fuels Fourth Quarter Results, Microsoft Investor Relations, July 29, 2026.
Alphabet Announces Second Quarter 2026 Results, Alphabet Inc. (SEC Form 8-K exhibit), July 22, 2026.
AWS Regions and Availability Zones, Amazon Web Services, accessed October 2026.
What are Azure regions?, Microsoft Learn, updated June 29, 2026.
Cloud locations, Google Cloud, last updated September 23, 2026.
Gemini Enterprise Agent Platform release notes, Google Cloud documentation, accessed October 2026.
Google Unveils Gemini Enterprise Agent Platform, Expands Vertex AI into Full Agent Stack, AIwire, April 23, 2026.
Google Cloud Next 2026: AI agents, A2A protocol, Workspace Studio, and the full-stack bet, The Next Web, April 2026.
Frontier agents, Trainium chips, and Amazon Nova: key announcements from AWS re:Invent 2025, Amazon, December 2025.
OpenAI models, Codex and Managed Agents on Amazon Bedrock, Amazon News, April 28, 2026, updated later in 2026.
Anthropic’s Claude Comes to Microsoft Foundry as Azure Adds Another Frontier Model Option, Pure AI, June 30, 2026.
Get started with Microsoft Foundry SDKs and endpoints, Microsoft Learn, accessed October 2026.
Foundry Gets New Name, Anthropic Models, Directions on Microsoft, date not shown on the retrieved page.
Microsoft launches VMs based on Cobalt 200 chip in preview, Maia 200 already in production, DatacenterDynamics, 2026 (Build 2026 coverage; exact date not shown).
Flexera 2026 State of the Cloud Report: The convergence of cloud and value, Flexera, March 2026.
Microsoft Azure eliminates cloud data egress fees, CIO Dive, March 2024.
Savings Plans FAQs, Amazon Web Services, accessed October 2026.
Azure savings plan for compute, Microsoft Azure, accessed October 2026.
Committed use discounts overview, Google Cloud documentation, accessed October 2026.
Amazon Introduces Elastic VMware Service for VMware Workload Migration to AWS in Preview, InfoQ, December 2024.
