Google Cloud

Google Cloud is the enterprise cloud business of Google and the name of a reportable segment in Alphabet's financial statements. The segment is broader than Google Cloud Platform (GCP): it combines infrastructure and platform services, Google Workspace, and other enterprise services. Alphabet says the segment earns revenue mainly from consumption-based fees and subscriptions for those products.^[1][2]

That accounting boundary is important. Alphabet does not publish a separate GCP revenue figure, so Google Cloud segment revenue should not be described as GCP revenue or as revenue from AI infrastructure alone. For the quarter ended March 31, 2026, the segment reported revenue of $20.028 billion and operating income of $6.598 billion.^[2]

Scope and service model

GCP is the infrastructure and platform part of Google Cloud. It provides on-demand computing, storage, networking, databases, analytics, application hosting, and machine-learning services. Google Workspace and other enterprise applications are part of the Google Cloud segment for financial reporting, but they are not GCP.^[1][2]

Google's architecture guidance places Compute Engine, Cloud Storage, and networking services in infrastructure as a service (IaaS). It identifies App Engine, Google Kubernetes Engine, and BigQuery as examples of platform as a service (PaaS). The practical boundary varies by service: customers manage more of the software stack with IaaS and delegate more of it with managed platforms.^[17]

History

Google announced a preview of App Engine on April 7, 2008, describing it as a way for developers to run web applications on Google's infrastructure.^[3] Compute Engine entered limited preview on June 28, 2012 and became generally available on December 3, 2013.^[4]

In September 2016, Google introduced the Google Cloud name as an umbrella for GCP, the productivity applications then renamed G Suite, machine-learning tools and APIs, and other enterprise products.^[5] The scope described at that launch was wider than GCP, which helps explain why the two names are not interchangeable.

Diane Greene led the business during that reorganization. In November 2018, Google announced that Thomas Kurian would join that month and transition into the leadership role in early 2019.^[6]

Technical foundations

Google published a series of systems papers before and during the development of its public-cloud business. They document internal systems and design ideas rather than specifications for every current Google Cloud service.

Scroll sideways for more →

These systems addressed different layers of large-scale computing. GFS and Bigtable focused on storage, MapReduce on batch data processing, Dremel on interactive analysis, Spanner on distributed transactions, and Borg on cluster management.^{[7][8][9][10][11][12]} Their papers describe the systems as they existed at the time of publication and should not be used as current service-level agreements or product specifications.

Google open sourced Kubernetes in 2014. The Kubernetes project describes it as combining Google's experience running production workloads at scale with ideas and practices from the wider community.^[13] Kubernetes is an independent open-source project, not a proprietary Google Cloud service, although Google Cloud operates a managed Kubernetes service.

AI infrastructure and Vertex AI

Google has used custom tensor processing units alongside conventional processors for machine-learning workloads. The first published TPU study examined an inference accelerator deployed in Google's data centers from 2015. It described a 65,536-element 8-bit multiply-accumulate matrix unit with peak throughput of 92 tera-operations per second.^[14]

The performance results in that 2017 paper were comparisons with a specific Intel Haswell CPU and Nvidia K80 GPU on Google's measured inference workload.^[14] They do not establish how later TPU pods compare with later GPU systems. Comparisons across accelerator generations also require matching numerical precision, chip count, memory, interconnect, power boundary, software, and workload. For that reason, this article does not repeat vendor comparisons between unlike per-chip, rack, or pod configurations.

Vertex AI became generally available in May 2021. Google described it as a managed machine-learning platform that brought model building, training, deployment, monitoring, metadata, and pipeline tools into a unified interface and API.^[15] The product has changed since launch, so the original announcement is evidence for its starting scope rather than a complete list of current features or models.

Google's archived service terms dated May 29, 2026 state that Google will not use customer data to train or fine-tune AI or machine-learning models without the customer's prior permission or instruction.^[16] This contractual restriction does not remove the customer's responsibility to decide what data to send, configure access controls, review service-specific retention behavior, and evaluate outputs.

Security and customer responsibility

Google Cloud uses a shared-responsibility model. Google is responsible for the underlying network and infrastructure, while customers remain responsible for their data and access policies. The division of other controls depends on whether a workload uses IaaS, PaaS, serverless services, or software as a service.^[17]

Google completed its acquisition of Mandiant on September 12, 2022, and Mandiant joined Google Cloud while retaining its brand.^[18] Google completed the acquisition of Wiz on March 11, 2026; Wiz also joined Google Cloud and retained its brand and multicloud focus.^[19]

The Wiz closing value was not the $32 billion headline value of the 2025 agreement. Alphabet's first-quarter 2026 filing records a $29.5 billion acquisition value after purchase-price adjustments and excluding post-combination compensation arrangements. Its preliminary purchase-price allocation totals $29.467 billion.^[2] The filing includes Wiz's results in the Google Cloud segment after closing.^[2]

Financial results

Alphabet reports Google Cloud as a segment and gives both revenue and operating income. The figures below are reported values, in billions of US dollars, rather than estimates.^[1][2]

Scroll sideways for more →

The annual figures show that segment operating income increased alongside revenue from 2023 through 2025.^[1] The quarterly comparison shows the same direction between the first quarters of 2025 and 2026.^[2] These totals include GCP, Workspace, and other enterprise services. They should not be allocated among products, customers, AI workloads, or acquisitions without a separate disclosure.

Reliability

On June 12, 2025, a global incident caused increased 503 errors across multiple Google Cloud, Google Workspace, and Google Security Operations products. Google's incident report traced the failure to a new Service Control code path that lacked appropriate error handling and feature-flag protection. A policy update containing blank fields was rapidly replicated to regional data stores, triggered a null pointer, and put Service Control binaries into crash loops.^[20]

Google reported that recovery took longer in some large regions because restarting tasks overloaded a dependent Spanner table. The company also said its cloud health infrastructure was affected, delaying the first public incident report, and listed changes to rollout controls, validation, backoff behavior, architecture, and communications.^[20] The event illustrates that multi-region infrastructure can still have globally correlated control-plane failures.

Competition and regulatory findings

The United Kingdom Competition and Markets Authority completed a cloud-services market investigation in July 2025. Its market definition covered the United Kingdom and European Economic Area. For 2024, the CMA placed Google at a 5 to 10 percent share of supply in both IaaS and PaaS, behind Microsoft and Amazon Web Services.^[21] This is a scoped regulatory estimate, not a worldwide market-share figure and not a measure of the broader Google Cloud accounting segment.

The CMA found that very few customers switched cloud provider each year and identified egress fees, differentiated interfaces, latency, skills, and limited transparency as barriers to switching or using multiple clouds. It also found that certain Microsoft software-licensing practices weakened the competitiveness of AWS and Google.^[21] The CMA concluded that AI-related cloud services were growing but had not yet materially changed cloud competitive dynamics on the evidence available for its investigation.^[21]

Interpreting Google Cloud claims

Several distinctions help prevent misleading comparisons:

• Google Cloud segment versus GCP: segment revenue includes Workspace and other enterprise services.^[1][2]
• Agreement value versus closing value: the Wiz agreement was announced at $32 billion, while the completed acquisition was recorded at $29.5 billion after adjustments, with a preliminary total purchase price of $29.467 billion.^[1][2]
• Internal systems versus current products: papers on GFS, MapReduce, Bigtable, Dremel, Spanner, and Borg document particular historical systems, not every present Google Cloud implementation.^{[7][8][9][10][11][12]}
• Chip versus system comparisons: accelerator figures are meaningful only when the precision, workload, hardware count, interconnect, memory, power boundary, and software are comparable.^[14]
• Provider versus customer controls: use of a managed cloud does not transfer every security, data-governance, or access-control obligation to the provider.^[17]

See also

• Cloud computing
• Google
• Vertex AI
• Cloud TPU
• Amazon Web Services
• Microsoft Azure

References

1. ^Alphabet Inc., "Annual Report on Form 10-K for the year ended December 31, 2025", 2026-02-04. sec.gov/...goog-20251231.
2. ^Alphabet Inc., "Quarterly Report on Form 10-Q for the quarter ended March 31, 2026", 2026-04-29. sec.gov/...goog-20260331.
3. ^Google Developers Blog, "Google App Engine at Campfire One", 2008-04-07. developers.googleblog.com/...ngine-at-campfire-one.
4. ^Google Cloud, "Compute Engine release notes archive", entries dated 2012-06-28 and 2013-12-03. docs.cloud.google.com/...release-notes-archive.
5. ^Diane Greene, "Introducing Google Cloud", Google Cloud Blog, 2016-09-29. cloud.google.com/...introducing-google-cloud.
6. ^Diane Greene, "Transitioning Google Cloud after three great years", Google Cloud Blog, 2018-11-16. cloud.google.com/...-cloud-after-three-great-years.
7. ^Sanjay Ghemawat, Howard Gobioff, and Shun-Tak Leung, "The Google File System", 19th ACM Symposium on Operating Systems Principles, 2003. research.google.com/...gfs-sosp2003.pdf.
8. ^Jeffrey Dean and Sanjay Ghemawat, "MapReduce: Simplified Data Processing on Large Clusters", OSDI, 2004. research.google.com/...mapreduce-osdi04.pdf.
9. ^Fay Chang et al., "Bigtable: A Distributed Storage System for Structured Data", OSDI, 2006. research.google.com/...bigtable-osdi06.pdf.
10. ^Sergey Melnik et al., "Dremel: Interactive Analysis of Web-Scale Datasets", Proceedings of the VLDB Endowment, 2010. research.google.com/...36632.pdf.
11. ^James C. Corbett et al., "Spanner: Google's Globally-Distributed Database", OSDI, 2012. research.google.com/...spanner-osdi2012.pdf.
12. ^Abhishek Verma et al., "Large-scale cluster management at Google with Borg", EuroSys, 2015. research.google.com/...43438.pdf.
13. ^The Kubernetes Authors, "Kubernetes Overview", last modified 2026-05-30. kubernetes.io/...overview.
14. ^Norman P. Jouppi et al., "In-Datacenter Performance Analysis of a Tensor Processing Unit", ISCA, 2017. arxiv.org/...1704.04760.pdf.
15. ^Craig Wiley, "Google Cloud unveils Vertex AI, one platform, every ML tool you need", Google Cloud Blog, 2021-05-18. cloud.google.com/...-ai-unified-platform-for-mlops.
16. ^Google Cloud, "Service Specific Terms", archived version dated 2026-05-29, section 18. cloud.google.com/...index-20260529.
17. ^Google Cloud Architecture Center, "Shared responsibilities and shared fate on Google Cloud", last reviewed 2023-08-21. docs.cloud.google.com/...esponsibility-shared-fate.
18. ^Thomas Kurian, "Google + Mandiant: Transforming Security Operations and Incident Response", Google Cloud Blog, 2022-09-12. cloud.google.com/...pletes-acquisition-of-mandiant.
19. ^Thomas Kurian, "Welcoming Wiz to Google Cloud: Redefining security for the AI era", Google Cloud Blog, 2026-03-11. cloud.google.com/...e-completes-acquisition-of-wiz.
20. ^Google Cloud Service Health, "Multiple GCP products are experiencing Service issues", incident report for 2025-06-12. status.cloud.google.com/...ow5i3PPK96RduMcb1SsW.
21. ^UK Competition and Markets Authority, "Cloud services market investigation: Summary of final decision", 2025-07-31. assets.publishing.service.gov.uk/...l_decision.pdf.