Executive Summary

The Confidential Computing Market was valued at 24.2 USD Billion in 2025 and is projected to reach 624.9 USD Billion by 2035, registering a CAGR of 38.4% over the forecast period.

Financial-sector compliance is the sharper mechanism: the EU’s Digital Operational Resilience Act, binding since January 2025, was cited by 77% of a 2025 Confidential Computing Consortium study’s adopters. Loosened US export licensing for advanced accelerators, effective January 2026, widens the accelerator base for confidential cloud instances.

North America held 42.0% of the market in 2025, ahead of Asia Pacific at 29.0% and Europe at 21.0%. Hardware led the Component axis and Cloud led on deployment mode.

Re-architecting legacy applications for enclave execution remains the binding constraint for regulated institutions. Participation spans chip designers, hyperscalers, HSM and TPM specialists, and integrators delivering attestation tooling.

Key Takeaways

• The market grows from USD 24.24 Billion in 2025 to USD 624.86 Billion by 2035, a 38.4% CAGR helped by consumption-based pricing that is lifting vendor ARR faster than legacy seat licensing.
• Hardware leads the component split, tracking the silicon and infrastructure buildout that precedes each new hyperscaler region announcement.
• Cloud is the largest deployment mode. Data-residency obligations under the GDPR, China’s PIPL, and the EU Data Act and Data Governance Act are steering some enterprise workloads toward sovereign and regional cloud instances instead.
• North America holds the largest regional share, at 42.0% in 2025, a position reinforced by SEC cyber-disclosure rules and the CCPA/CPRA framework that keep compliance spend embedded in US platform budgets.
• Ten profiled suppliers compete on platform breadth and integration surface, with EU AI Act conformity deadlines and NIS2 security obligations now factoring into vendor roadmaps for buyers selling into the EU.

Market Definition and Scope

The Confidential Computing Market covers hardware, software and services that protect data in use through hardware-based trusted execution environments, a boundary distinct from encryption applied only at rest or in transit. It spans enclave-capable CPUs, hardware security modules and trusted platform modules; attestation and orchestration platforms; and consulting, integration and managed services, deployed on-premise, in the cloud or across hybrid estates for enterprise and government workloads.

Scope is set by what regulators now require of data in use, not just data at rest. The EU’s NIS2 directive and the EU Data Act and Data Governance Act’s rules on data sharing and sovereignty push regulated European workloads toward hardware-rooted isolation, and GDPR’s data-minimisation principle extends that pressure to any processor handling EU personal data. China’s PIPL adds a parallel, localisation-driven mandate for multinationals running workloads in-country. In the US, CCPA/CPRA obligations and the SEC’s cyber-incident disclosure rules give public companies a reason to show data protection controls at the processing layer, not only in storage; vendors typically map enclave deployments to the NIST CSF and SP 800-53 control families to satisfy that evidentiary burden. The EU AI Act’s rules for high-risk AI systems are now extending the same logic to training and inference workloads that touch personal or proprietary data.

Excluded are general-purpose cloud security and encryption-at-rest or in-transit products lacking hardware-rooted isolation, and standard server CPUs without enclave capability, since neither meets the trusted-execution boundary this market defines.

Growth Drivers and Restraints

DORA compliance deadlines are pulling EU financial institutions toward hardware-rooted isolation

The EU’s Digital Operational Resilience Act, Regulation (EU) 2022/2554, has applied to banks, insurers and their ICT third-party providers since 17 January 2025, mandating ICT risk-management frameworks and systematic third-party oversight. A 2025 Confidential Computing Consortium study found that 77% of adopters cited regulatory requirements such as DORA as a reason to evaluate confidential computing. The effect concentrates in the Software segment, where attestation and orchestration platforms let banks demonstrate hardware-based data-in-use protection to supervisors without re-platforming core systems.

Loosened US export licensing is expanding the advanced-accelerator supply cloud operators can deploy

The US Bureau of Industry and Security’s rule at 91 FR 1684, effective 15 January 2026, moved licence review for NVIDIA H200- and AMD MI325X-class chips from a presumption of denial to case-by-case approval, conditional on exporter certifications and independent US-based performance testing, and is reported to carry a 25% tariff or revenue-share arrangement. The change governs which accelerators GPU cloud operators can install, directly widening the Hardware segment’s supply of enclave-capable processors available to build confidential virtual-machine instances at scale.

Federal and enterprise procurement baselines are bundling attestation into standard security stacks

FedRAMP authorization requirements for federal cloud procurement, alongside SOC 2 Type II and ISO/IEC 27001 certification expectations from enterprise buyers, increasingly require providers to document hardware-rooted isolation as part of standard due-diligence checklists rather than a specialist add-on. NIST SP 800-53 control baselines reference confidential-computing-style safeguards for data-in-use protection. The mechanism pulls demand into the Services segment, where consulting and implementation teams integrate attestation tooling into existing identity and access-management stacks for large enterprise buyers migrating regulated workloads to Cloud deployment.

Legacy-estate re-architecture cost is slowing on-premise migration

Moving an application into a trusted execution environment typically requires re-compiling or re-architecting it to run inside a memory-isolated enclave, and the BIS export rule’s requirement for independent US-based performance testing adds a further certification cost layer to hardware sourcing. Regulated institutions running large on-premise estates absorb this cost most, slowing their shift out of On-Premise deployment.

DORA’s compliance timeline is outpacing available security-engineering capacity

DORA’s third-party ICT risk-management obligations, binding since 17 January 2025, compete for the same limited pool of security engineers now needed to configure attestation and key-management tooling across EU financial entities. Mid-market buyers without dedicated platform teams absorb this bottleneck most, deferring rollouts to the Managed Services sub-segment rather than building in-house capability.

Market Trends

Hyperscaler AI-deployment investment is shifting spend toward services

Microsoft launched Microsoft Frontier Company on 2 July 2026, an AI-implementation business backed by a USD 2.5 billion commitment and roughly 6,000 engineers embedded with clients including London Stock Exchange Group and Unilever, two days after Amazon committed USD 1 billion to a similar initiative. Bundling forward-deployed engineering with governance and attestation work is pulling enterprise budgets toward the Services segment, ahead of standalone Hardware or Software purchases, through the forecast period.

Export-policy easing is redrawing where confidential cloud capacity gets built

The BIS’s January 2026 rule change, moving H200- and MI325X-class accelerator licensing from presumption of denial to case-by-case review, is reshaping which regions can host GPU-backed confidential virtual machines and under what certification burden. Cloud operators serving China and Macau now plan accelerator placement around exporter certification and third-party testing requirements rather than blanket denial, a planning shift that favors Public Cloud capacity build-out in jurisdictions with clearer licensing paths.

DORA-style resilience rules are standardizing attestation as a procurement checkbox

DORA’s application across 20 categories of EU financial entity and their ICT providers since 17 January 2025 has made systematic third-party ICT risk management a standing requirement rather than a periodic audit exercise. A 2025 Confidential Computing Consortium study recorded 77% of adopters citing rules of this kind as grounds to evaluate confidential computing, pushing attestation evidence into routine vendor questionnaires and shifting the Software segment toward always-on compliance reporting.

Segment Analysis

By Component

• Hardware (largest) – Processors and chips with built-in trusted execution environments, such as secure enclaves, that isolate and encrypt data while it is being processed

• Central Processing Units (CPUs)
• Hardware Security Modules (HSMs)
• Trusted Platform Modules (TPMs)
• Software – Development kits, attestation tools, and orchestration platforms that let applications run inside and verify hardware-based secure enclaves

• Confidential Computing Platforms
• Confidential Computing Solutions
• Services – Consulting, integration, and managed offerings that help organizations design, deploy, and maintain confidential computing environments within their systems
• Professional Services

• Consulting Services
• Implementation & Integration Services
• Support & Maintenance Services
• Managed Services

Hardware leads the Confidential Computing Market in 2025, ahead of software and services. Enterprises must first procure processors and modules built with trusted execution environments, central processing units, hardware security modules and trusted platform modules, since enclave isolation and encryption of data in use are properties of the chip rather than the application layered on top of it; software and services can only attach once that hardware foundation is in place. Services is growing fastest, pulled by forward-deployed implementation programmes such as Microsoft’s Frontier Company and Amazon’s parallel enterprise AI unit, which bundle governance and integration work around confidential workloads rather than sell hardware alone.

By Deployment

• On-Premise – Confidential computing infrastructure deployed within an organization’s own data centers using dedicated trusted execution environment hardware under direct operator control
• Cloud (largest) – Confidential computing capabilities offered as a managed service through public cloud providers using their hosted trusted execution environment instances and virtual machines

• Public Cloud
• Private Cloud
• Hybrid – A deployment approach combining on-premise confidential computing hardware with cloud-based trusted execution environments to process sensitive workloads across both environments

Cloud leads deployment in 2025, ahead of on-premise and hybrid models. Hyperscalers have made confidential virtual machines a default rather than a premium option: Azure runs AMD SEV-SNP-based DCasv6 and ECasv6 confidential VMs across 57 regions, and Google Cloud offers its Confidential G4 instances at the same price as standard compute across every consumption model it sells, removing the cost penalty that once discouraged enclave adoption. Hybrid is growing fastest, pulled by data-sovereignty rules such as the EU Data Act, which pushes organisations to keep regulated workloads processing on controlled on-premise hardware while extending the same attestation model into public cloud regions for burst capacity and disaster recovery.

Regional Analysis

North America

North America held 42.0% of the market in 2025, worth USD 10.18 Billion.

Asia Pacific

The region took 29.0% of 2025 revenue, or USD 7.03 Billion.

Europe

The third-largest regional market, Europe accounted for 21.0% in 2025 and USD 5.09 Billion.

Competitive Landscape

The Confidential Computing Market is led by a group of established hyperscale and semiconductor vendors rather than a single dominant player. Competition centers on control of trusted-execution-environment silicon, breadth of confidential-VM coverage across cloud regions, price parity with standard compute, and the depth of attestation and key-management tooling enterprises need to trust enclaves with regulated data. Microsoft Corporation, Amazon Web Services, Google LLC, IBM Corporation, Intel Corporation, Advanced Micro Devices, Arm Holdings, NVIDIA Corporation, Fortanix and Anjuna Security compete across this hardware-to-software stack.

Microsoft advanced Azure confidential computing in June 2026, demonstrating Confidential Live Migration for Intel TDX virtual machines at Build 2026 and running AMD SEV-SNP-based DCasv6 and ECasv6 confidential VMs across 57 Azure regions, widening the operational footprint available to regulated workloads. Google Cloud confirmed in June 2026 that its Confidential G4 virtual machines carry no separate confidential-computing price premium, available On Demand, via Reservations, Dynamic Workload Scheduler Flex Start and Spot, removing a cost barrier that had slowed enclave adoption for AI inference workloads. IBM introduced LinuxONE 5 in May 2025, a platform built to combine confidential and quantum-safe computing while processing up to 450 billion AI inference operations per day, anchoring IBM’s roadmap around long-horizon cryptographic resilience rather than enclave performance alone.

Strategic Outlook

The clearest whitespace sits in hybrid deployment for regulated European workloads, where system integrators and managed-service providers stand to capture the integration work the EU Data Act’s portability mandate creates. That opportunity depends on cross-vendor attestation, Intel TDX, AMD SEV-SNP and Arm CCA verified through a common trust model, maturing enough that enterprises can move workloads between on-premise and cloud enclaves without re-certifying each time.

By 2035, the component mix should tilt further toward services as hyperscalers standardise confidential-VM pricing at parity with ordinary compute, turning the remaining purchase decision into implementation, governance and attestation work rather than a hardware evaluation.

Confidential Computing Market Report Scope