Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Scan for outdated or missing drivers - takes under a minute3Repair Windows errors before they cause bigger problemsSome links on this page are affiliate links: if you buy through them we may earn a commission, at no extra cost to you.
A study commissioned by the Confidential Computing Consortium (CCC) and conducted by IDC reports that 75% of surveyed organizations are adopting confidential computing. But that figure includes organizations still piloting or testing: 57% said they were in that stage, while 18% reported production use. The results, published December 3, 2025, suggest growing interest in protecting sensitive data during processing—especially for AI—not that three-quarters of organizations have deployed the technology at scale.
What the study found—and what “adopting” means
The CCC, a Linux Foundation project community, commissioned IDC to survey more than 600 IT leaders across 15 industries for Unlocking the Future of Data Security: Confidential Computing as a Strategic Imperative. The public announcement says the study examined adoption, use cases, benefits, barriers and regulatory influences. Its headline adoption figure combines two distinct stages: 57% of respondents said their organizations were piloting or testing confidential computing, and 18% said they had it in production.
Those figures make confidential computing a technology many organizations are evaluating, but they do not show that 75% have it deployed in production, or that production use is broad across each organization. The public summary does not provide the full questionnaire, sampling frame, respondent-selection or weighting methods, response rate, or independent replication. The findings should therefore be read as IDC survey results commissioned by an industry consortium that promotes the technology—not as an independently established industry census.
Recommended Free Tools
Among respondents, 88% cited improved data integrity as a primary benefit, 73% cited confidentiality with technical assurances, and 68% cited improved regulatory compliance. These are reported perceptions, not measured proof that the technology improves integrity or compliance by those percentages. The summary also identifies accelerated innovation and cost efficiency as reported outcomes.
#1 Best Overall
What confidential computing protects
Security controls traditionally focus on data at rest (stored) and in transit (moving across a network). Confidential computing aims to protect data in use: while software is processing it. It typically uses hardware-backed trusted execution environments (TEEs) to isolate selected code and data from some parts of the surrounding system, such as a hypervisor, host administrator or other workloads. The exact protection depends on the hardware, service design and threat model.
Common components include memory encryption, secure or measured boot, remote attestation and policies that release encryption keys only to a workload running in an approved state. Implementations may isolate an application, a virtual machine, a container or, in some configurations, accelerator processing. Attestation is important because encryption alone does not tell a data owner what code is running. A verifier checks evidence about the hardware and workload against an approved policy; a key service can then release secrets only if the evidence meets that policy.
This adds a protection layer; it does not replace encryption at rest or in transit, identity and access controls, secure development, patching, endpoint protections or governance. Nor does a TEE make the software inside it trustworthy by default. A vulnerability, malicious update or compromised build pipeline can still put sensitive data at risk.
Why AI is sharpening the question
AI workloads concentrate valuable and sensitive assets. Training data may include medical records, financial information or trade secrets. Inference requests can expose personal details or business plans. Model weights may be valuable intellectual property. AI agents may also receive broad permissions and process data with less direct human involvement than a conventional application.
Running workloads in a public cloud can raise questions about who could access plaintext in memory or influence the host environment. Confidential computing can help address some of those concerns when the design protects the relevant parts of the workload and has a credible attestation and key-release process. It is also relevant when organizations want to analyze shared data without giving one another unrestricted access to the underlying records.
The study identifies secure model training, confidential inference, AI agents working with regulated data, privacy-preserving analytics and cross-organization data collaboration as use cases respondents are considering or pursuing. These are plausible fits, but protection must cover the actual data path. A confidential CPU-side VM does not automatically protect data once it reaches an unsupported accelerator, and a TEE does not stop a model from revealing sensitive information in its outputs.
Confidential computing also does not solve prompt injection, data poisoning, hallucinations, excessive agent permissions, insecure application code or AI governance. It protects a defined execution boundary, not the correctness or safety of everything executed within it.
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →Adoption drivers, benefits and barriers in the survey
The CCC-commissioned IDC summary reports the following responses. Percentages reflect the survey as described in the public announcement; they are not independent deployment measurements.
| Survey finding | Reported figure | How to read it |
|---|---|---|
| Organizations adopting confidential computing | 75% | Includes 57% piloting or testing and 18% in production |
| Improved data integrity cited as a primary benefit | 88% | Respondent-reported benefit |
| Confidentiality with technical assurances cited | 73% | Respondent-reported benefit |
| Improved regulatory compliance cited | 68% | Respondent-reported benefit |
| Workload security and external threats as a driver | 56% | Survey response |
| Protection of personally identifiable information as a driver | 51% | Survey response |
| Compliance as a driver | 50% | Survey response |
| Attestation validation identified as a barrier | 84% | Survey response |
| Skills gaps identified as a barrier | 75% | Survey response |
The announcement says 77% were more likely to consider confidential computing because of DORA-related data-in-use requirements. That is a reported influence on consideration, not evidence that DORA universally mandates confidential computing. Public-cloud users were the most likely group to implement it, at 71%; hybrid or distributed-cloud users followed at 45%.
Production deployment varied by sector in the reported results: 37% in financial services, 29% in healthcare and 21% in government. The study reports full-production figures of 26% for Canada, 24% for the United States, and 20% each for China and the United Kingdom. It also found stronger interest in multi-party privacy-preserving collaboration among healthcare respondents (78%) than financial services (61%) or government (26%). These are survey figures, not verified country-level or sector-wide deployment statistics.
The leading reported barrier—attestation validation, at 84%—is especially significant. Organizations have to choose which hardware and firmware roots of trust to accept, who verifies platform evidence, which workload measurements are approved and how keys are released. They also need a process for software updates that alter measurements, failed attestations, incident investigation and audit evidence. Skills gaps (75%), the perception that the technology is niche (77%), interoperability and implementation complexity add to the work.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
When confidential computing is a strong fit
It is most compelling when the threat model includes a cloud or infrastructure operator, host-level access, or a need for parties to collaborate without fully trusting one another. Candidate workloads include inference over regulated records, cross-institution fraud detection, healthcare research, key-handling services, protection of proprietary model weights, and sensitive applications hosted in public cloud. Sovereignty or jurisdiction requirements may also make it worth evaluating, though a TEE alone does not satisfy every sovereignty obligation.
Rank #4
It may add little for public data without a meaningful confidentiality requirement, or where the central problem is weak authorization rather than host access. It can also be a poor fit for applications dependent on unsupported drivers or hardware, or those that need unrestricted debugging and host-level visibility. For large AI workloads, confirm that confidential GPU support exists for the exact accelerator, cloud, region and framework; do not assume VM-level protection extends to GPU memory or every intermediate value.
Implementation trade-offs to plan for
- Attestation and keys: Define the acceptable hardware, firmware and workload state; identify the verifier; set key-release policy; approve updates; and specify what happens when evidence fails. Keep a rollback image and a documented emergency recovery path.
- Visibility and response: Isolation can restrict provider and host visibility by design. Debugging, profiling, malware detection and forensics may become harder. Decide how to monitor the workload and investigate incidents without relying on inspection the architecture is meant to prevent.
- Residual exposure: Memory protection does not eliminate side channels or leaks through timing, access patterns, traffic volume, errors, logs, inputs or outputs. A compromised application inside the TEE can still misuse the data it is allowed to process.
- Performance and cost: Measure latency, throughput, startup time, memory needs and accelerator availability on the intended workload. Include engineering effort, attestation infrastructure, migration, audit, monitoring and rollback costs—not just a cloud surcharge.
- Portability: TEE types, attestation formats, hardware generations and management integrations differ. A solution built around one provider’s services may be difficult to move, even if the application itself is portable.
Cloud options are not interchangeable
AWS Nitro Enclaves creates isolated environments from EC2 instances. AWS documents that enclaves have no persistent storage, interactive access or external networking; communication is through a secure local connection with the parent instance. The service supports cryptographic attestation and integration with AWS Key Management Service. AWS says Nitro Enclaves carries no additional usage charge, but customers still pay for EC2 and other services. Documentation says it is supported on most Intel-, AMD- and AWS Graviton-based Nitro instances, allows up to four enclaves per parent instance, and is not supported on Outposts, Local Zones or Wavelength Zones. The limited networking and storage model can require application decomposition and enclave-specific key-management work; it is not a drop-in home for every AI workload. See AWS Nitro Enclaves documentation and AWS’s confidential computing overview.
AWS also describes protections inherent to its Nitro System as not requiring customer code changes. That should not be confused with every enclave-based design being transparent: using enclaves, attestation, key release or confidential accelerators can require architectural changes.
Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchWindows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallGoogle Cloud Confidential VM offers hardware-backed VM options, with additional charges that vary by technology and machine family. Google’s pricing page, as displayed August 18, 2026, lists on-demand surcharges for specified configurations including AMD SEV, AMD SEV-SNP on N2D and Intel TDX on C3. It lists separate pricing for confidential GPU configurations; G4 confidential-computing charges and the associated NVIDIA license fee are shown as free during preview, with charges applying after general availability. Availability and final cost depend on region, machine and accelerator. Google says Confidential Space, intended for controlled data collaboration, has no additional charge beyond the Confidential VM and other resources used. Check the current Confidential VM pricing page before budgeting; listed prices can change.
Best Value
Oracle’s sovereign-cloud principles document says customers can implement confidential-computing VMs or bare-metal servers without an extra charge on top of OCI Compute pricing. Confirm the exact region and instance availability with Oracle’s document and current service terms. Specialist platforms such as Fortanix and Anjuna add management or deployment layers around confidential workloads; they may help with lifecycle, policy or application integration, but also introduce additional vendors, costs and dependencies. Compare them against native tooling rather than assuming an abstraction removes the underlying hardware and attestation constraints.
A practical pilot plan
- Choose the asset and exposure. Inventory sensitive datasets, model weights and inference inputs. Identify who could access them today and whether the concern is a host operator, hypervisor, insider, co-tenant or cross-company partner.
- Select one bounded workload. Start with a contained case, such as inference over sensitive records or a key-handling service, rather than attempting to confidentialize an entire AI estate.
- Map the full data path. Determine where plaintext appears across CPU and GPU memory, storage, networking, logs and outputs. Confirm that the selected platform protects the relevant components.
- Define trust and attestation policy. Specify acceptable hardware and software measurements, who verifies them, which keys can be released, how updates are approved and how evidence is retained.
- Test failure and recovery. Exercise rejected attestations, changed images, firmware updates, unavailable regions, key-service outages and rollback. Set out break-glass access and audit requirements before production.
- Measure operational and economic impact. Compare latency, throughput, engineering effort, observability, cloud charges and support needs with the existing workload. Include the cost of maintaining verification and update policies.
- Expand only on evidence. Broaden deployment if the pilot meaningfully reduces a defined risk and the organization can operate the controls. Keep confidential computing as one layer in a wider security and AI-governance program.
Is it really a “strategic imperative”?
The study supports a narrower, defensible conclusion: confidential computing is moving from a specialist capability into the set of infrastructure options organizations are considering for sensitive cloud workloads, regulated AI and data collaboration. The case is strongest when the organization needs technical assurances about an infrastructure boundary or wants to compute across parties that cannot simply exchange raw data.
Whether it is an imperative depends on the workload, threat model, data sensitivity, regulatory obligations, available hardware and ability to operate attestation and key policies. The survey is evidence of interest, not proof that every organization needs the technology now—or that pilots will become successful production systems. For many teams, the sensible next move is a focused pilot with explicit security, performance and recovery criteria.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Quick Recap
Read the CCC’s announcement of the IDC study.
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
