Secure AI at Scale: How Corvex and Intel Enable Confidential Computing with Intel® Trust Domain Extensions and Intel® Trust Authority for Security-First Environments

Rodolfo Leffa, Project Manager, Corvex

Paul O’Neill, Confidential Computing Lead, Intel

Sathi Nair, Product Manager Intel Trust Authority, Intel

‍

Secure AI at Scale: How Corvex and Intel Enable Confidential Computing with Intel® Trust Domain Extensions  and Intel® Trust Authority for Security-First Environments

‍

As AI workloads reach unprecedented scale, a new class of risk is emerging: protecting sensitive data, proprietary models, and regulated workloads in highly complex AI computing environments. For enterprises in healthcare, financial services, government, and IP-intensive segments like AI model builders, traditional cloud security controls are no longer sufficient.

Confidential Computing architectures address the problem of security in AI computing environments and are rapidly becoming foundational requirements, not optional enhancements, for secure AI infrastructure.

Corvex and Intel are deploying confidential AI infrastructure that supports secure, verifiable, and privacy-preserving AI execution at cloud scale, enabling enterprises to run sensitive and regulated workloads with strong security assurances and establishing the foundation for trusted hybrid AI deployments.

 “Intel® Trust Domain Extensions and Intel® Trust Authority strengthen Corvex’s confidential compute offering by delivering hardware-rooted Trusted Execution Environments that isolate entire virtual machines from the host, hypervisor, and other tenants. Combined with Corvex’s secure single-tenant GPU clusters and NVIDIA Confidential GPU modes, Intel TDX and ITA enable end-to-end encrypted AI workflows from CPU orchestration through GPU execution,” said Seth Demsey, Co-Founder and Co-CEO of Corvex.

Confidential Computing: Critical Security for AI Workloads

Confidential computing focuses on protecting data in use, complementing traditional encryption methods that safeguard data at rest and in transit. While encryption has long secured stored data and network communications, sensitive information has traditionally remained exposed during processing, creating a critical security gap. Confidential computing addresses this gap by leveraging Trusted Execution Environments (TEEs)— isolated CPU and GPU execution environments in which data and application code have an additional layer of security throughout computation. TEEs are established using hardware-based security technologies such as Intel® Trust Domain Extensions (TDX), which provide strong memory encryption, isolation, and when combined with Intel Trust Authority adds strong cryptographic attestation. This added layer of security control supports protection of sensitive workloads even from privileged system software, cloud administrators, and malicious insiders.

By enforcing a hardware-rooted trust boundary, TEEs establish that only authorized and verified code can access sensitive data, enabling secure, verifiable, and privacy-preserving computation in shared and cloud-based environments. Integrating Intel® Trust Domain Extensions (TDX) and NVIDIA Confidential Computing with remote attestation, Corvex enables cryptographic verification of system integrity and protects data in use while maintaining performance, scale, and operational reliability. For proprietary AI workloads in cloud environments, confidential computing from Corvex provides the same security assurances previously available only in isolated, on-premises infrastructure.

Intel® Trust Authority: Independent Verification at Scale

Intel® Trust Authority (ITA) is a SaaS-based independent, third-party remote attestation service, that provides cryptographic verification of TEE at cloud scale. ITA enables organizations to establish verifiable trust in confidential computing platforms, ensuring workloads run only in approved, hardware-secure environments, providing verification of TEEs. Built on Intel SGX and Intel TDX, ITA serves customers in healthcare, financial services, and regulated industries who require:

• Independent attestation and verification: Cryptographically verifies that workloads are executing in approved confidential computing environments, independent of cloud service providers.
• Protection of sensitive data and AI models
• Privacy preservation of enterprise data, prompts, and model IP throughout execution in external cloud environments.
• Auditable proof of trust and compliance: Provides verifiable evidence that confidential computing environments are correctly configured and enforced — supporting regulatory compliance and audit requirements.
• Cloud-agnostic trust verification: Enables a consistent trust model across multi-cloud and hybrid environments.
• Composite attestation for confidential AI: Supports end-to-end attestation across CPU and GPU domains, including verification of Intel® TDX confidential virtual machines and NVIDIA GPUs enabling trusted, accelerated confidential AI pipelines

Performance, Reliability, and Security at Scale: The Corvex Platform with Intel® TDX and Intel® Trust Authority

Integrating Intel TDX and NVIDIA Confidential Computing with remote attestation, Corvex enables cryptographic verification of system integrity and protects data in use while maintaining performance, scale, and operational reliability. For demanding training and inference workloads in cloud environments, confidential computing from Corvex provides the same security assurances previously available only in isolated, on-premises infrastructure. Benefits of the Corvex platform include:

• Confidential AI execution using Intel® TDX and NVIDIA Confidential Computing, enabling hardware-isolated workloads with remote attestation
• Optimized factory-scale GPU clusters with latest-generation NVIDIA GPUs, purpose-built for training and inference
• Managed Kubernetes optimized for AI workloads, simplifying deployment and lifecycle management
• Tier III data center architecture with redundant power, AI-native networking, and advanced cooling
• 24×7 on-site expert operators for root-cause analysis across drivers, PCIe, networking, and firmware layers

Real-World Applications

Corvex’s secure AI Factory program aims to serve diverse sectors including AI product and model builders, technologically sophisticated enterprises, federal agencies, healthcare, and financial services organizations. Its platform delivers high price-performance by maximizing compute density while maintaining elasticity and allowing flexibility for burst capacity during peak demand periods. Popular use cases include:

• Government & Critical Infrastructure: Public-sector agencies require demonstrable assurance that sensitive workloads cannot be accessed or tampered with. Corvex's platform, integrated with Intel® Trust Authority, enables independent cryptographic verification of TEEs, supporting zero-trust principles and auditability for defense and critical infrastructure use cases.
• Regulated AI Inference and Training: Corvex enables healthcare, financial services, and life sciences organizations to run AI workloads at scale while maintaining strict data confidentiality. Hardware-enforced TEEs protect models and datasets during execution, with remote attestation providing cryptographic proof of secure environments, allowing production AI deployment without exposing IP or regulated data.
• Enterprise & SaaS Multi-Tenant Platforms: Corvex combines factory-scale GPU clusters with hardware-isolated execution, enabling strong workload isolation for multi-tenant platforms. Organizations can scale shared AI infrastructure while maintaining tenant separation and protecting proprietary models without sacrificing performance.
• Collaborative Research (Future Roadmap): Corvex plans to support collaborative AI research where institutions contribute encrypted datasets to shared training workflows without exposing raw data. Hardware-enforced isolation enables cross-institution research in healthcare, life sciences, and federal programs where data sharing is constrained by privacy or sovereignty requirements.

Conclusion

Together, Corvex and Intel Confidential Computing establish a new standard for secure AI infrastructure, combining factory-scale performance, hardware-enforced isolation, and independent verification to enable confidential AI execution at cloud scale.

Now, organizations deploying AI in regulated, sovereign, or IP-sensitive environments can leverage Corvex’s confidential AI platform, built on Intel® Trust Domain Extensions (TDX) and Intel® Trust Authority, to run secure, verifiable, and privacy-preserving AI workloads with confidence.

‍