Why confidential computing is essential for enterprise AI

Confidential computing is emerging as a foundational technology for enterprise AI by protecting data in use through hardware-based trusted execution environments, according to an analysis of the security gap that leaves decrypted in-memory data vulnerable to malicious insiders and cyberattacks. The piece argues that as AI workloads require access to larger and more sensitive datasets — customer records, intellectual property, financial and healthcare data — organizations face a tradeoff between accelerating AI adoption and protecting their most critical information. It ties the trend to sovereign AI requirements from governments and public sector organizations seeking to leverage AI while maintaining strict sovereignty controls.

Artificial intelligence has entered a new phase. For most large organizations, the conversation is no longer whether AI can create business value. It is how quickly AI can be deployed across the enterprise while maintaining security, governance, compliance, and control. From customer service and software development to drug discovery and financial modeling, AI is increasingly being applied to organizations’ most valuable assets: their data. The quality of AI outcomes depends directly on the quality, breadth, and sensitivity of the information used to train, fine-tune, and run models. Yet many enterprises face a fundamental dilemma. The data that creates the greatest business value is often the same data that carries the highest levels of risk. Customer records, intellectual property, financial data, healthcare information, government data, and operational insights cannot simply be exposed to new platforms or shared broadly without appropriate safeguards. As a result, many organizations find themselves caught between the desire to accelerate AI adoption and the need to protect their most critical information. This is where confidential computing is rapidly emerging as a foundational technology for the AI era. For decades, enterprise security strategies have focused on protecting data in two key states: at rest and in transit. Data stored on systems can be encrypted. Information moving across networks can be secured through encrypted communications channels. These protections have become standard practice across industries. But a third state has historically been more difficult to protect: data in use. Whenever applications process information, data must traditionally be decrypted in memory so systems can perform calculations. During this period, sensitive information can become vulnerable to threats ranging from malicious insiders to sophisticated cyberattacks. In an AI-driven world, this challenge becomes even more significant. AI workloads often require access to larger datasets, broader collaboration across teams and partners, and increasingly complex computing environments. The more valuable the workload, the more attractive it becomes as a target. Confidential computing addresses this longstanding gap by protecting data while it is being actively processed. Using hardware-based trusted execution environments, applications and data can remain isolated and protected, ensuring sensitive information can be analyzed without exposing it unnecessarily. For organizations pursuing enterprise-scale AI, this capability is becoming increasingly important. Traditional business applications generally operate within well-defined boundaries. AI changes the equation. Modern AI systems increasingly depend on access to proprietary information that differentiates one organization from another. Public models and publicly available datasets can provide a starting point, but competitive advantage often comes from combining AI with unique enterprise knowledge. A pharmaceutical company may need AI models to analyze proprietary research. A manufacturer may want to use operational data from factories worldwide. A financial institution may need to process highly sensitive client information. Governments and public sector organizations may seek to leverage AI while maintaining strict sovereignty requirements. In each case, AI value is directly tied to data sensitivity. The challenge for CIOs and business leaders is clear: if organizations cannot safely use their most valuable information, they limit AI’s potential. If they loosen security controls to gain AI benefits, they introduce unacceptable risk. Confidential computing helps resolve this tension by enabling organizations to use sensitive data more confidently and at greater scale. At the same time AI adoption is accelerating, another major trend is reshaping technology strategies: sovereign AI. Organizations across industries and geographies are increasingly focused on maintaining control over their data, models, infrastructure, and operations. Regulatory requirements continue to evolve. Data residency concerns are growing. Intellectual property has become a strategic asset. Governments and enterprises alike are seeking greater assurance regarding where data resides, who has access to it, and how it is used. For many leaders, sovereignty is no longer simply a compliance issue. It is a business imperative. Successful AI strategies now require organizations to balance innovation with governance. They must be able to move quickly while maintaining visibility and control. Confidential computing plays a critical role in this equation because it strengthens protections around sensitive workloads regardless of whether they are running in an on-premises environment, a private cloud, a colocation facility, or a hybrid architecture. By helping ensure that data remains protected throughout processing, organizations gain stronger assurances around security and governance without sacrificing agility. As sovereign AI moves from concept to implementation, confidential computing is becoming a key enabling technology. The next generation of AI will increasingly depend on collaboration. Healthcare organizations may seek to combine research data across institutions. Financial firms may need to share insights while protecting customer privacy. Supply chain partners may want to use AI models that leverage information from multiple participants without exposing proprietary details. Historically, these scenarios have been difficult to enable because collaboration often required organizations to relinquish direct control over sensitive information. Confidential computing introduces new possibilities by creating protected environments where data can be processed while remaining shielded from unauthorized access. This allows organizations to collaborate with greater confidence, helping unlock value that would otherwise remain trapped in isolated data silos. For many industries, this capability could become one of the most important accelerators of AI innovation over the next decade. One of the long-standing challenges in cybersecurity has been balancing protection with operational efficiency. Historically, stronger security controls were often associated with increased complexity, higher costs, or reduced performance. Enterprise AI changes the requirements dramatically. Organizations need environments capable of supporting large-scale training, fine-tuning, and inference workloads while simultaneously meeting stringent security expectations. Security can no longer be treated as an overlay added after deployment. It must be integrated into the architecture itself. This is why the industry is increasingly focusing on secure-by-design AI infrastructure. Rather than forcing organizations to choose between performance and protection, today’s confidential computing solutions are designed to help enable both. This approach becomes particularly important as organizations deploy AI workloads that support mission-critical functions, revenue-generating operations, and strategic decision-making. As enterprise AI adoption expands, organizations need more than individual technologies. They need integrated solutions that combine accelerated computing, secure infrastructure, data management, and operational oversight into a cohesive platform. Together, HPE and NVIDIA are helping organizations build the secure AI environments required for this new generation of workloads. By combining NVIDIA’s accelerated computing technologies with HPE’s expertise in enterprise infrastructure, supercomputing, security, and hybrid architectures, organizations can deploy AI solutions designed to address performance, governance, and sovereignty requirements simultaneously. Confidential computing forms a critical component of this strategy. It helps enable organizations to process sensitive information with stronger protections while supporting the scale and performance demanded by AI workloads. This becomes especially important for industries where trust is non-negotiable, including financial services, healthcare, government, defense, telecommunications, and critical infrastructure. For these organizations, security is not simply about risk reduction. It is a prerequisite for innovation. The most successful AI initiatives of the coming years will not be determined solely by model sophistication or computing power. They will be defined by trust. Customers want assurance that their data is protected. Regulators expect stronger governance. Boards demand resilience and risk management. Business leaders need confidence that AI systems can operate securely at scale. Confidential computing helps address these requirements by extending protection to one of the last major vulnerabilities in the data lifecycle: information being actively processed. As a result, organizations can move beyond viewing security as a constraint and begin treating it as an enabler of growth. The ability to securely use sensitive data can unlock new AI applications, accelerate innovation, enable broader collaboration, and support stronger sovereign AI strategies. Organizations that establish this foundation early may gain a significant advantage as AI becomes increasingly central to business operations. The organizations creating the greatest value from AI are those capable of combining powerful infrastructure, trusted governance, and access to high-quality proprietary data. Confidential computing sits at the intersection of all three. By protecting data in use, enabling secure collaboration, supporting sovereign AI objectives, and helping organizations maintain control over their most valuable information, confidential computing is becoming a foundational technology for the AI era. The future of AI will not simply belong to organizations that can generate the most intelligence. It will belong to those that can do so with the highest levels of trust, security, and control. In that future, confidential computing is not just a security feature. It is the foundation that makes enterprise AI possible. Learn how HPE and NVIDIA are helping organizations design secure, sovereign AI environments that protect sensitive data while accelerating innovation at hpe.com/ai https://www.hpe.com/ai . As AI becomes increasingly central to economic competitiveness, scientific advancement, and national priorities, organizations require infrastructure that balances performance with security and sovereign control. Together, HPE and NVIDIA co-engineer rack-scale AI systems that integrate AI computing, high-performance networking, and supercomputing expertise to support large-scale AI workloads. This provides enterprises, governments, and research institutions with a trusted foundation for sovereign AI initiatives while maintaining control over critical data, models, and operations.