Intel and AMD have long promoted their trusted execution environments—such as Intel’s Software Guard Extensions (SGX) and AMD’s Secure Encrypted Virtualization (SEV)—as secure enclaves that protect sensitive data and code from unauthorized access, even within compromised systems. These technologies are foundational to modern network security, used by cloud providers, financial institutions, and government agencies to safeguard encryption keys, authentication credentials, and private workloads. However, recent research has revealed a critical vulnerability: these trusted enclaves can be compromised through physical attacks. Researchers demonstrated that by using side-channel techniques—such as power analysis, electromagnetic probing, and timing attacks—attackers with physical access to a machine can extract secrets from the enclaves, effectively bypassing the security guarantees they were designed to provide. The attacks exploit subtle variations in how processors consume power or emit electromagnetic signals during computation. By monitoring these signals with specialized equipment, attackers can infer what data is being processed inside the enclave, even if the data remains encrypted in memory. In some cases, attackers have successfully extracted cryptographic keys and decrypted sensitive information. Despite the severity of these findings, both Intel and AMD have maintained that physical attacks are not within their threat model. They argue that trusted enclaves are designed to protect against software-based threats and remote exploits, not against adversaries with direct, physical access to hardware. As such, they recommend that users protect their systems from physical tampering through environmental controls, secure facilities, and access restrictions. But many users—especially those in cloud environments—did not receive this message. In practice, cloud providers often deploy servers in shared data centers where physical security may not be fully under the control of individual customers. This creates a gap between the perceived security of trusted enclaves and the reality of their exposure to physical threats. Security experts warn that relying solely on software-level protections without considering physical attack vectors undermines the entire security architecture. They urge organizations to adopt layered defenses, including hardware-based tamper detection, secure boot mechanisms, and continuous monitoring for anomalies in power and signal patterns. The revelations have sparked renewed debate about the limits of hardware-based security and the need for more robust threat modeling. As AI, cloud computing, and data privacy become increasingly central to digital infrastructure, the assumption that “trusted” hardware is immune to physical compromise may no longer be sustainable.