Anthropic users on Windows 11 are reporting a persistent memory allocation issue in the latest release of Claude Desktop, where the application consistently launches an unneeded 1.8 gigabyte Hyper-V virtual machine upon startup. The issue, first documented in mid-February 2026, affects chat-only users who have no intention of utilizing the platform's Cowork or agent features. Despite VirtualMachinePlatform being the sole enabled virtualization feature on affected systems, the app triggers the Hyper-V Host Compute Service during boot and launch cycles, spawning a Vmmem process that consumes approximately 11 percent of a standard 16 gigabyte system's total RAM. Diagnostic analysis indicates that the virtual machine initialization is driven by an RPC interface event tied to stale session files accumulated from previous Cowork sessions. Investigators identified thousands of orphaned session directories containing Docker-style identifiers within the local application data folder. These residual files persist indefinitely, forcing the desktop client to provision virtual infrastructure on every boot regardless of user intent. The underlying JSON configuration errors generating in the event log suggest a failure in session state validation rather than a complete system misconfiguration. The immediate impact on end users is notable performance degradation. Idle memory utilization jumps from typical browser-level consumption to over 62 percent before any productive work begins. Combined with standard operating system and application overhead, this behavior routinely pushes total RAM usage past the 70 percent threshold, resulting in noticeable system sluggishness and forcing users to manually terminate background virtualization processes after each launch. Currently, the only stable mitigation requires users to disable the Windows VirtualMachinePlatform feature entirely, which simultaneously breaks Cowork functionality, or to script continuous termination of the vmwp and vmcompute processes. In response, affected users and technical reviewers are urging Anthropic to implement demand-based initialization for containerized agent environments. Proposed resolutions include automatic purging of expired session data upon closure and a graceful fallback to lightweight chat-only mode when virtual infrastructure is unavailable or unnecessary. The report highlights a growing friction point in desktop AI applications that rely on local virtualization for advanced features. As consumer hardware remains constrained by fixed memory allocations, efficient resource provisioning will become increasingly critical for widespread adoption of on-device AI workflows. Anthropic has not yet issued an official statement or patch timeline regarding the reported behavior.