Google Cloud Warns AMD Turin CPU Side‑Channel Flaws Threaten VM Privacy
Google Cloud alerts customers of three AMD Turin CPU side‑channel bugs that could allow virtual machines to read each other’s memory.
On Thursday, November 12, 2026, Google Cloud announced a security advisory concerning three previously unknown side‑channel flaws in its AMD Turin CPUs.1 The company warned that the defects enable code running on the same processor core within a virtual machine (VM) to infer sensitive data from other processes—including the host operating system and the kernel—by exploiting subtle cache‑timing variations.1 Google stated it has already applied mitigations against attacks between VMs and between VMs and the host.1 New issues require additional guest‑OS patches and a stop‑and‑start reboot cycle for affected instances.1
The vulnerabilities are variants of the “cross‑VM side channel” class of attacks similar to Spectre and Meltdown. In practice, an attacker with malicious code inside a VM can correlate timing differences against known inputs and recover encryption keys or database credentials used by other applications sharing the same physical core. Because the design allows data from one process to leak into another via cache timing, the impact extends across all workloads scheduled on those cores.
Google recommends customers apply the latest guest‑OS updates immediately and perform a full restart of their VMs to flush caches completely.1 These steps mitigate risk but also introduce measurable performance overhead: depending on workload intensity users may experience 20–40% slower execution while the caches remain flushed. Mitigation mirrors earlier infrastructure‑wide software updates required after Spectre‑type disclosures underscoring how modern cloud CPUs must balance security hardening with operational efficiency.
Flaws could affect any customer running AMD Turin chips in its public‑cloud environment regardless of operating system version.1 The new bugs expose gaps that allow privileged code within one VM to read data from another—an outcome that directly undermines privacy guarantees promised by virtualization.
Google’s response aligns with industry best practices for dealing with speculative execution vulnerabilities: issue rapid patches, advise customers restart VMs, and transparently communicate performance trade‑offs.1 The company urged vigilance given the high potential impact on sensitive workloads such as encryption key management or proprietary database operations.
In summary, Google Cloud has identified three AMD Turin CPU side‑channel defects capable of leaking data across processes inside virtual machines. Customers must apply guest‑OS updates, reboot their instances, and accept possible 20–40% performance reduction while caches remain flushed. This development reaffirms how even state‑of‑the‑art CPUs can harbor subtle vulnerabilities that threaten tenant isolation in cloud infrastructures.
Made by Ren, the ArRENCE AIChat with Ren, plan your day, generate images, and more.
Meet Ren →