Cowork shifts AI inference and VM to cloud, easing battery drain for users

Cowork announced on October 5, 2026 that its new release runs both model inference and the supporting virtual machine entirely in the cloud, eliminating the local disk and battery overhead that plagued the previous version. The change means a user can close a laptop without halting work, and the desktop app now mediates any file‑access tool calls.

Key points

• Old Cowork used a cloud‑hosted model but a local Anthropic‑provided VM.
• Users complained about disk usage, battery drain, and interrupted sessions.
• New Cowork places inference and VM in the cloud, giving each session an isolated sandbox.
• The desktop client now handles device‑side file access on demand.
• The shift aims to support phone usage and continuous work without local resource penalties.

What happened

The original Cowork architecture executed model inference in the cloud while delegating tool calls to a VM shipped to the user’s computer. Simon Willison’s quotation notes that “People loved what they were able to do with Claude but didn't love the disk, battery, and performance cost of running the VM locally.” Feedback also highlighted that closing a laptop stopped all work, a limitation for mobile professionals.

In response, the development team re‑engineered the platform so that both inference and the VM now reside in the cloud. Each user session receives its own sandboxed environment, preventing state sharing across sessions. When the VM requires a file from the user’s device, the desktop application performs the file‑access tool call, preserving security while offloading compute.

The team believes the redesign resolves several pain points. The quotation states, “We think this solves a lot of problems we've heard about (like using Cowork from a phone, keeping work running, or getting all the same power without losing battery to the VM).” The announcement was posted on Simon Willison’s blog on October 5, 2026.

The backstory

Cowork’s early adoption hinged on its ability to integrate Claude’s large‑language‑model capabilities with a flexible tool‑calling interface. The hybrid model—cloud inference plus a local VM—offered high performance but introduced resource constraints on laptops and desktops. As remote work and mobile computing grew, users increasingly demanded a solution that would not sacrifice battery life or require a constantly open laptop.

Industry trends have shown a shift toward fully cloud‑native AI services, driven by economies of scale and the desire to simplify client‑side footprints. Cowork’s move aligns with this broader migration, echoing similar strategies from competing platforms that have abstracted away local execution environments.

Technical trade‑offs

Moving the VM to the cloud reduces local CPU and storage demands, but it introduces latency for any file‑access operations that must round‑trip through the desktop client. The sandboxed sessions improve isolation, yet they rely on robust network connectivity to maintain responsiveness. Security benefits stem from centralizing execution, but the model now depends on the provider’s cloud security posture.

Developers gain a more consistent environment across devices, but they must adapt to a new workflow where the desktop app mediates all device‑side interactions. Monitoring and debugging may shift from local logs to cloud‑based telemetry, requiring updated tooling.

Market implications

The redesign positions Cowork as a more mobile‑friendly AI assistant, potentially expanding its appeal to enterprises that prioritize BYOD policies and remote work continuity. Competitors that still depend on local VMs may lose traction among users sensitive to battery life and device performance.

However, enterprises with strict data‑residency requirements might view the cloud‑only approach as a risk, especially if the sandbox cannot guarantee on‑premise data handling. Providers that can offer hybrid options—local execution for sensitive workloads and cloud execution for general tasks—could capture the niche that Cowork now vacates.

What builders and businesses should do

• Evaluate whether your workloads can tolerate the added network latency of cloud‑based VM calls.
• Update security assessments to reflect the shift from local to cloud execution.
• Adjust monitoring pipelines to capture cloud‑side telemetry for debugging.
• Consider mobile deployment strategies now that Cowork supports phone usage without local VM overhead.
• Review data‑residency policies to ensure compliance with the new cloud‑only model.

The bottom line

Watch for Cowork’s performance metrics after rollout; any latency spikes or security incidents will shape adoption curves in the next quarter.