What challenges did Proxima Fusion face before AI implementation?

Running complex physics simulations for stellarator design was slow and required expert input to interpret results and relaunch simulations. The team needed a way to multiply the number of simulations a user could orchestrate while preserving control over proprietary engineering workflows.

What was the impact of the AI solution?

AlphaClaw is in live beta with 30 researchers. The company says the system dramatically accelerates the design timeline. Proxima Fusion plans to expand the rollout to 80 engineers.

Source

GCPEvidence: Low40/100

Proxima Fusion builds engineering agents for fusion simulations with Gemini Enterprise Agent Platform

Use case typeTraining simulationUpdated 5 days ago

Proxima Fusion, a Munich-based fusion energy company, is using Google Cloud to accelerate the design of stellarators for clean fusion energy. The company launched project AlphaClaw to multiply the number of simulations engineers can orchestrate and reduce the manual effort needed to interpret and relaunch runs.

Organization: Proxima Fusion
Industry: Energy & Utilities
Location: Germany
Published: July 2026

Reported outcomes

Strategic outcomes

Speed & agilityAccelerated the design timelineScale & capacityExpanded simulation capacity

Why do we believe this deployment?Customer identity, provider attribution, maturity, and source checks

Customer: Proxima Fusion
Provider: GCP
Maturity: Unknown
Linked source: Google Cloud Customers

No explicit deployment-stage evidence found.

Customer identity supportedSource describes one deploymentMaturity evidence evaluated

Primary read

Use case focus

Showing 2 of 2

1Training simulation
2Workflow orchestration

Running complex physics simulations for stellarator design was slow and required expert input to interpret results and relaunch simulations.
The team needed a way to multiply the number of simulations a user could orchestrate while preserving control over proprietary engineering workflows.

Proxima Fusion built a Slack-based engineering agent called AlphaClaw.
Natural-language requests in Slack trigger Cloud Run, Pub/Sub queues the workload, and Compute Engine runs long simulations.
Gemini Enterprise Agent Platform drives the automation using Gemini models via the Gemini CLI in headless mode.
The agent works with internal code repositories, Cloud SQL, and Cloud Storage, and posts charts, plots, and pull requests back into Slack.
A secondary reviewer agent inspects intended actions and pauses runs that look unfamiliar or risky until a human approves them.

Technologies

Gemini Enterprise Agent Platform Gemini Cloud Run Pub/Sub Compute Engine Cloud SQL Cloud Storage Gemini CLI

AlphaClaw is in live beta with 30 researchers.
The company says the system dramatically accelerates the design timeline.
Proxima Fusion plans to expand the rollout to 80 engineers.

Architecture

Slack-based agentic simulation workflow. Natural-language requests enter through Slack, Cloud Run acts as a gateway, Pub/Sub queues jobs, Compute Engine executes long-running simulations, and Gemini Enterprise Agent Platform runs Gemini models in headless mode with access to internal repositories, Cloud SQL, and Cloud Storage. A second reviewer agent checks planned actions and requests human approval for unfamiliar or risky steps.

Sources & evidence1

Evidence: Low40/100Evidence strength