AI news · October 7, 2026
Clockwork raises $31 million to reduce lost GPU training time
Clockwork is selling recovery from GPU-cluster failure, not another model. Its LinkPass and TorchPass products target the expensive gap between a failed node and a useful resumed run.
- new funding
- $31M
- reported goodput loss
- 28.5% to 10.3%
- GPU cluster in the test
- 4,096
Clockwork announced a $31 million funding round on October 5 and said its resilience software is in production at LinkedIn and Together AI, with WhiteFiber expanding its use. LinkPass is designed to reroute around failed network links, while TorchPass handles migration and recovery for training workloads. The company also described a multi-node platform that can snapshot work without code changes and support asynchronous checkpoints for reinforcement learning.
Clockwork says LinkedIn avoids tens of thousands of GPU-hours of downtime each month. On its own site, it reports reducing goodput loss from 28.5 percent to 10.3 percent on a 4,096-GPU cluster. The result will depend on the cluster topology, workload, and recovery policy.
The business signal is strong because infrastructure waste is easy to price when a team can measure failed training time. A builder should ask for a replay on its own workload: inject a link or node failure, measure the lost useful compute, then compare recovery time and correctness with the existing checkpoint path.
What you can do with it
Instrument one long-running training or inference job before buying a resilience layer. Record failure frequency, recovery time, replay cost, and lost useful work. Then run a controlled fault injection and require the recovered job to match a clean baseline before counting the saved GPU-hours.
Our take
Reliability is a sharper AI-infrastructure wedge than vague claims about faster clusters. Clockwork's reported numbers are not proof for every buyer, but the measurement path is direct: if recovery reduces lost compute on a real workload, the product can justify its cost without needing a new model narrative.
Links Clockwork Funding release
Source: Clockwork ↗ — Made With Models writes the brief; the reporting is theirs.