Clockwork.io, an AI infrastructure startup based in Palo Alto, has secured $31 million in a new funding round. The investment was co-led by Premji Invest, Wing Venture Capital, and Seligman Ventures, with additional support from existing backers NEA and e& Capital. This latest injection of capital brings the company's total funding to $73 million.
The startup develops software designed to maintain AI workloads during hardware or network failures. Its LinkPass product reroutes traffic during network outages, while TorchPass shifts tasks away from failing GPUs. These tools allow training jobs to continue without the need to restart from previous checkpoints, effectively reducing downtime and preventing the waste of expensive computing resources.
The software is already in use by major companies, including LinkedIn, Wells Fargo, and Uber. LinkedIn reports that the technology prevents tens of thousands of GPU-hours of downtime every month. Additionally, Together AI has integrated TorchPass into its GPU clusters, and WhiteFiber is expanding its implementation of the software across its own infrastructure.
Clockwork.io recently introduced new features for TorchPass. These updates enable the capture of training job states across multiple nodes and allow for background application checkpointing. These capabilities aim to shorten recovery times and facilitate the transfer of model weights between training and inference systems without requiring changes to the existing training code.
The company intends to allocate the new funds toward expanding the deployment of its software across training, inference, and reinforcement-learning workloads. Furthermore, Clockwork.io plans to acquire more enterprise clients and strengthen its distribution channels through partnerships with cloud and neocloud providers like Nebius and NScale.
