{"success":true,"data":{"pressRelease":{"id":"160642","rtpr_id":"nPn9P7p41a-20261005","ticker":"WYFI","exchange":"NASDAQ","all_tickers":["WYFI"],"title":"Clockwork.io Raises $31M as LinkedIn, Together AI and WhiteFiber Adopt Its Resilience Software to Stop Wasting GPU-Hours","author":"PR Newswire","published_at":"2026-10-05T13:00:10.096Z","article_body":"Clockwork.io Raises $31M as LinkedIn, Together AI and WhiteFiber Adopt Its Resilience Software to Stop Wasting GPU-Hours\nPR Newswire\n\nPALO ALTO, Calif., Oct. 5, 2026\n\nLinkedIn prevents tens of thousands of GPU-hours of downtime monthly; new\nTorchPass innovations preserve AI workload progress without code changes and\nspeed reinforcement learning.\n\nPALO ALTO, Calif., Oct. 5, 2026 /PRNewswire/ -- Clockwork.io\n(https://edge.prnewswire.com/c/link/?t=0&l=en&o=4788996-2&h=2176116736&u=http%3A%2F%2Fclockwork.io%2F&a=Clockwork.io)\n, whose fault-tolerance software keeps AI training, reinforcement learning and\ninference workloads running through infrastructure failures, today announced\n$31 million in new funding, production deployments at LinkedIn and Together\nAI, and expanded adoption by WhiteFiber.\n\nThe company also introduced two new capabilities for its TorchPass solution,\neach capturing the state of a running distributed AI job. Multi-node platform\nsnapshots, an industry first for training, save an entire running job across\nevery node without changes to the training code and preserve it for recovery.\nFast, asynchronous application checkpoints, taken in the background while the\njob runs, accelerate reinforcement learning. They deliver updated model\nweights to the inference replicas that generate rollouts, the examples the\nmodel learns from, so those replicas spend less time waiting or working from a\nstale model.\n\nFault tolerance has become a requirement for AI at scale\n\nLarge distributed AI workloads can span thousands of GPUs that must stay in\nsync: one failed GPU, dropped link or frozen server can stall the entire job.\nMeta reported unexpected interruptions averaging roughly one every three hours\nduring a 54-day period of Llama 3 training on 16,384 GPUs.\n\nThe typical response is to reload a checkpoint, a saved copy of the job's\nprogress. Recovery can take up to 90 minutes, leaves healthy GPUs waiting, and\nrequires the job to repeat work completed since that checkpoint. Customers pay\nfor idle GPUs and repeated computation, and models take longer to complete. As\njobs grow, each restart puts more GPU time at risk.\n\nAt this scale, keeping useful work running through failures is an\ninfrastructure requirement. Clockwork.io\n(https://edge.prnewswire.com/c/link/?t=0&l=en&o=4788996-2&h=2176116736&u=http%3A%2F%2Fclockwork.io%2F&a=Clockwork.io)\n meets it with a fault-tolerance suite that platform teams deploy as a layer\nbetween the hardware and the workload. LinkPass reroutes traffic around a\nfailed link so the job never sees the fault. TorchPass moves work from a\nfailing GPU to a healthy one so training continues instead of rolling back.\nBoth are in production. TorchPass's new platform snapshots, announced today,\ncapture the state of a running distributed job so the whole job can be\nrestored when a failure is too large to migrate around.\n\n\"Failures are inevitable at AI scale. Losing hours of useful work to them\nshould not be,\" said Suresh Vasudevan, CEO of Clockwork.io. \"Fault tolerance\nis a goodput multiplier: it keeps GPUs doing useful work instead of waiting\nfor recovery or repeating work already done. We built our software alongside\nenterprises and cloud providers operating some of the largest GPU fleets, so\nit handles the failures they actually see. That protection belongs in the\ninfrastructure enterprises and cloud providers rely on every day.\"\n\nAdoption expands across enterprises, hyperscalers and neoclouds\n\nEnterprises running their own GPU fleets, hyperscalers and neoclouds are\nadopting Clockwork.io\n(https://edge.prnewswire.com/c/link/?t=0&l=en&o=4788996-2&h=2176116736&u=http%3A%2F%2Fclockwork.io%2F&a=Clockwork.io)\n for the same reason: more of their GPU-hours go to useful work.\n\nLinkedIn has deployed LinkPass network fault tolerance across its AI\ninfrastructure fleet and prevents tens of thousands of GPU-hours of downtime\neach month.\n\n\"At AI infrastructure scale, a single network issue should never sideline\nhealthy GPUs or interrupt running workloads. Before Clockwork.io, one\nInfiniBand NIC flap could remove an eight-GPU server from service, while a\nswitch port flap could drain a second server, doubling the impact to 16 GPUs,\"\nsaid Raghu Hiremagalur, SVP, CTO Infrastructure, LinkedIn. \"Clockwork.io\nhelped transform that operating model. Its network fault-tolerance technology\nautomatically reroutes traffic onto healthy paths, allowing jobs to continue\nuninterrupted while link, optic, cable, or NIC faults are repaired. In\naggregate, Clockwork.io\n(https://edge.prnewswire.com/c/link/?t=0&l=en&o=4788996-2&h=3538194085&u=http%3A%2F%2Fclockwork.io%2F&a=Clockwork.io+)\nprevents tens of thousands of GPU-hours of downtime per month across our\nfleet. By turning what were once disruptive operational incidents into\nmanageable maintenance events, Clockwork.io has helped improve infrastructure\nutilization and operational efficiency.\"\n\nTogether AI is bringing TorchPass to market as a service on its GPU Clusters.\nAt the PyTorch Conference, the two companies will demonstrate a live\nmulti-node training job continuing through injected network and GPU failures\nwithout restarting.\n\n\"Our customers grade us on goodput, the share of their GPU-hours that actually\nmove the model forward,\" said Pavneet Ahluwalia, Product Lead, Together AI.\n\"Node repair already detects faults and provisions replacement capacity\nautomatically. Clockwork.io's\n(https://edge.prnewswire.com/c/link/?t=0&l=en&o=4788996-2&h=1066710848&u=http%3A%2F%2Fclockwork.io%2F&a=Clockwork.io%27s)\n TorchPass and LinkPass build on that foundation and are designed to keep\njobs moving through GPU faults and link failures, preserving progress. We are\nbringing them to market as the next layer of resilience in the platform.\"\n\nWhiteFiber (NASDAQ: WYFI), an existing customer, is expanding its use of\nClockwork.io\n(https://edge.prnewswire.com/c/link/?t=0&l=en&o=4788996-2&h=3538194085&u=http%3A%2F%2Fclockwork.io%2F&a=Clockwork.io+)\nsoftware across its growing global GPU-as-a-service footprint.\n\n\"Pressure-testing a cluster's reliability before it reaches production is\ncritical, because a customer who inherits a hidden fabric fault pays for it\nlater in failed jobs and lost GPU-hours,\" said Tom Sanfilippo, Chief\nTechnology Officer, WhiteFiber. \"Marginal optics, misconfigured NICs, and\nlinks that pass a basic test but degrade under load can slip through.\nClockwork.io's automated fleet audit validates every link and node at once,\nlocalizes faults in minutes, and lets us correct them before acceptance. We\nbring clusters up faster, and a customer's first training run lands on a\nfabric validated end-to-end, not just powered on. With market demand growing\nas rapidly as it is, getting validated capacity to customers quickly is\ncritical to our business, and it is why we are expanding Clockwork.io\n(https://edge.prnewswire.com/c/link/?t=0&l=en&o=4788996-2&h=2176116736&u=http%3A%2F%2Fclockwork.io%2F&a=Clockwork.io)\n across our clusters.\"\n\nNew TorchPass capabilities put workload protection in platform teams' hands\n\nClockwork.io\n(https://edge.prnewswire.com/c/link/?t=0&l=en&o=4788996-2&h=3538194085&u=http%3A%2F%2Fclockwork.io%2F&a=Clockwork.io+)\nextended TorchPass beyond GPU migration with two capabilities platform teams\nhave not had: a snapshot of a whole distributed job that they can take\nthemselves, and application checkpoints fast enough to run in the background.\n\nFor training, platform snapshots save a running job's execution state across\nall of its nodes so the job can be restored after an interruption. Platform\nteams and AI infrastructure engineers deploy it for supported workloads\nwithout waiting for application owners to modify their code or add\ncheckpointing logic. Enterprise teams can protect training jobs across their\nfleet with one mechanism, and cloud providers can protect customer jobs whose\ncode they do not control. Where teams checkpoint at the application level,\nTorchPass's fast checkpoints can be taken more often, so less progress is lost\nand less computation repeated after a failure.\n\nFor inference, large models run across two or more servers, so one bad link\ncan take down a whole replica and cut off a user session or agent task\nmid-stream. LinkPass keeps those multi-server replicas serving through link\nfailures.\n\nReinforcement learning depends on both training and inference. Copies of the\nmodel generate rollouts, the trainer learns from them, and updated weights\nmust reach the copies before they can generate with the latest version.\nTorchPass's application checkpoints carry the updated weights to the rollout\nreplicas sooner, while LinkPass keeps those replicas serving through link\nfailures.\n\n\"Cluster fault tolerance used to be a training problem. It is now an inference\nproblem too,\" said Dylan Patel, Founder, CEO, and Chief Analyst at\nSemiAnalysis, whose ClusterMAX ratings benchmark GPU cloud providers. \"In our\nClusterMAX, TorchPass cuts training goodput loss from 14% to under 3% for a\ngold-rated neocloud. Reinforcement Learning (RL) ties the two together:\ninference replicas generate rollouts, the trainer learns from them, and the\nupdated weights go back to the replicas. Clockwork.io\n(https://edge.prnewswire.com/c/link/?t=0&l=en&o=4788996-2&h=2176116736&u=http%3A%2F%2Fclockwork.io%2F&a=Clockwork.io)\n keeps replicas serving through link flaps and network failures. Its\nextremely fast checkpoints accelerate weight transfer back into the rollout\nfleet, so neither direction stalls the run. One fault-tolerance layer under\ntraining, inference, and RL is where this has to be solved.\"\n\nEnterprise platform teams and cloud providers can contact Clockwork.io\n(https://edge.prnewswire.com/c/link/?t=0&l=en&o=4788996-2&h=2176116736&u=http%3A%2F%2Fclockwork.io%2F&a=Clockwork.io)\n to evaluate the software for their workloads or explore partnership\nopportunities.\n\nThe round and what it funds\n\nThe round was co-led by Premji Invest, Wing Venture Capital, and Seligman\nVentures, with participation from existing investors NEA and e& Capital.\nIt brings Clockwork.io's\n(https://edge.prnewswire.com/c/link/?t=0&l=en&o=4788996-2&h=1066710848&u=http%3A%2F%2Fclockwork.io%2F&a=Clockwork.io%27s)\n total funding to $73 million. The company will use the capital to accelerate\nthe rollout of its fault-tolerance suite across training, inference and\nreinforcement learning, expand enterprise adoption, and scale delivery through\ncloud partners.\n\n\"The one thing that scales perfectly is unreliability: put enough GPUs in one\nmachine and something is always failing,\" said Greg Papadopoulos, Venture\nPartner, NEA. \"The old playbook: stop the job, reload a checkpoint, makes no\nsense at today's scale. Clockwork treats failure as the normal state:\nTorchPass migrates training off a failing GPU live, and now snapshots an\nentire running job with no code changes. It's already saving tens of thousands\nof GPU-hours a month. We first backed Clockwork.io\n(https://edge.prnewswire.com/c/link/?t=0&l=en&o=4788996-2&h=3538194085&u=http%3A%2F%2Fclockwork.io%2F&a=Clockwork.io+)\nin 2021 and are thrilled to keep supporting them as they define the\nperformance layer of the AI cluster.\"\n\nAbout Clockwork.io\n(https://edge.prnewswire.com/c/link/?t=0&l=en&o=4788996-2&h=2176116736&u=http%3A%2F%2Fclockwork.io%2F&a=Clockwork.io)\n\nClockwork.io\n(https://edge.prnewswire.com/c/link/?t=0&l=en&o=4788996-2&h=2176116736&u=http%3A%2F%2Fclockwork.io%2F&a=Clockwork.io)\n pioneers Software-Driven AI Fabrics™, a programmable layer between\nhardware and workload that makes GPU clusters observable, fault-tolerant, and\nfully utilized across any accelerator, network, or cloud. AI workloads need\nthe whole cluster to act as one machine, yet failures and bottlenecks idle\nGPUs. Clockwork.io's\n(https://edge.prnewswire.com/c/link/?t=0&l=en&o=4788996-2&h=1066710848&u=http%3A%2F%2Fclockwork.io%2F&a=Clockwork.io%27s)\n FleetLens platform recovers that lost capacity: nanosecond-accurate\ntelemetry pinpoints the GPU, node, or link slowing or stalling a job,\nvalidates clusters before launch, and, with LinkPass and TorchPass, keeps\nworkloads running through infrastructure failures and degradations, from\ntraining to inference. SemiAnalysis has independently benchmarked TorchPass as\nrecovering from failures faster than checkpoint-restart and leading\nopen-source frameworks. LinkedIn, Together AI, WhiteFiber, Wells Fargo,\nNebius, NScale, and DCAI trust Clockwork.io to power their AI infrastructure.\nLearn more at\n(https://edge.prnewswire.com/c/link/?t=0&l=en&o=4788996-2&h=281815303&u=https%3A%2F%2Fwww.clockwork.io%2F&a=%C2%A0)\nwww.clockwork.io\n(https://edge.prnewswire.com/c/link/?t=0&l=en&o=4788996-2&h=4006713019&u=https%3A%2F%2Fwww.clockwork.io%2F&a=www.clockwork.io)\n.\n\nView original content to download\nmultimedia:https://www.prnewswire.com/news-releases/clockworkio-raises-31m-as-linkedin-together-ai-and-whitefiber-adopt-its-resilience-software-to-stop-wasting-gpu-hours-302897586.html\n(https://www.prnewswire.com/news-releases/clockworkio-raises-31m-as-linkedin-together-ai-and-whitefiber-adopt-its-resilience-software-to-stop-wasting-gpu-hours-302897586.html)\n\nSOURCE Clockwork.io\n\n\n\nMedia Contact: Dana Trismen, clockwork.io@unshakablemarketinggroup.com\n\nPhoto: \nhttps://mmx.prnewswire.com/media/MS2001849/Clockwork-press-illustration-no-text-1920x1080_2-1.jpg?id=OA2985523\nhttps://mmx.prnewswire.com/media/MS2001848/1145681.jpg?id=OA2985522\n\nCopyright (c) 2026 PR Newswire Association,LLC. All Rights Reserved.","article_body_html":"","raw_payload":{"data":{"id":"nPn9P7p41a-20261005","title":"Clockwork.io Raises $31M as LinkedIn, Together AI and WhiteFiber Adopt Its Resilience Software to Stop Wasting GPU-Hours","author":"PR Newswire","ticker":"WYFI","created":"2026-10-05T13:00:10.096Z","tickers":["WYFI"],"exchange":"NASDAQ","article_body":"Clockwork.io Raises $31M as LinkedIn, Together AI and WhiteFiber Adopt Its Resilience Software to Stop Wasting GPU-Hours\nPR Newswire\n\nPALO ALTO, Calif., Oct. 5, 2026\n\nLinkedIn prevents tens of thousands of GPU-hours of downtime monthly; new\nTorchPass innovations preserve AI workload progress without code changes and\nspeed reinforcement learning.\n\nPALO ALTO, Calif., Oct. 5, 2026 /PRNewswire/ -- Clockwork.io\n(https://edge.prnewswire.com/c/link/?t=0&l=en&o=4788996-2&h=2176116736&u=http%3A%2F%2Fclockwork.io%2F&a=Clockwork.io)\n, whose fault-tolerance software keeps AI training, reinforcement learning and\ninference workloads running through infrastructure failures, today announced\n$31 million in new funding, production deployments at LinkedIn and Together\nAI, and expanded adoption by WhiteFiber.\n\nThe company also introduced two new capabilities for its TorchPass solution,\neach capturing the state of a running distributed AI job. Multi-node platform\nsnapshots, an industry first for training, save an entire running job across\nevery node without changes to the training code and preserve it for recovery.\nFast, asynchronous application checkpoints, taken in the background while the\njob runs, accelerate reinforcement learning. They deliver updated model\nweights to the inference replicas that generate rollouts, the examples the\nmodel learns from, so those replicas spend less time waiting or working from a\nstale model.\n\nFault tolerance has become a requirement for AI at scale\n\nLarge distributed AI workloads can span thousands of GPUs that must stay in\nsync: one failed GPU, dropped link or frozen server can stall the entire job.\nMeta reported unexpected interruptions averaging roughly one every three hours\nduring a 54-day period of Llama 3 training on 16,384 GPUs.\n\nThe typical response is to reload a checkpoint, a saved copy of the job's\nprogress. Recovery can take up to 90 minutes, leaves healthy GPUs waiting, and\nrequires the job to repeat work completed since that checkpoint. Customers pay\nfor idle GPUs and repeated computation, and models take longer to complete. As\njobs grow, each restart puts more GPU time at risk.\n\nAt this scale, keeping useful work running through failures is an\ninfrastructure requirement. Clockwork.io\n(https://edge.prnewswire.com/c/link/?t=0&l=en&o=4788996-2&h=2176116736&u=http%3A%2F%2Fclockwork.io%2F&a=Clockwork.io)\n meets it with a fault-tolerance suite that platform teams deploy as a layer\nbetween the hardware and the workload. LinkPass reroutes traffic around a\nfailed link so the job never sees the fault. TorchPass moves work from a\nfailing GPU to a healthy one so training continues instead of rolling back.\nBoth are in production. TorchPass's new platform snapshots, announced today,\ncapture the state of a running distributed job so the whole job can be\nrestored when a failure is too large to migrate around.\n\n\"Failures are inevitable at AI scale. Losing hours of useful work to them\nshould not be,\" said Suresh Vasudevan, CEO of Clockwork.io. \"Fault tolerance\nis a goodput multiplier: it keeps GPUs doing useful work instead of waiting\nfor recovery or repeating work already done. We built our software alongside\nenterprises and cloud providers operating some of the largest GPU fleets, so\nit handles the failures they actually see. That protection belongs in the\ninfrastructure enterprises and cloud providers rely on every day.\"\n\nAdoption expands across enterprises, hyperscalers and neoclouds\n\nEnterprises running their own GPU fleets, hyperscalers and neoclouds are\nadopting Clockwork.io\n(https://edge.prnewswire.com/c/link/?t=0&l=en&o=4788996-2&h=2176116736&u=http%3A%2F%2Fclockwork.io%2F&a=Clockwork.io)\n for the same reason: more of their GPU-hours go to useful work.\n\nLinkedIn has deployed LinkPass network fault tolerance across its AI\ninfrastructure fleet and prevents tens of thousands of GPU-hours of downtime\neach month.\n\n\"At AI infrastructure scale, a single network issue should never sideline\nhealthy GPUs or interrupt running workloads. Before Clockwork.io, one\nInfiniBand NIC flap could remove an eight-GPU server from service, while a\nswitch port flap could drain a second server, doubling the impact to 16 GPUs,\"\nsaid Raghu Hiremagalur, SVP, CTO Infrastructure, LinkedIn. \"Clockwork.io\nhelped transform that operating model. Its network fault-tolerance technology\nautomatically reroutes traffic onto healthy paths, allowing jobs to continue\nuninterrupted while link, optic, cable, or NIC faults are repaired. In\naggregate, Clockwork.io\n(https://edge.prnewswire.com/c/link/?t=0&l=en&o=4788996-2&h=3538194085&u=http%3A%2F%2Fclockwork.io%2F&a=Clockwork.io+)\nprevents tens of thousands of GPU-hours of downtime per month across our\nfleet. By turning what were once disruptive operational incidents into\nmanageable maintenance events, Clockwork.io has helped improve infrastructure\nutilization and operational efficiency.\"\n\nTogether AI is bringing TorchPass to market as a service on its GPU Clusters.\nAt the PyTorch Conference, the two companies will demonstrate a live\nmulti-node training job continuing through injected network and GPU failures\nwithout restarting.\n\n\"Our customers grade us on goodput, the share of their GPU-hours that actually\nmove the model forward,\" said Pavneet Ahluwalia, Product Lead, Together AI.\n\"Node repair already detects faults and provisions replacement capacity\nautomatically. Clockwork.io's\n(https://edge.prnewswire.com/c/link/?t=0&l=en&o=4788996-2&h=1066710848&u=http%3A%2F%2Fclockwork.io%2F&a=Clockwork.io%27s)\n TorchPass and LinkPass build on that foundation and are designed to keep\njobs moving through GPU faults and link failures, preserving progress. We are\nbringing them to market as the next layer of resilience in the platform.\"\n\nWhiteFiber (NASDAQ: WYFI), an existing customer, is expanding its use of\nClockwork.io\n(https://edge.prnewswire.com/c/link/?t=0&l=en&o=4788996-2&h=3538194085&u=http%3A%2F%2Fclockwork.io%2F&a=Clockwork.io+)\nsoftware across its growing global GPU-as-a-service footprint.\n\n\"Pressure-testing a cluster's reliability before it reaches production is\ncritical, because a customer who inherits a hidden fabric fault pays for it\nlater in failed jobs and lost GPU-hours,\" said Tom Sanfilippo, Chief\nTechnology Officer, WhiteFiber. \"Marginal optics, misconfigured NICs, and\nlinks that pass a basic test but degrade under load can slip through.\nClockwork.io's automated fleet audit validates every link and node at once,\nlocalizes faults in minutes, and lets us correct them before acceptance. We\nbring clusters up faster, and a customer's first training run lands on a\nfabric validated end-to-end, not just powered on. With market demand growing\nas rapidly as it is, getting validated capacity to customers quickly is\ncritical to our business, and it is why we are expanding Clockwork.io\n(https://edge.prnewswire.com/c/link/?t=0&l=en&o=4788996-2&h=2176116736&u=http%3A%2F%2Fclockwork.io%2F&a=Clockwork.io)\n across our clusters.\"\n\nNew TorchPass capabilities put workload protection in platform teams' hands\n\nClockwork.io\n(https://edge.prnewswire.com/c/link/?t=0&l=en&o=4788996-2&h=3538194085&u=http%3A%2F%2Fclockwork.io%2F&a=Clockwork.io+)\nextended TorchPass beyond GPU migration with two capabilities platform teams\nhave not had: a snapshot of a whole distributed job that they can take\nthemselves, and application checkpoints fast enough to run in the background.\n\nFor training, platform snapshots save a running job's execution state across\nall of its nodes so the job can be restored after an interruption. Platform\nteams and AI infrastructure engineers deploy it for supported workloads\nwithout waiting for application owners to modify their code or add\ncheckpointing logic. Enterprise teams can protect training jobs across their\nfleet with one mechanism, and cloud providers can protect customer jobs whose\ncode they do not control. Where teams checkpoint at the application level,\nTorchPass's fast checkpoints can be taken more often, so less progress is lost\nand less computation repeated after a failure.\n\nFor inference, large models run across two or more servers, so one bad link\ncan take down a whole replica and cut off a user session or agent task\nmid-stream. LinkPass keeps those multi-server replicas serving through link\nfailures.\n\nReinforcement learning depends on both training and inference. Copies of the\nmodel generate rollouts, the trainer learns from them, and updated weights\nmust reach the copies before they can generate with the latest version.\nTorchPass's application checkpoints carry the updated weights to the rollout\nreplicas sooner, while LinkPass keeps those replicas serving through link\nfailures.\n\n\"Cluster fault tolerance used to be a training problem. It is now an inference\nproblem too,\" said Dylan Patel, Founder, CEO, and Chief Analyst at\nSemiAnalysis, whose ClusterMAX ratings benchmark GPU cloud providers. \"In our\nClusterMAX, TorchPass cuts training goodput loss from 14% to under 3% for a\ngold-rated neocloud. Reinforcement Learning (RL) ties the two together:\ninference replicas generate rollouts, the trainer learns from them, and the\nupdated weights go back to the replicas. Clockwork.io\n(https://edge.prnewswire.com/c/link/?t=0&l=en&o=4788996-2&h=2176116736&u=http%3A%2F%2Fclockwork.io%2F&a=Clockwork.io)\n keeps replicas serving through link flaps and network failures. Its\nextremely fast checkpoints accelerate weight transfer back into the rollout\nfleet, so neither direction stalls the run. One fault-tolerance layer under\ntraining, inference, and RL is where this has to be solved.\"\n\nEnterprise platform teams and cloud providers can contact Clockwork.io\n(https://edge.prnewswire.com/c/link/?t=0&l=en&o=4788996-2&h=2176116736&u=http%3A%2F%2Fclockwork.io%2F&a=Clockwork.io)\n to evaluate the software for their workloads or explore partnership\nopportunities.\n\nThe round and what it funds\n\nThe round was co-led by Premji Invest, Wing Venture Capital, and Seligman\nVentures, with participation from existing investors NEA and e& Capital.\nIt brings Clockwork.io's\n(https://edge.prnewswire.com/c/link/?t=0&l=en&o=4788996-2&h=1066710848&u=http%3A%2F%2Fclockwork.io%2F&a=Clockwork.io%27s)\n total funding to $73 million. The company will use the capital to accelerate\nthe rollout of its fault-tolerance suite across training, inference and\nreinforcement learning, expand enterprise adoption, and scale delivery through\ncloud partners.\n\n\"The one thing that scales perfectly is unreliability: put enough GPUs in one\nmachine and something is always failing,\" said Greg Papadopoulos, Venture\nPartner, NEA. \"The old playbook: stop the job, reload a checkpoint, makes no\nsense at today's scale. Clockwork treats failure as the normal state:\nTorchPass migrates training off a failing GPU live, and now snapshots an\nentire running job with no code changes. It's already saving tens of thousands\nof GPU-hours a month. We first backed Clockwork.io\n(https://edge.prnewswire.com/c/link/?t=0&l=en&o=4788996-2&h=3538194085&u=http%3A%2F%2Fclockwork.io%2F&a=Clockwork.io+)\nin 2021 and are thrilled to keep supporting them as they define the\nperformance layer of the AI cluster.\"\n\nAbout Clockwork.io\n(https://edge.prnewswire.com/c/link/?t=0&l=en&o=4788996-2&h=2176116736&u=http%3A%2F%2Fclockwork.io%2F&a=Clockwork.io)\n\nClockwork.io\n(https://edge.prnewswire.com/c/link/?t=0&l=en&o=4788996-2&h=2176116736&u=http%3A%2F%2Fclockwork.io%2F&a=Clockwork.io)\n pioneers Software-Driven AI Fabrics™, a programmable layer between\nhardware and workload that makes GPU clusters observable, fault-tolerant, and\nfully utilized across any accelerator, network, or cloud. AI workloads need\nthe whole cluster to act as one machine, yet failures and bottlenecks idle\nGPUs. Clockwork.io's\n(https://edge.prnewswire.com/c/link/?t=0&l=en&o=4788996-2&h=1066710848&u=http%3A%2F%2Fclockwork.io%2F&a=Clockwork.io%27s)\n FleetLens platform recovers that lost capacity: nanosecond-accurate\ntelemetry pinpoints the GPU, node, or link slowing or stalling a job,\nvalidates clusters before launch, and, with LinkPass and TorchPass, keeps\nworkloads running through infrastructure failures and degradations, from\ntraining to inference. SemiAnalysis has independently benchmarked TorchPass as\nrecovering from failures faster than checkpoint-restart and leading\nopen-source frameworks. LinkedIn, Together AI, WhiteFiber, Wells Fargo,\nNebius, NScale, and DCAI trust Clockwork.io to power their AI infrastructure.\nLearn more at\n(https://edge.prnewswire.com/c/link/?t=0&l=en&o=4788996-2&h=281815303&u=https%3A%2F%2Fwww.clockwork.io%2F&a=%C2%A0)\nwww.clockwork.io\n(https://edge.prnewswire.com/c/link/?t=0&l=en&o=4788996-2&h=4006713019&u=https%3A%2F%2Fwww.clockwork.io%2F&a=www.clockwork.io)\n.\n\nView original content to download\nmultimedia:https://www.prnewswire.com/news-releases/clockworkio-raises-31m-as-linkedin-together-ai-and-whitefiber-adopt-its-resilience-software-to-stop-wasting-gpu-hours-302897586.html\n(https://www.prnewswire.com/news-releases/clockworkio-raises-31m-as-linkedin-together-ai-and-whitefiber-adopt-its-resilience-software-to-stop-wasting-gpu-hours-302897586.html)\n\nSOURCE Clockwork.io\n\n\n\nMedia Contact: Dana Trismen, clockwork.io@unshakablemarketinggroup.com\n\nPhoto: \nhttps://mmx.prnewswire.com/media/MS2001849/Clockwork-press-illustration-no-text-1920x1080_2-1.jpg?id=OA2985523\nhttps://mmx.prnewswire.com/media/MS2001848/1145681.jpg?id=OA2985522\n\nCopyright (c) 2026 PR Newswire Association,LLC. All Rights Reserved."},"type":"article","timestamp":"2026-10-05T13:00:10.169572541Z","server_sent_at_ms":1791205210169},"received_at":"2026-10-05T13:00:10.325Z","source_url":"https://www.prnewswire.com/news-releases/clockworkio-raises-31m-as-linkedin-together-ai-and-whitefiber-adopt-its-resilience-software-to-stop-wasting-gpu-hours-302897586.html"},"analysis":null,"durationMs":null,"modelName":null}}