Clockwork.io Raises $31M To Stop Wasted GPU-Hours

Clockwork.io raised $31 million to expand its GPU fault-tolerance software after LinkedIn, Together AI and WhiteFiber adopted it to stop losing GPU-hours to hardware failures.

Trace Cohen
By Trace Cohen Β· Value Add Pulse Funding Desk
Founder & Editor-in-chief, Value Add VC
2 min read
Clockwork.io Raises $31M To Stop Wasted GPU-Hours: key figures
ShareXLinkedInEmail

THE RUNDOWN

1

GPU idle-time from hardware failures is an invisible tax on every AI lab's compute spend -- Clockwork.io's pitch is recovering that waste without the customer having to rewrite training code, a low-friction sell that explains fast enterprise adoption.

2

LinkedIn, Together AI and WhiteFiber all adopting the same resilience layer suggests infrastructure reliability is becoming a shared utility across large enterprise AI fleets, neoclouds and GPU-as-a-service providers alike, not a one-off internal tool each builds itself.

3

Together AI planning to resell TorchPass as a service shows how fast infrastructure startups can become embedded inside larger GPU-cloud platforms rather than selling directly to end customers -- a distribution path worth tracking for infra-layer founders.

4

At $73M in total funding against a single infrastructure-reliability niche, Clockwork.io's backers are betting GPU failure-recovery becomes as standard a line item as networking or storage -- a bet on category durability, not a one-time trend.

The VC Read

Value Add VC analysis

LinkedIn, Together AI and WhiteFiber adopting the same tool in one cycle is the diligence signal, not the $31M check size. Worth asking Clockwork.io directly what percentage of GPU-hours saved translates into renewed contracts versus one-time pilot deployments -- infra tools die fast when the savings don't show up in a renewal conversation.

Analysis

Clockwork.io has raised $31 million to expand its fault-tolerance software for AI infrastructure, led by Premji Invest, Wing Venture Capital and Seligman Ventures, with participation from NEA and e& Capital, bringing its total funding to $73 million, according to The SaaS News and Converge Digest.

What the software actually does

The Palo Alto-based company, founded in 2018 as TickTock Networks out of Stanford research and renamed in 2021, builds two core products: LinkPass, which reroutes network traffic around failed links, optics, cables or NICs, and TorchPass, which migrates work from a failing GPU to healthy capacity without forcing a full restart from an earlier checkpoint. Clockwork.io is extending TorchPass with multi-node platform snapshots that capture an entire distributed job's execution state without requiring changes to training code, plus asynchronous checkpoints that run while a workload keeps training.

The problem Clockwork.io is solving is a real and expensive one: a single failed GPU, NIC, optical link or switch port in a large distributed training run can force recovery from an earlier checkpoint, leaving the rest of an otherwise-healthy GPU cluster idle while it waits. At the scale frontier labs now train at, that idle time compounds into real money.

β€œAt the scale frontier labs now train at, that idle time compounds into real money.”

Early customers signal fast adoption

LinkedIn has deployed LinkPass across its AI infrastructure fleet and reports the software prevents tens of thousands of GPU-hours of downtime every month. Together AI, a GPU-cloud provider that competes with names like CoreWeave and Crusoe, plans to offer TorchPass as a service directly to its own customers, while WhiteFiber is expanding Clockwork.io across its global GPU-as-a-service infrastructure. That combination -- a large enterprise AI fleet, a neocloud, and a GPU-as-a-service provider all adopting the same resilience layer in the same funding cycle -- is a stronger adoption signal than the dollar amount of the round itself.

What the headline misses

Clockwork.io has not disclosed revenue, a valuation, or customer count beyond the three named deployments, which makes it hard to size how much of the GPU-cloud market has actually standardized on its tools versus built in-house alternatives. Hyperscalers like Google and Amazon have their own internal failure-recovery systems for their own fleets, and nothing here suggests Clockwork.io has displaced those -- its traction so far is concentrated among neoclouds and GPU-as-a-service providers that lack hyperscaler-scale infrastructure teams.

Watch whether Clockwork.io discloses a hyperscaler-scale customer win next, which would be the real test of whether this becomes infrastructure-layer standard or stays a neocloud-tier tool.

ShareXLinkedInEmail

More from the Pulse archive: Startup & VC Funding coverage β†’ Β· All topics

Key Sources

2 sources

Reported by The SaaS News Β· First reported by Converge Digest Β· Analysis by Value Add Pulse.

← Back to Pulse

THE WIRE in your inboxβ€” Tech, startup & VC news with The VC Read, a few times a week. Free to subscribe, no spam.

No spam, unsubscribe anytime. Sent via Substack Β· Privacy Β· or subscribe on Substack β†—

Read Next

FUNDING
Open-Source AI Infrastructure's Funding Renaissance
Illustration for: Open-Source AI Infrastructure's Funding Renaissance
FUNDING~$965M combined

Open-Source AI Infrastructure's Funding Renaissance

Ollama's $65 million Series B is the latest sign that open-source AI tooling -- once seen as a commodity layer -- is commanding real institutional capital alongside the neoclouds and frontier labs.

FUNDING
Crunchbase: The Week's 10 Biggest Funding Rounds, Led by Joulent's $1.75B Energy Raise
Crunchbase: The Week's 10 Biggest Funding Rounds, Led by Joulent's $1.75B Energy Raise logo
FUNDING$1.75B lead deal (Joulent)

Crunchbase: The Week's 10 Biggest Funding Rounds, Led by Joulent's $1.75B Energy Raise

Crunchbase News' roundup of the week's 10 biggest funding rounds (June 27-July 2) was led by Houston-based energy infrastructure provider Joulent's $1.75 billion raise from National Grid Ventures, aimed squarely at AI and compute-intensive industries' power needs. Together AI's $800M Series C and Venice AI's $65M unicorn round both placed inside the top 10, alongside biotech, legal-tech, sports and video-AI rounds spanning a genuinely broad set of categories beyond pure frontier-model AI.

FUNDING
TwelveLabs Raises $100M Series B to Build 'Video Superintelligence,' Deepens AWS Partnership
Illustration for: TwelveLabs Raises $100M Series B to Build 'Video Superintelligence,' Deepens AWS Partnership
FUNDING$100M

TwelveLabs Raises $100M Series B to Build 'Video Superintelligence,' Deepens AWS Partnership

TwelveLabs closed a $100 million Series B on July 1, co-led by NEA and Naver Ventures with Amazon participating directly, pushing total funding above $200 million. The company builds foundation models that understand video the way humans do, and is expanding from video-understanding models into a full-stack agentic system combining perception, knowledge and reasoning, while committing AWS Trainium as its preferred inference chip.