AI infrastructure startup Clockwork has raised $20.6 million and launched its breakthrough software, FleetIQ, to fix inefficiencies in GPU-powered systems.
Quick Summary – TLDR:
- Clockwork raised $20.6 million in a funding round led by NEA, with backing from AMD, Broadcom, and tech leaders like Lip-Bu Tan and John Chambers.
- The new funding will accelerate development and adoption of its AI infrastructure software, FleetIQ.
- FleetIQ optimizes communication between GPUs, clusters, and clouds, improving performance, uptime, and cost-efficiency.
- The company’s roots trace back to Stanford University research, now powering real-world enterprise workloads like those at Uber.
What Happened?
Clockwork, a San Francisco-based AI compute startup and Stanford spinout, has secured $20.6 million in a funding round to solve one of AI’s most pressing problems: underutilized GPUs and inefficient infrastructure. Alongside the raise, the company has officially launched FleetIQ, a software platform designed to optimize and modernize AI workloads at scale.
Led by venture firm New Enterprise Associates (NEA), the round saw participation from AMD, Broadcom, and tech visionaries such as Intel CEO Lip-Bu Tan, former Cisco CEO John Chambers, and Carl Ledbetter. Clockwork plans to use the funding to scale adoption of its new product and expand its footprint into enterprise and neocloud GPU deployments.
11/
— clockworkio (@clockworkio) September 10, 2025
Clockwork’s Software-Driven Fabric (SDF) is redefining the foundation of AI infrastructure.
Read the full announcement here 👉 https://t.co/xLjIrozWLL pic.twitter.com/SiF85v2551
Clockwork Takes on AI’s Efficiency Crisis
AI’s explosive growth has pushed GPU demand through the roof, but many companies are failing to get the most out of their hardware. Billions of dollars’ worth of GPUs are running well below their potential, with frequent job restarts, crashes, and wasted energy. FleetIQ is Clockwork’s answer to this problem.
Built on research from Stanford, the platform acts as a software-driven fabric (SDF), sitting between workloads and infrastructure. It intelligently manages communication across GPUs, clusters, and clouds, helping companies extract more value from their existing systems.
Key features of FleetIQ:
- Improves communication between GPUs and clusters
- Microsecond-level visibility across fleets and workloads
- Real-time fault tolerance and routing to avoid failures
- Works across diverse environments: NVIDIA, AMD, custom accelerators
- Compatible with InfiniBand, Ethernet, and RoCE
FleetIQ is designed to cut down system downtime and prevent bottlenecks that stall AI operations. It addresses a key issue in today’s AI stack: even when compute is available, poor communication between components leads to constant delays and crashes.
Uber and Analysts Endorse the Innovation
The product is already showing impact in enterprise settings. Uber, one of Clockwork’s early users, highlighted the significant improvements in network observability and fault detection the platform provided. According to Uber, this tech reduced issue detection time from hours to just minutes.
Industry experts are also taking notice. Dylan Patel, CEO of SemiAnalysis, emphasized that communication has become the new bottleneck in AI infrastructure, saying, “With accelerators running in lockstep, a single link flap or congestion spike can crater utilization.”
Backing From Industry Heavyweights
This funding round marks a major milestone for Clockwork, valuing the company at four times its previous round just two years ago. The presence of chip giants AMD and Broadcom not only adds financial weight but strategic depth, reinforcing Clockwork’s position in the market.
NEA’s involvement further validates the company’s trajectory. With over $27 billion in assets under management and a track record of 280+ IPOs, NEA’s support signals confidence in Clockwork’s long-term vision.
SQ Magazine Takeaway
Honestly, I think Clockwork is solving one of the least glamorous but most crucial problems in AI right now. Everyone talks about training huge models, but nobody wants to admit how much GPU power gets wasted. Clockwork’s approach makes AI smarter not by building faster chips, but by using what we already have more intelligently. FleetIQ feels like the kind of tool that becomes essential behind the scenes in every major AI deployment. If they can deliver at scale, this might be one of the most important infrastructure plays of the decade.
