WILMINGTON, DE — Zymtrace said it has raised $12.2 million to date, including a newly closed $8.5 million seed round, to expand its platform designed to identify performance bottlenecks in enterprise GPU clusters used for artificial intelligence workloads.
The company said the seed round was led by Venture Guides, with participation from Mango Capital, Fly Ventures, and 6 Degrees Capital.
Strategic angel investors in the round included Thomas Wolf, co-founder of Hugging Face; Christian Bach, founder of Netlify; AI systems optimization specialist Christopher Fregly; and Reece Chowdhry of Concept Ventures.
Zymtrace previously raised an unannounced $3.7 million pre-seed round led by Fly Ventures and Mango Capital, with participation from Entropy Industrial Capital.
The company said the new funding will support product development, expanded enterprise deployments, and growth of its U.S. go-to-market team.
Zymtrace’s platform is designed to analyze distributed computing systems to identify inefficiencies in how GPUs and CPUs process AI workloads. The company said many GPU clusters operate at roughly 35% to 40% utilization, leaving significant computing capacity unused.
When performance bottlenecks occur, engineers often must spend days or weeks investigating the cause using multiple tools, the company said.
Zymtrace’s software profiles GPU and CPU workloads across distributed systems and traces performance issues to specific code functions without requiring changes to the underlying application.
“The cheapest GPU you can buy is the one you already own,” said Israel Ogbole, co-founder and chief executive officer of Zymtrace. “The bottleneck is rarely the hardware. It’s the code that runs on it.”
Customers have used the platform to improve performance, the company said. Ben Carr, co-founder and chief technology officer of Anam, said Zymtrace helped his team identify inefficient workloads affecting GPU usage.
“Zymtrace pinpointed where our workloads were stalling and showed us how to resolve the issues,” Carr said. “We improved inference latency by 2.5x and increased throughput by 90% for our Cara3 model.”
Zymtrace said its platform uses an eBPF-based architecture that allows continuous system profiling with limited impact on performance.
The company said the technology can identify bottlenecks related to GPU kernel execution, memory usage, CPU scheduling, and distributed communication.
Investor Sage Nye, partner and founding team member at Venture Guides, said improving infrastructure efficiency will be increasingly important as demand for AI computing grows.
“As infrastructure increasingly becomes the limiting factor to growth, performance gains and efficiency aren’t optional,” Nye said.
Fly Ventures partner Fredrik Bergenlid said improving GPU utilization will be critical as computing costs rise.
“The future of AI won’t only be defined by who can acquire the most GPUs, but by who gets the most out of them,” Bergenlid said.
Zymtrace was founded by engineers who previously developed and open-sourced an eBPF CPU profiling agent later donated to OpenTelemetry while working at Elastic. The technology is now used in production by companies including Cisco, Datadog, Grafana, and IBM.
The company said it aims to help enterprises improve throughput, reduce infrastructure costs, and identify performance bottlenecks in large-scale AI systems.
For the latest news on everything happening in Chester County and the surrounding area, be sure to follow MyChesCo on Google News and MSN.