Company: PEAK:AiO
Founder and CSO, President, CEO: Mark Klarzynski, Roger Cummings
Website: www.peakaio.com
About PEAK:AiO
Headquartered in Manchester, UK, PEAK:AIO builds AI storage infrastructure that makes GPU compute immediately productive. PEAK:AIO was established based on a straightforward observation: the high cost of storage for modern AI tasks is often prohibitive, leaving billions of pounds in GPU hardware underutilized.
The company’s story starts with the NHS in 2019 when the NHS AIDE program set out to build what would become the world’s largest federated clinical AI network comprising 10 NHS Trusts, 18 million patients, and a complete NVIDIA DGX stack. During the implementation of this comprehensive system, the delivery team recognized that available enterprise storage solutions would drain a major part of the funding intended for GPU resources. In such a context, any capital diverted to storage directly reduces the investment available for clinical AI models essential for saving lives.
PEAK:AIO was built to solve that arithmetic problem: a platform that starts at a single server from any vendor, deploys in hours, and delivers enterprise-grade performance at a fraction of the cost of alternatives.
That constraint turned out to also define the majority of the AI storage market globally. Enterprise storage vendors require 8 to 12 server deployments costing anywhere from £300,000 to £450,000 just to get started, while PEAK:AIO can start with a single server, typically costing around £40,000. 80% of AI deployments running on 1 to 128 GPU clusters, including healthcare, research, national laboratories, pharmaceuticals, robotics, and government, have no other option purpose-built for their scale.
Real-world results consistently validate this approach. At Los Alamos National Laboratory, PEAK:AIO delivered a 5.7x throughput improvement on existing server hardware with no new capital expenditure, taking the same servers from 70 GB/s to 400 GB/s throughput with software alone.
Work at the Zoological Society of London saw image-processing performance accelerate from 3 images every minute to 30 images every second. This 600x gain allowed for the processing of datasets in just one day that previously required months. A 2017 Dell server at a Tier 1 University achieved 92.5 percent of the theoretical maximum for a 100-gigabit network, confirming linear performance scaling from 1 to 20 nodes.
The company recently unveiled the first open-source pNFS metadata server, a milestone achieved through collaboration with Los Alamos National Laboratory (LANL). According to Gary Grider, HPC Division Leader at LANL, the partnership with PEAK:AIO achieved more progress in metadata scalability in six months than in the prior six years.
PEAK:AIO customers include Los Alamos National Laboratory, NHS AIDE, Oxford Robotics Institute, University of Liverpool, University of Strathclyde MediForge Hub, and the Zoological Society of London.
From its UK roots, PEAK:AIO is growing rapidly across the US and international markets, supported by a network of OEM and technology partnerships. The company is proven at the heart of some of the world’s most demanding AI projects, providing a flexible, cost-effective, and high-performance foundation for modern AI infrastructure.
Want to find out what TechRound can do for your business?
