Quake AI, which runs about 22,000 Nvidia GPUs, wants partners to build services on bare metal. We set out the four hires we would make first for a UK GPU infrastructure practice, in order.
Quake AI, the infrastructure unit of Nasdaq-listed RUM Group, is looking for cloud service and channel partners to build services on its bare-metal GPU cloud, IT Europa reports. According to the report, it runs about 22,000 Nvidia H100 and H200 GPUs and uses shared co-location space in the UK, in Chester and London. For UK integrators and MSPs, the hiring decision is whether to build a GPU infrastructure practice of their own and, if so, which engineers to hire first.
Quake AI says it wants many more channel partners
Quake AI was formed after RUM completed its $767m acquisition of German firm Northern Data in June, IT Europa reports, and it has more than 50 end customers served from facilities across North America and Europe. Mike Masci, Chief Financial Officer at RUM Group, told IT Europa: "You can't just deploy a platform and go past the channel." He added: "We are embracing the channel, we love our existing partners, but we want many more." He also said: "We want partners that can offer their services on top of ours to their end customers." Masci acknowledged that "the partner ecosystem we want to be part of don't know us".
The offer is specialist. IT Europa lists bare-metal access, dedicated environments, customers' own operating images and encryption keys, and direct engineer support. Masci said: "Bare metal is a specialism," and argued that customers "want to be as close to the metal as possible."
Services on bare metal need infrastructure engineers, not only cloud sellers
Our view is that building services on bare metal is a different job from reselling cloud consumption. Without a managed software layer between the customer and the hardware, the partner usually takes on operating system images, drivers, cluster scheduling, high-speed networking, storage performance and key management. That is work for engineers with Linux and high-performance computing experience. For many UK resellers and MSPs, that pool is small, and it is also sought by AI companies, research computing teams and the cloud providers themselves.
Masci framed the competitive test as "who can keep their GPUs both running and highly optimised." Partners that sell services on top of a GPU cloud are likely to be judged on the same thing, which is why the order of hiring matters.
The first four hires for a GPU infrastructure practice, in order
1. A GPU infrastructure architect who has run clusters in production
In our view, the first hire should own the design and the customer's technical confidence. Look for someone who has built and operated GPU clusters at a cloud provider, a research centre or an AI company, rather than someone who has only sized hardware for a sale.
2. A Linux and HPC platform engineer
Because customers bring their own operating images, someone has to build, patch and support them alongside drivers and schedulers. We expect this to be the hardest role to fill, because many of the strongest candidates work outside the channel.
3. A pre-sales consultant who can size AI workloads
Masci said: "We want to seek out AI natives, as there is potentially more margin in them." Those customers are likely to ask whether bare metal or a hyperscaler suits their training or inference, so partners need a technical seller who can size GPUs, storage and networking and explain the trade-offs honestly.
4. A security engineer for keys and data location
Our view is that customer-held encryption keys and UK co-location give partners a credible offer for regulated buyers. Someone still has to design key management, access controls and evidence for auditors, and that is a security engineering role rather than a sales one.
Four checks before you commit to the first salary
- Test demand before the architect hire. Hold a few conversations with AI-native or research customers to confirm the work exists before you commit to a senior salary.
- Look beyond the channel for engineers. University research computing teams, HPC centres and data centre operators are likely sources of Linux and GPU skills.
- Extend the skills you already have. Your network and storage engineers may already cover part of the stack, so map their gaps before you advertise.
- Benchmark pay against the right market. GPU and HPC engineers are often priced against AI companies and cloud providers rather than channel employers, so check the real rate first.
The decision for UK integrators and MSPs is whether to build GPU infrastructure skills now, while providers such as Quake AI are still recruiting partners, or to wait and partner with someone who already has them. If you are weighing that choice, see how we support resellers and MSPs, use our salary benchmarking to price GPU and HPC roles before you advertise, or get in touch.
Sources
- Newly launched Quake AI in search for GPU channel partners, IT Europa, 22 October 2026
Get our view on your team.
Tell us a little about your business and a senior consultant will come back to you the same working day, in confidence.
Or book a call with Chris