Home/Partners
OEM, distributor & channel partnersWe sell your hardware. We do not sell against it.
Velyrix is the deployment, commissioning and managed operations arm for GPU clusters sold through the channel. Our Bring Your Own GPU model means every customer we bring to a hall is a customer buying servers from an OEM — not from us.
A neocloud that increases your sell-through instead of replacing it
Most cloud providers buy hardware once and then rent it out, competing with your direct and channel business for the same end customer. Velyrix is built the other way round.
BYOG drives hardware sales
Our flagship model is customer-owned hardware. Every BYOG cluster we host is a purchase order placed with an OEM or distributor — frequently one that would otherwise have gone to a rental cloud and been lost to the channel entirely.
Channel neutrality
We do not resell servers as a business line. Where a customer asks us to transact, we do it at a disclosed fee, through your authorized channel, and we register the opportunity with you first.
We de-risk your deals
The most common reason a GPU deal stalls is that the customer has nowhere to put the hardware and no team to run it. We remove both objections, which shortens your sales cycle.
Services delivery arm
Deployment, commissioning, fabric build and managed operations delivered under your brand or ours, so you can quote a complete solution without building a field services organisation.
Benchmark & POC lab
Space, power and engineering time for customer proofs of concept, competitive benchmarks and reference architecture validation on your platforms.
Geographic reach
25 markets across 18 countries in North America, Europe and the Asia-Pacific allied economies, with partner facilities for receiving, staging and field engineering. Ashburn and Dallas are live today; other markets are contracted against a specific opportunity. Every location is export-clean for the platforms you sell.
Five ways to work together
Named deployment and field services partner for your AI infrastructure customers. Joint reference architectures, validated deployment playbooks per platform, engineer training on your systems, and a coordinated escalation path into your support organisation.
Velyrix as the fulfilment destination for GPU clusters moving through distribution: receiving, staging, integration, burn-in and drop-ship to a hosted environment, with asset and serial reporting back to you.
Resell Velyrix colocation, BYOG hosting, deployment and managed services alongside your hardware quote. Margin on services, deal registration protection and pre-sales engineering support on the call with you.
We deliver deployment, commissioning and support under your brand, in your customer's facility or ours, with your logo on the acceptance report. Subcontract agreement, NDA and non-solicitation as standard.
Introduce a customer who needs power, cooling or operations, and receive a referral fee on the contracted value. No certification, no quota, no commitment.
An engineering bench you can quote without hiring one
Receiving & logistics
Secure dock, appointment scheduling, inbound inspection with photographic record, serial and asset capture, de-trash and disposal, staging space and customs documentation support.
Integration & rack build
Rack, cable, label and dress to plan, including rail kits, PDU mapping, coolant plumbing and manifold work for liquid-cooled platforms. Installed to your published procedures.
Fabric engineering
Rail-optimised InfiniBand and Spectrum-X design, cable plans, optics, link training, BER sweeps and topology verification against the reference architecture.
Commissioning & ATP
72-hour burn-in, DCGM diagnostics, NVLink and NCCL benchmarks, thermal survey and a signed acceptance test report your customer can hand to their auditors — co-branded on request.
Warranty-preserving install
ESD-controlled handling to ANSI/ESD S20.20, torque specifications, firmware baselines and serial-level records, so manufacturer warranty and support entitlements stay intact.
Install-base reporting
Quarterly reporting of deployed assets, serials, firmware levels, failure rates and RMA history for your platforms in our halls — useful field data you rarely get from a cloud operator.
Sales enablement
Joint account planning, pre-sales engineering on customer calls, sizing and TCO models, site tours and reference customers under NDA.
Co-marketing
Joint reference architectures, case studies, webinars and event presence, with MDF-eligible activity planned quarterly.
Demo & POC hosting
Short-term hosted environments for customer evaluations and competitive benchmarks on your hardware, with engineering support included.
Clear lines between your support organisation and ours
Ambiguity about who owns a ticket is what turns a failed GPU into an unhappy customer. The split below is agreed in writing before the first deployment.
| Tier | Owned by | Scope | Response target |
|---|---|---|---|
| L1 — Monitoring & triage | Velyrix NOC | Alerting, first diagnosis, node drain, job protection, customer notification, ticket ownership | 15 min (P1) · 1 h (P2) |
| L2 — Field engineering | Velyrix | On-site diagnosis, component swap from spares, cabling and optics replacement, re-validation and return to service | 4 h on site |
| L3 — Platform engineering | Velyrix + OEM | Firmware and driver defects, fabric anomalies, performance regressions, reproduction cases and escalation packages | Next business day |
| L4 — OEM / vendor | OEM | Hardware RMA, engineering change notices, out-of-spec failure analysis, warranty determination | Per your contract |
RMA coordination
We open the case under the customer's or partner's support contract, supply logs, serials and failure evidence in your required format, track the dispatch, fit the part and close the loop with a written resolution.
Spare parts depot
On-site consignment or Velyrix-held spares for GPUs, PSUs, NICs, optics, DIMMs and NVMe, sized to fleet population and agreed failure rates, so a four-hour replacement target is realistic.
Failure analytics
XID, ECC, thermal and link-error trending per platform and per batch, shared with your engineering teams — early warning on a bad production lot benefits everyone.
Deployment playbooks we already hold
Each platform below has a validated Velyrix deployment playbook: rack elevation, power schedule, cooling method, cable plan and acceptance test profile.
| OEM | Representative platforms | GPUs | Cooling | Playbook |
|---|---|---|---|---|
| Dell Technologies | PowerEdge XE9680, XE9680L, XE9712, XE7745 | H100, H200, B200, GB200/GB300 | Air, DLC | Validated |
| Supermicro | SYS-821GE-TNHR, SYS-A21GE-NBRT, 4U DLC, GB200/GB300 NVL72 | H100, H200, B200, B300, GB300 | Air, DLC | Validated |
| GIGABYTE | G593 series, G894 series, XL44 rack-scale, G4A3 | H100, H200, B200, B300 | Air, DLC | Validated |
| HPE | Cray XD670, ProLiant XD685, Cray Supercomputing EX | H100, H200, B200 | Air, DLC | Validated |
| Lenovo | ThinkSystem SR675 V3, SR685a V3, SR780a V3, Neptune DWC | H100, H200, B200 | Air, DWC liquid | Validated |
| NVIDIA | DGX H100, DGX H200, DGX B200, DGX SuperPOD reference | H100, H200, B200 | Air, DLC | Validated |
| ASUS | ESC N8-E11, ESC NB8I-E11 | H100, H200, B200 | Air, DLC | On request |
| QCT | QuantaGrid D74H-7U, S74G-2U | H100, H200, GB200 | Air, DLC | On request |
| Wiwynn / Foxconn | ORv3 and MGX rack-scale designs | B200, B300, GB300 | DLC | On request |
Platform names are trademarks of their respective owners and are listed to describe deployment capability. Inclusion in this matrix indicates that Velyrix holds a deployment playbook for the platform; it does not by itself indicate a commercial partnership, certification or endorsement by the manufacturer.
The rules we hold ourselves to
- Deal registration first. Where a partner registers an opportunity, we protect it for the agreed period and will not approach that account with a competing motion.
- No end-run. We do not use a partner-introduced customer relationship to displace that partner's hardware position at renewal.
- Transparent transacting. If we are asked to buy hardware on a customer's behalf, we disclose the fee and buy through an authorized channel.
- Trademark discipline. We use partner names and marks descriptively and to your brand guidelines, and we do not claim certifications or partner status we have not been granted.
- Confidentiality. Mutual NDA before pricing or roadmap discussion; customer data never shared between partners.
- Trade compliance. Restricted-party screening, end-user verification and anti-diversion controls on every deal involving controlled hardware — protecting your compliance position as well as ours.
Register a deal or start a partnership
Send the details below to [email protected] and a partner manager will respond within one business day.
- Partner company, country and primary contact
- End customer and opportunity value (or "confidential")
- Platform and quantity — e.g. 16× 8-GPU H200 nodes
- Region and target site, or "Velyrix to propose"
- Services required: hosting, deployment, fabric, managed operations
- Expected decision and delivery dates
Onboarding is normally: mutual NDA, partner agreement, technical and commercial enablement session, then first joint opportunity — typically two to three weeks.
Own the GPUs. Let us run them.
Buy your NVIDIA servers from any OEM or distributor you like, ship them to a Velyrix hall, and we handle the rest — deployment, fabric, cooling, monitoring and support. Or rent ours. Either way, you get a plan in one business day.
Frequently asked questions
Does Velyrix compete with its OEM and channel partners?
No. Velyrix does not run a hardware resale business line. Our flagship BYOG model has the end customer purchase servers from an authorized OEM, distributor or reseller, which means a Velyrix engagement typically creates a hardware purchase order for the channel rather than removing one.
Can we sell Velyrix services under our own brand?
Yes. White-label deployment, commissioning and managed operations are delivered under the partner's brand, including co-branded or partner-branded acceptance reports, governed by a subcontract agreement with NDA and non-solicitation terms.
How does deal registration work?
Send the opportunity details to [email protected]. Registered deals are acknowledged within one business day and protected for an agreed period, during which Velyrix will not pursue a competing motion into that account.
Who owns the customer support relationship?
The split is agreed in writing per engagement. Typically Velyrix owns L1 monitoring and triage and L2 field engineering, Velyrix and the OEM share L3 platform engineering, and hardware RMA and warranty determination sit with the OEM under the customer's or partner's support contract.
Do you hold spare parts for our platforms?
Yes, either as partner-owned consignment stock or Velyrix-held spares, sized against the deployed fleet population and agreed failure rates so that a four-hour on-site replacement target is achievable.
Will your engineers install our systems to our documented procedures?
Yes. Each supported platform has a Velyrix deployment playbook built from the manufacturer's published installation guidance, with ESD control to ANSI/ESD S20.20, torque specifications, firmware baselines and serial-level documentation, so warranty and support entitlements are preserved.