Nexus in every data centre. Offered by the operators your customers already trust.
Your customers want AI that stays inside your walls. Nexus lets data centre operators, sovereign and regional clouds and hyperscalers offer a complete AI backplane on their own GPUs: a private space for each customer, the knowledge graph, agents and governance included. And you can run your own sites on it too, from energy reporting to the certificates your customers ask for.
From your halls to your customers, in six steps.
Nexus installs on the infrastructure you already run. Each step builds on the last, and your own operations can move onto it at any point.
Install in your halls
On your own clusters, as a machine image or on appliances you ship. Nexus runs fully inside your boundary.
Add your GPUs
Point Nexus at your graphics cards and the models you host. Requests are spread across them by rules you set.
Offer it to customers
Each customer gets its own organisation and its own space, with walls between them that are tested every hour.
Meter and bill
Use is counted per customer, unit and model, so you can bill for it and plan capacity ahead.
Customers build on it
Your customers bring their data, build their ontology and switch on the platforms they need.
Run your own sites
Use the same platform for your own obligations: energy reporting, grid rules, NIS2, certifications and physical security.
A full AI backplane, sold from your own halls.
Customers sign in to a platform that runs in your data centre and keeps their data there. Each one is isolated from the others, and each can switch on the capabilities it needs: the knowledge graph, agents, collaboration and governance.
A private space for each customer
Every customer has its own space in the cluster, with limits it cannot exceed and walls it cannot cross.
- Isolated per customer and unit
- Boundaries tested every hour
- Room given without downtime
Ship it in a box
For customers who want it on their own premises, Nexus arrives on an appliance that checks itself before it is plugged in.
- Checked on arrival
- Claimed by the customer
- Managed from your side
Your graphics cards, shared fairly and fully used.
Pool your GPUs, give each customer a share, and let requests spread across local cards and any hosted capacity you allow. Budgets and quotas stop one customer from crowding out the rest.
GPU pools
- Shares per customer
- Training and serving pools
- Queues you can reorder
GPU budgets and quotas
- Hours per customer
- Warnings before limits
- Raised with approval
Spreading the load
- Local first, by rule
- Fallbacks you approve
- Data kept where it must be
Every request counted, ready to bill.
Use is counted per customer, unit, model and job, so billing is a report rather than a project. The same figures forecast the room your clusters will need over the next 90 days.
Usage explorer
- Per customer, unit and model
- Exported for billing
- Spikes explained
Cost and capacity ahead
- 90-day forecast
- Cost per extra node
- Before customers notice
Run your own sites on the same platform.
Operators carry a long list of obligations: energy reporting, grid connection rules, NIS2, financial-sector rules for their customers, certifications and physical security. Nexus keeps them in one place, with the evidence gathered from your own systems.
Every obligation, one place
- All rules, all sites
- Deadlines tracked
- Evidence linked
One site
- Power, cooling and halls
- Grid rules per site
- History kept
Energy reporting
- Figures from your own meters
- Filed on time
- Year on year
Halls and racks
- Space and power per hall
- Cooling at a glance
- Capacity for new customers
Physical security
- Visitors and access
- Doors and cameras watched
- Records kept
Answer security questions with proof, not attachments.
Banks, insurers and public bodies ask their providers for evidence. Give them proof they can check themselves, and keep your certifications current with evidence that collects itself.
Customer assurance
- Answers linked to evidence
- Checked by the customer
- Reused across questionnaires
DORA for your financial customers
Give each bank and insurer what DORA asks of its providers.
- Per customer
- Exit plans and tests
- Incident clocks
Certification
- Controls and evidence
- Findings closed on time
- An auditor’s room
Your customers’ data never leaves your country.
Show an examiner which country each kind of data is stored and processed in. Nexus runs with no outside connection if you need it to, and every piece of proof says where it came from.
- Data location per kind of data
- Runs fully offline if required
- Every piece of proof traced to its source
For data centre operators.
Who is this for?
Colocation and data centre operators, sovereign and regional clouds, GPU clouds and hyperscalers. Anyone who runs compute that customers want AI on.
Do we need our own GPUs?
It helps, because most customers want their AI close to their data. Nexus also works with hosted models where your rules allow.
Can customers see each other?
No. Each customer has its own organisation and its own space in the cluster, and the boundaries between them are tested every hour.
Can it carry our name?
Customers can reach it at your own address. Talk to us about how the partnership works.
Can we run our own operations on it too?
Yes. The same platform keeps your energy reporting, grid rules, NIS2, certifications and physical security in one place.
Go deeper
Offer Nexus from your data centre.
Tell us about your halls, your GPUs and your customers. We will show you what running Nexus for them looks like.