Meet Aleria at NVIDIA GTC Berlin · 21 and 22 October · GTC Berlin · Booth 3054Book 20 minutes
A desk in Abu Dhabi with a desktop, a laptop and a phone, all running Aleria An office floor where people and Aleria experts work together The company building, its systems connected through Aleria Fusion The city, with the company and an Aleria data center across it An Aleria data center with power, cooling and fibre A data hall with rows of GPU racks run as Aleria Cloud One rack with a tenant's NVIDIA GPU tray pulled out The UAE with its border lit The globe, with a beacon in every country that runs the stack Back on the desk: the board pack on the monitor, approved

Five layers of AI. One company. Inside your border.

AI that does real work has to read your ledgers, contracts and email. So Aleria builds and runs every layer it needs in your country, from the data center to the app on your screen.

Running today1,000+ customers in 48 countries see who

Example runA board pack through all five layers · 0 bytes left the country

Aleria as an exploded paper model in five plates. Top: people at desks with Aleria on their screens and an AI expert. Then the company's systems wired to one node, a gateway ring routing to models, rows of GPU racks with cooling, and at the bottom NVIDIA GPU trays and a power substation. A green thread runs from a screen at the top down through every plate.
  1. 01ApplicationsAleria OS · Video AIAI experts that do the work, for every employee.
  2. 02Your dataAleria FusionReads every system you run, with each person's permissions.
  3. 03ModelsInference GatewayAleria and open models behind one API.
  4. 04InfrastructureAleria Cloud · data centersA GPU cloud in data centers we build and run.
  5. 05Chips and energyNVIDIA · local power PartnersNVIDIA GPUs, and power from local partners.
The five layers of Aleria, inside your border: applications, your data, models, infrastructure, chips and energy.
Scroll

Now follow it.

The board pack goes down through every layer of Aleria and comes back done. An example run.

You ask Aleria OS for the quarterly board pack.

One chat for every employee, on any device. Behind it, AI experts do the work with your company's skills and systems.

A finance expert takes the job.

Experts are AI colleagues with a role, skills and access rights, as many as your company needs. This one plans the pack and shares out the work, next to your team.

Aleria Fusion reads every system the job needs.

Fusion connects the systems you run, from the ledger and the ERP to documents and email, and reads only what this person may open, where the data lives.

Inference Gateway sends each call to the right model.

One API in front of Aleria's own models and open ones, with keys, limits and logs. However many calls a job makes, each goes to the model that fits, inside the country.

The calls arrive at Aleria's own data centers.

Aleria builds and runs them itself, as an NVIDIA Cloud Partner. Inside are NVIDIA DGX systems and DDN storage, and we run the power, cooling, fibre and security around them.

Aleria Cloud gives the job the GPUs it needs.

From one GPU to whole clusters, it finds them in your own tenant, walled off from other customers, with your keys and your logs.

NVIDIA GPUs write the pack.

They run in NVIDIA Grace Blackwell systems, on power from local partners.

The whole trip stayed in one country, under its law.

The app, your data, the models, the GPUs and their power all ran inside its borders. The network log shows that 0 bytes of your data left the country.

Four ways to run itHosted by AleriaYour data centerAir-gappedA national AI factory

Each country gets its own closed loop.

The same five layers run separately in every country. A request in one country never touches another's. The next loop can be in your country.

Your board pack is ready.

Every figure is sourced and every risk the board should see is flagged. One company ran all five layers, inside one border.

An example run of one request.

Skip the example

The institutions that own their intelligence will define the next decade.

Eric LeandriChief Executive Officer, Aleria

So Aleria builds all five layers for them, runs them inside their own border, and answers for every one.

Your data stays inside your border, and you can prove it.

Put Aleria behind a firewall that blocks everything, run a week of real work, and read the log. You should find zero outbound connections.

How sovereignty works
  • Hosted by AleriaOur operators run it in our data centers, under UAE law. Start this week.
  • Your data centerOur engineers install it on your hardware, and it stays closed to us by default.
  • Air-gappedNo internet at any layer. Updates arrive as signed packages.
  • A national AI factoryWe build and run all five layers inside your country.

What customers ask us first.

Short answers here. The long ones, each with a check, are on the Sovereignty page.

Where do the GPUs run?

In Aleria's AI factories, which we build as an NVIDIA Cloud Partner, or on your own hardware. Each country's stack runs separately, so a request in one country never touches another country's capacity.

What does “inside your border” mean?

Every layer that touches your data runs in your country: the app, the connectors, the models, the GPUs, the logs and the keys. It is operated by Aleria under that country's law, or by your team on your own servers.

What leaves the country?

Nothing that contains your data. Software updates come in as signed packages, and the network log shows every byte that goes out, so your security team can check it.

What happens if we cut the internet?

An air-gapped deployment keeps working. Models, data and the app all run inside your network, and licensing works offline.

Who can see our data?

Only the people you allow. Fusion reads with each person's own permissions, and your keys encrypt the data. Aleria staff need your approval to access a tenant.

Can we run it in our own country?

Yes. Aleria OS, Fusion and the Gateway run on your own servers or your national cloud, in your country, under your rules. Or we build an AI factory with you there. See AI factories.

Do we have to buy all five layers?

No. Start at any layer: Aleria OS for a team from $29 a month, the Inference Gateway for developers, GPU capacity on Aleria Cloud, or a whole national AI factory.

See your own work done by Aleria.

Bring one report your team builds every week. We will run it on Aleria in the briefing.

At NVIDIA GTC Berlin? Book 20 minutes at booth 3054, 21 and 22 October