Industries · Sovereign AI

Your AI, your data, your infrastructure.

Nations and national clouds building sovereign AI must move massive training data in-country, at line rate, and under their own control. Zettar fills national-scale networks, runs on HPE ProLiant enterprise servers, stays on-premises, and keeps every byte encrypted and accountable — so your AI factory stays fed without surrendering sovereignty.

Built forNational AISovereign cloudsGovernmentNational research networks
The problem

Sovereignty keeps data in-country — but the data still has to move.

  • Petabytes of training data still have to move between national sites, fast.
  • National GPU factories sit idle waiting on data — the most expensive infrastructure you own, earning nothing.
  • Vendor lock-in is unacceptable — national data has to stay in open, native formats under national control.
  • Security, provenance, and jurisdiction are non-negotiable — every byte must be encrypted and accountable.
How Zettar helps

National-scale data movement, on your terms.

Fills national networks

Line rate, no ceiling

Linear scale-out with no software ceiling. Where the network provides it — up to 1.6 Tbps on Taiwan's research networks — zx scales out to fill it. At NCHC in Taiwan, two Mini appliances were measured at about 91% of their 10 Gbps capacity.

Sovereign by design

On-premises, in-country

Built on HPE ProLiant enterprise servers and run entirely on your premises — no forced cloud, so data need not leave the country.

Feeds the AI factory

Keep GPUs fed

Scale-out file, object, and streaming movement that keeps national GPU clusters fed — turnkey and pretuned, so your teams and hardware vendors tune nothing.

Secure & accountable

End-to-end integrity

Encrypted with TLS and unconditional end-to-end checksums, with roots in U.S. DOE exascale science — distance-insensitive, validated over real 5,000- and 12,375-mile transfers.

Proof, not promises
Taiwan's National Center for High-performance Computing (NCHC) deployed two Zettar Mini appliances between its Tainan and Taichung datacenters, running at 9.1 Gbps — about 91% of the Mini's 10 Gbps capacity.
— NCHC deployment, Taiwan (2025)

Backed by the record: 1 PB in 29 hours at 96% link utilization with SLAC & ESnet (U.S. DOE). Read the record →

≈91%
of a Mini appliance’s 10 Gbps capacity at NCHC — line rate at the edge
Build your business case

Put a number on it — and make your case.

Sovereign AI programs are judged on time to capability. Every week data sits in transit is a week the national cluster isn’t delivering — while the hardware depreciates.

FAQ

Common questions

What is sovereign AI data movement?

Moving the massive training data behind national AI in-country, at line rate, on infrastructure the nation controls — without surrendering data sovereignty to a foreign cloud.

Does Zettar keep data in-country?

Yes — zx runs on-premises and in-country, with no forced cloud, so data stays under your jurisdiction and control.

Are we locked in?

No — zx moves data in native formats with no proprietary wrappers, so nothing is trapped, and the appliance runs entirely on your premises, in-country, under your control. It is built on HPE ProLiant enterprise servers.

How much of our national network can it use?

There is no software ceiling — zx scales out to fill national-scale networks, up to 1.6 Tbps where the network provides it, as on Taiwan's research networks.

What security controls does it provide?

Every transfer is encrypted with TLS and protected by unconditional end-to-end checksums, on a data mover with roots in U.S. DOE exascale science.

Get started

Feed your AI factory — without surrendering sovereignty.

See the zx Appliance fill national-scale networks at line rate, on-premises and under your control.