Who are SambaNova’s sovereign/in-country deployment partners (EU/UK/AU) and how do we engage them for procurement?

Sovereign and in‑country deployments are how SambaNova customers get frontier‑grade, open‑source AI performance while keeping data, compute, and operations inside national borders. If you’re operating in the EU, the UK, or Australia and you need compliant, high‑throughput inference close to your users, SambaNova works through a focused set of regional partners that run SambaNova infrastructure as sovereign AI clouds.

This guide walks through who those sovereign / in‑country deployment partners are and how to engage them for procurement, from initial scoping through contract and rollout.

The Quick Overview

• What It Is: A network of regional data center and cloud partners that operate SambaNova‑powered sovereign AI services in Europe, the UK, and Australia, with infrastructure, data, and operations kept in‑country or in‑region.
• Who It Is For: Public sector, highly regulated industries, and enterprises that must meet strict data residency, GDPR/EU AI Act, or national sovereignty requirements while running state‑of‑the‑art LLM and agentic workloads.
• Core Problem Solved: You get high‑throughput, low‑latency inference on open models without shipping sensitive data to foreign hyperscalers or building and operating your own data center stack.

How It Works

At a high level, SambaNova provides the full “chips‑to‑model” inference stack—SN50 and SN40L‑16 RDUs and racks, SambaStack, SambaOrchestrator, and OpenAI‑compatible APIs—while regional partners provide in‑country data centers, connectivity, and commercial relationships. Together, they deliver sovereign AI inference as a managed service or private deployment.

You choose the geography and sovereignty profile you need; SambaNova works with the appropriate partner to stand up or allocate capacity, and your teams integrate via standard, OpenAI‑compatible endpoints.

1. Workload and sovereignty scoping:
   You define workloads (LLMs, agents, RAG), compliance requirements (GDPR, EU AI Act, sector regulation), and residency constraints (EU‑only, UK‑only, AU‑only). SambaNova maps those requirements to the right in‑country partner and deployment pattern.

2. Solution design and capacity planning:
   SambaNova and the sovereign partner size the underlying SambaRack systems (SN50 or SN40L‑16), design the networking and security perimeter, and confirm model mix (e.g., Llama, DeepSeek, gpt‑oss). This is where model bundling strategies and expected tokens per second / tokens per watt are finalized.

3. Procurement and go‑live:
   Commercials are executed either directly with the regional partner (usage‑based cloud or reserved capacity) or through a joint agreement that may include SambaManaged options. After that, you connect via OpenAI‑compatible APIs, run validation, and move workloads into production with full in‑country compliance.

Sovereign / In‑Country Partners by Region

European Union: Infercom (SambaManaged‑Powered)

Region: EU (with a launch focus on Luxembourg and broader European coverage)
Service Type: Sovereign Inference‑as‑a‑Service, powered by SambaManaged

Infercom operates what SambaNova has announced as Europe’s first sovereign AI inference service, powered by SambaNova’s full stack. The service is:

• Designed for EU compliance: Built from the ground up to support GDPR and EU AI Act requirements.
• Inference‑first: Optimized for high‑throughput, low‑latency generative AI workloads rather than generic cloud compute.
• Open model friendly: Delivers the performance of frontier‑class open models while keeping data and models in European data centers.

Infercom is best suited when:

• You need EU‑resident inference for multi‑tenant or dedicated workloads.
• You want to avoid the overhead of running your own data center stack but still require full data sovereignty.
• You care about operational proof—Infercom and SambaNova jointly design for compliance and performance, rather than retrofitting generic infrastructure.

United Kingdom: Argyll (Sovereign AI Cloud)

Region: United Kingdom
Service Type: Renewable‑powered sovereign AI cloud powered by SambaNova

Argyll has partnered with SambaNova to deliver the UK’s first renewable‑powered sovereign AI cloud, designed for organizations that want in‑country AI while also improving their energy profile.

Key characteristics:

• UK‑resident compute and data: Designed so workloads, logs, and model artifacts remain within the UK.
• Renewable‑powered infrastructure: Targets organizations with both sovereignty and sustainability mandates.
• SambaNova full stack: SambaRack systems with RDUs plus SambaStack and SambaOrchestrator, exposed via standard APIs.

Argyll is ideal when:

• You require UK‑only data residency and are subject to UK‑specific regulatory or public sector guidelines.
• Sustainability (carbon targets, renewable sourcing) sits alongside performance and sovereignty as a procurement driver.
• You want a cloud‑like procurement model—capacity consumption inside the UK without running your own racks.

Europe (Broader): OVHcloud for Flagship AI Inference

Region: Pan‑European, with data center locations across the EU
Service Type: AI endpoints and inference services powered by SambaNova

OVHcloud has selected SambaNova to power its flagship AI endpoints inferencing service. This gives European customers a familiar, large‑scale cloud provider with SambaNova hardware and inference stack under the hood.

You get:

• European data centers: For workloads that must stay within OVHcloud’s EU facilities.
• Elastic AI endpoints: Cloud‑style provisioning for LLM inference, backed by SambaNova’s RDU architecture.
• Open ecosystem alignment: OVHcloud’s open cloud stance combined with SambaNova’s emphasis on open‑source models.

OVHcloud is a fit when:

• You want AI inference integrated into broader cloud workloads already running on OVHcloud.
• You prefer dealing with a major European cloud provider while still getting SambaNova performance characteristics.
• You need multi‑region EU coverage rather than a single‑country sovereign profile.

Australia: Southern Cross AI (Sovereign AI Cloud)

Region: Australia
Service Type: Sovereign AI cloud powered by SambaNova

Southern Cross AI (referenced by SambaNova as a sovereign partner in Australia) focuses on in‑country AI for Australian organizations, powered by SambaNova’s inference stack.

Typical attributes:

• Australian data residency: Designed to keep sensitive workloads within Australian borders.
• Sovereign focus: Tailored to government, defense, and regulated industries with strict locality rules.
• Integrated SambaNova stack: RDUs, tiered memory architecture, SambaStack, and orchestration managed on your behalf.

Southern Cross AI is best when:

• You must adhere to Australian data sovereignty and public sector procurement requirements.
• Latency to Australian end‑users is a priority and you want inference closer to where your data is generated.
• You’re not interested in building your own data center capex footprint but need enterprise‑grade performance.

Note: The exact list of sovereign partners in each region evolves. For the most current partner roster and capabilities, contact SambaNova directly.

Why These Partners Matter for Sovereign AI

All of these partners operate on a common technical foundation:

• RDUs with three‑tier memory architecture: The SN50 and SN40L‑16 are designed to keep models and prompts hot, reducing data movement and increasing tokens per watt—critical when power and cooling are constrained in sovereign facilities.
• Model bundling on SambaStack: Multiple frontier‑scale models can be bundled on one node, so complex agent workflows don’t bounce between endpoints or data centers. This is especially important for sovereign environments where cross‑border routing is not acceptable.
• SambaOrchestrator for operations: Autoscaling | Load Balancing | Monitoring | Model Management across in‑country racks, giving both the partner and you clear operational controls and observability.
• OpenAI‑compatible APIs: You can port your application in minutes from existing OpenAI‑style integrations to these sovereign environments, minimizing refactor risk.

The result is sovereign AI that behaves like modern cloud inference—but with the guarantees, locality, and controls required by national regulations.

How to Engage These Partners for Procurement

Step 1: Define Your Sovereignty and Workload Requirements

Before you start the procurement conversation, align internally on:

• Jurisdiction: EU, UK, Australia, or multi‑region.
• Data residency: Hard constraints (e.g., “must not leave EU”) vs. strong preferences.
• Compliance: GDPR, EU AI Act, local data protection laws, sector regulators (finance, health, defense).
• Workload profile:
  • Types of models (e.g., Llama, DeepSeek, gpt‑oss)
  • Agentic patterns (multi‑step workflows, tool‑use, long‑running sessions)
  • Expected token throughput (tokens/second) and concurrency.
• Connectivity model: Private network links, VPN, zero‑trust access, or public internet endpoints.

Capturing this upfront lets SambaNova and the partner size the right combination of SN50 vs. SN40L‑16, and decide how much capacity to reserve for peak workloads.

Step 2: Contact SambaNova as the Front Door

While you can approach some partners directly, the most efficient route—especially for complex workloads—is to engage SambaNova first:

• Use the contact form:
  Get Started → https://sambanova.ai/contact
• Provide:
  • Regions (EU / UK / AU) and sovereignty constraints
  • Industry and regulatory context
  • Estimated workloads (models, TPS, data sensitivity)
  • Timeline and procurement structure (RFP, pilot, direct purchase)

SambaNova’s team will:

• Map your requirements to the appropriate regional sovereign partner (Infercom, Argyll, OVHcloud, Southern Cross AI, or a combination).
• Determine whether you need multi‑partner coverage (e.g., EU‑plus‑UK) for distributed operations.
• Propose deployment patterns (multi‑tenant managed service vs. dedicated clusters).

Step 3: Joint Solution Architecture and Sizing

In this phase, SambaNova and the in‑country partner collaborate with your platform and security teams to define:

• Topology:
  • Rack count (SambaRack SN50 vs. SN40L‑16), power envelope, cooling profile.
  • Network segmentation, private connectivity, and access controls.
• Model portfolio and bundling:
  • Which models to keep hot in tiered memory for fast agent loops.
  • How to bundle models so one node can serve multi‑step workflows without cross‑border traffic.
• SLA and performance targets:
  • Tokens per second per model (e.g., gpt‑oss‑120b at over 600 tokens/second, DeepSeek‑R1 up to 200 tokens/second as independently measured by Artificial Analysis, depending on configuration).
  • Latency objectives, availability targets, and autoscaling behavior.

Output from this phase typically includes:

• A reference architecture diagram for your region(s).
• A capacity plan tied to your projected usage and growth.
• A security and compliance outline to align with your internal risk teams.

Step 4: Commercial Procurement with the Regional Partner

Procurement usually flows through the sovereign partner, with SambaNova embedded as the technology provider. Typical options include:

• Usage‑based cloud model:

  • Pay per token, per hour, or per endpoint usage through the partner’s portal.
  • Best if you need elasticity and want to avoid capacity planning complexity.

• Reserved or dedicated capacity:

  • Commit to a specific capacity slice (e.g., dedicated SambaRack deployments) in the in‑country data center.
  • Best for predictable, high‑throughput workloads and tighter isolation requirements.

• SambaManaged + Partner:

  • In some cases, SambaNova provides SambaManaged operational services atop the partner’s infrastructure, combining SambaNova’s operational maturity with local data center presence.
  • Best when you want SambaNova deeply involved in ongoing operations and optimization.

Procurement teams will work with:

• The regional partner as the primary commercial counterparty.
• SambaNova for technical validation, performance attestations, and architecture guarantees.

Step 5: Integration, Testing, and Production Cutover

Once contracts are in place:

1. Endpoint provisioning:

   • The partner exposes OpenAI‑compatible endpoints (via SambaCloud‑style APIs) in your selected region.
   • You receive API keys, networking details, and environment documentation.

2. Pilot and validation:

   • Port existing OpenAI‑compatible applications in minutes by updating base URLs and keys.
   • Validate latency, throughput, and correctness on your representative workloads.
   • Confirm logging, monitoring, and identity integration meet your internal standards.

3. Scaled rollout:

   • Production traffic is gradually shifted to the sovereign environment.
   • SambaOrchestrator metrics are used to tune autoscaling | load balancing | model management as load grows.
   • You iterate on model bundling and tiered memory configurations to keep key models and prompts hot.

Features & Benefits Breakdown

Core Feature	What It Does	Primary Benefit
In‑country sovereign data centers	Keeps compute, storage, and logs within EU, UK, or AU borders	Meets data residency and sovereignty mandates without building your own facilities
SambaNova RDUs + tiered memory	Reduces data movement and keeps models/prompts hot across three memory tiers	Maximizes tokens per watt and throughput under strict power/cooling envelopes
OpenAI‑compatible sovereign APIs	Exposes LLMs and agents via familiar OpenAI‑style endpoints	Lets teams port applications in minutes, minimizing integration risk and time to production

Ideal Use Cases

• Best for regulated AI workloads: Because it keeps inference in‑country (EU/UK/AU) with infrastructure tuned for GDPR, EU AI Act, and public sector standards, while still delivering frontier‑class throughput.
• Best for agentic and multi‑model workflows: Because SambaStack and RDUs enable model bundling and fast switching between large models on a single node, avoiding cross‑region hops between endpoints in sovereign environments.

Limitations & Considerations

• Region availability and partner coverage: The exact capabilities and locations of sovereign partners evolve over time. For the latest partner list and regional SLAs, engage SambaNova directly to avoid designing around outdated assumptions.
• Capacity planning for large deployments: Sovereign environments often operate under stricter power, space, and change‑control constraints than hyperscale clouds. Large deployments may require longer lead times for rack capacity; early engagement with SambaNova and the partner mitigates this.

Pricing & Plans

Pricing for SambaNova’s sovereign / in‑country deployments is typically set and billed by the regional partner, with structure depending on your usage profile and compliance needs.

Common patterns include:

• Elastic Sovereign Cloud:
  Usage‑based pricing (per token or per endpoint) managed by the partner (Infercom, Argyll, OVHcloud, Southern Cross AI).

  • Best for: Teams needing flexible capacity and fast time‑to‑value without long‑term hardware commitments.

• Reserved / Dedicated Sovereign Capacity:
  Committed capacity on dedicated SambaRack systems inside the partner’s facilities, potentially combined with SambaManaged services.

  • Best for: Enterprises and agencies with predictable, high‑volume workloads who want cost predictability, isolation, and long‑term optimization.

For current pricing options and to align with your procurement model, contact SambaNova:

• Get Started: https://sambanova.ai/contact

Plan Fit Summary

• Elastic Sovereign Cloud: Best for teams needing rapid ramp‑up and variable load across EU/UK/AU without owning hardware, including early‑stage programs and mixed experimentation/production workloads.
• Reserved Sovereign Capacity: Best for large, always‑on workloads (contact centers, knowledge assistants, public services) where you want tighter control over capacity, performance, and cost per token.

Frequently Asked Questions

Which partner should I use if I operate across both the EU and the UK?

Short Answer: Engage SambaNova first; we’ll map your requirements across Infercom/OVHcloud for EU and Argyll for the UK based on where data must reside and how traffic flows.

Details:
Multi‑jurisdiction architectures often require strict boundaries—for example, EU citizen data that must stay in EU data centers, alongside UK workloads that must remain in the UK. SambaNova can help you:

• Use Infercom or OVHcloud for EU‑resident inference.
• Use Argyll for UK‑resident inference.
• Design routing, identity, and logging so requests are consistently served from the correct region.
• Harmonize model choices and APIs so your application code remains consistent while the infrastructure respects regional constraints.

By coming through SambaNova first, you avoid building ad‑hoc, inconsistent patterns with each partner and instead get a coherent, sovereign‑aware design.

Can I migrate existing OpenAI‑based applications to a sovereign partner without major rewrites?

Short Answer: Yes. SambaNova’s sovereign partners expose OpenAI‑compatible APIs, so you typically update endpoints and keys rather than rewriting application logic.

Details:
Because SambaNova’s inference stack is designed to be OpenAI compatible, most customers:

• Keep their existing client libraries and SDKs.
• Swap the base URL to the sovereign endpoint (EU/UK/AU) and rotate credentials.
• Validate behavior on a per‑model basis (e.g., moving from proprietary models to open models like Llama, DeepSeek, or gpt‑oss).

The heavy lifting—tokens per second, model bundling, tiered memory optimization—happens in the SambaNova stack operated by the partner. Your team focuses on policy, evaluation, and application logic, not on re‑architecting for a new API.

Summary

SambaNova’s sovereign and in‑country deployment partners—Infercom in the EU, Argyll in the UK, OVHcloud for flagship EU AI inference, and Southern Cross AI in Australia—provide a path to high‑performance, open‑model generative AI that stays inside your required borders. They run SambaNova’s full inference stack—RDUs, SambaRack, SambaStack, SambaOrchestrator, and OpenAI‑compatible APIs—so you can meet sovereignty and compliance requirements without sacrificing throughput, latency, or energy efficiency.

Instead of stitching together “one‑model‑per‑node” infrastructure across generic clouds, you get chips‑to‑model computing purpose‑built for agentic and multi‑model workloads, delivered in‑country by partners who understand local regulation and data‑center operations.

Next Step

Get Started