SambaNova backs sovereign, energy-efficient AI clouds across Australia, Europe, and the UK
SambaNova has signed three new sovereign AI agreements with SCX (Australia), Infercom (Europe), and Argyll (United Kingdom). The goal: deliver compliant, onshore AI inference at far lower energy draw than traditional GPU stacks-without major facility upgrades.
As inference overtakes training in real-world workloads, tokens per watt is becoming the metric that matters. These deployments aim to keep data local, cut operating costs, and run advanced models in places where electrical capacity is tight.
Why this matters
Data centers are hitting energy limits. IDC frames it simply: AI infrastructure today is like incandescent bulbs; SambaNova's approach is closer to LED-same output, dramatically less electricity and cost. For IT leaders, developers, and public sector teams, that means you can scale AI services without waiting on new substations, liquid cooling, or budget resets.
Three deals. One standard for efficient AI.
Australia: SCX
SCX is launching the country's first ASIC-based sovereign AI cloud, built with SambaNova. It runs entirely on renewables, offers roughly 10x lower inference energy costs, and meets national compliance requirements out of the box-built in Australia for Australian enterprises and agencies.
United Kingdom: Argyll
Argyll is rolling out the UK's first containerized sovereign inference cloud, running on wind, solar, and wave energy. Using air-cooled SN40L systems, the setup aligns with national data and energy strategies and gives British organizations a secure, domestic option for AI at scale.
Europe (Germany): Infercom
Infercom is launching Germany's first inference platform with full EU compliance. The service keeps sensitive data within European borders, meets GDPR expectations, and reduces energy use for startups, enterprises, and government users building production AI.
The technical angle
Traditional GPU racks can draw up to ~120 kW, which strains facilities and budgets. SambaNova's SN40L racks use about 10 kW-no liquid cooling, fewer upgrades, far lower OpEx.
Under the hood, SambaNova's Reconfigurable Dataflow Unit (RDU) targets high-throughput inference at a fraction of the electricity and cost. For operators, that means shorter lead times (SambaManaged can be live in ~90 days) and predictable scaling without ripping and replacing existing infrastructure.
What this means for IT, developers, and government
- Onshore by default: keep datasets and model outputs within national or EU borders.
- Compliance built-in: meet GDPR and sector rules without bolting on extra controls.
- Energy and cost control: 10 kW racks ease site constraints and reduce ongoing spend.
- Faster rollout: containerized options and air-cooled designs shorten deployment cycles.
- Inference at scale: optimize for tokens per watt as usage shifts from training to serving.
What to watch next
- Capacity ramps and regional availability timelines.
- Model support, latency benchmarks, and SLAs for public sector workloads.
- Integration paths with existing MLOps, data governance, and observability stacks.
Learn more
Organizations interested in sovereign, energy-efficient AI with strong compliance can visit sambanova.ai or email press@sambanova.ai.
If your team is building skills for AI deployment and governance, explore practical training by role at Complete AI Training.
About SambaNova
SambaNova enables enterprises to deploy advanced generative AI quickly and securely. Founded in 2017 and headquartered in Palo Alto, the company is backed by investors including SoftBank Vision Fund 2, BlackRock, Intel Capital, GV, Walden International, Temasek, GIC, Redline Capital, Atlantic Bridge Ventures, and Celesta. More at sambanova.ai or info@sambanova.ai.
Contacts
Press
SambaNova: Virginia Jamieson, virginia.jamieson@sambanova.ai, 650-279-8619
Your membership also unlocks:
