Australia's sovereign AI infrastructure provider delivers ASIC-powered inferencing capacity to Equinix’s AI ecosystem through Equinix Fabric, unveils Australia's first sovereign LLM coding model, and secures next-generation SambaNova SN50 hardware for multi-site national rollout

SYDNEY: #ai--SouthernCrossAI (SCX), Australia's full-stack sovereign AI infrastructure company, today announced it is now providing sovereign AI inferencing capacity to the Equinix Fabric AI ecosystem, making its ASIC-powered inference nodes discoverable and accessible to enterprises, government agencies, and developers across the Equinix ecosystem.

Alongside this milestone, SCX confirmed that its national infrastructure expansion will be powered by SambaNova's next-generation SN50 Reconfigurable Dataflow Unit (RDU), the world's fastest AI chip purpose-built for agentic AI inference.

SCX Joins Equinix Fabric AI Ecosystem

SCX's inclusion in Equinix Fabric AI ecosystem represents a significant step in making sovereign Australian AI inferencing capacity available at scale. Through Equinix Fabric, Equinix's software-defined interconnection service, organisations across the Asia-Pacific region can now establish secure, low-latency, private connections directly to SCX's ASIC-accelerated inference nodes - without transiting public internet infrastructure or routing data through overseas hyperscaler environments.

The partnership builds on SCX's existing operational node at Equinix's SY5 IBX data centre in Sydney, and positions Equinix Fabric as the primary interconnection backbone for SCX's planned multi-site national network. Future nodes confirmed at Equinix facilities across Australia will progressively extend sovereign inferencing capacity to enterprise and government customers throughout 2026 and beyond.

"We are excited to welcome SCX to the Equinix Fabric AI ecosystem. SCX represents exactly the kind of innovative sovereign AI provider that enterprises and government agencies need to access through Equinix Fabric - a locally governed, high-performance inferencing capability that meets Australia's data sovereignty requirements. Together, we are building the infrastructure for Australia's AI future."

- Chris Johnston, Interim Managing Director, Equinix Australia

"Joining Equinix Fabric AI ecosystem means that any organisation connected to Equinix Fabric can now reach our sovereign inferencing nodes with a private, direct connection. This is how we scale Australian AI - not by moving data overseas, but by making world-class inference available where Australian data already lives."

- David Keane, CEO & Co-Founder, SCX

SambaNova SN50: Next-Generation Hardware Confirmed for National Expansion

SCX has confirmed that its national infrastructure rollout across Australian Equinix facilities will incorporate SambaNova's newly announced SN50 Reconfigurable Dataflow Unit (RDU); SambaNova's fifth-generation chip, purpose-built for agentic AI inference at scale. The SN50 delivers five times the maximum speed and more than three times the throughput of NVIDIA Blackwell B200 GPUs for agentic inference workloads, while averaging just 20 kilowatts of power per SambaRack1; enabling deployment in existing air-cooled data centres without infrastructure modification.

The SN50's tiered memory architecture supports models of up to ten trillion parameters with context lengths of up to ten million tokens, and enables millisecond hot-swapping between models - a capability that is essential for the complex, multi-model agentic workflows increasingly required by enterprise and government AI deployments. The SN50 begins shipping to customers in the second half of 2026, and SCX's confirmed supply positions Australia as one of the first markets to deploy the technology at sovereign infrastructure scale.

The supply of SN50 chips underpins SCX's confirmed expansion to additional Equinix facilities across Australia. New nodes will progressively come online throughout 2026 and 2027, building out a national network of sovereign inferencing capacity accessible through Equinix Fabric from Sydney, Melbourne, Brisbane, and beyond.

SambaNova CRO Visits Sydney to Advance Strategic Partnership

Harry Ault, Chief Revenue Officer at SambaNova, visited Sydney to advance the strategic partnership between SambaNova and SCX, meeting with customers, government stakeholders, and ecosystem partners to discuss the future of sovereign AI infrastructure in Australia and the broader Asia-Pacific region.

"The Asia-Pacific region represents one of the most significant opportunities for sovereign AI infrastructure globally, and Australia is at the forefront of that transformation. SCX has built something genuinely unique - a full-stack sovereign AI platform that combines SambaNova's world-leading inference hardware with the connectivity and reach of the Equinix ecosystem. The deployment of the SN50 across SCX's national network will be a defining moment for Australian AI, delivering agentic inference capabilities at a price and performance level that simply wasn't possible before. We are proud to be backing SCX's expansion and deeply committed to Australia's sovereign AI future."

- Harry Ault, Chief Revenue Officer, SambaNova Systems

About SCX (SouthernCross AI)

SouthernCross AI (SCX) is Australia's full-stack sovereign AI infrastructure company, delivering onshore AI inference, model hosting, and platform services powered by ASIC-accelerated compute from SambaNova Systems. SCX's infrastructure is deployed within Equinix IBX data centres and is accessible through Equinix Fabric. The company's platform supports a growing portfolio of open-weight and proprietary models - including SCX's own Australian-optimised LLMs - with no data retention, no offshore routing, and no training on customer data. SCX is headquartered in Sydney, Australia. For more information, visit scx.ai.

This press release contains forward-looking statements. Actual results may differ materially from those projected. All trademarks are the property of their respective owners.

1https://www.hpcwire.com/aiwire/2026/02/25/sambanova-eyes-10-trillion-parameter-models-for-agentic-ai-with-new-chip/

Fonte: Business Wire