Strategy Blog
What Our DDN Partnership Means for You — Faster, More Efficient AI Inference
Our new partnership with DDN removes the single biggest bottleneck to AI performance — the data layer. Here's what that means for your workloads.
We have been building towards this for a while, and today we are excited to share the result: SCX is partnering with DDN, a global AI data intelligence leader, to bring their Infinia data platform into our sovereign AI inference cloud.
This is not just a technology integration — it is the piece that makes everything else we have built work better. And we cannot wait for you to try it.
The bottleneck no one talks about
When people talk about AI performance, they focus on the model and the chip. That makes sense — those are the headline numbers. But there is a hidden bottleneck that quietly kills performance in production: the data layer.
Every AI inference request has to move data — model weights, context, KV cache — in and out of memory. If that data pipeline is slow, your expensive AI hardware sits idle, waiting. No matter how fast your chip, a stalled data layer means wasted compute, higher latency, and slower response times for your users.
DDN Infinia is purpose-built to solve exactly this problem.
What DDN Infinia delivers
DDN shared some numbers ahead of our announcement that are worth sitting with:
- Sub-millisecond latency — data reaches the inference engine faster than you can blink
- Up to 27x faster KV cache loading — the core data operation that powers every AI request
- Extreme concurrency support — more simultaneous users without degradation
For our customers, this means your AI workloads run faster, handle more users, and make better use of the compute you are paying for.
Why this matters for your bottom line
We built SCX on a simple premise: that sovereign AI infrastructure could be both better and cheaper than the offshore hyperscaler alternative. The efficiency story is well established — our SambaNova-powered nodes deliver approximately 2.5x to 5.6x the performance per watt of GPU-based systems.
But efficiency is not only about power. It is about getting more value from every dollar you spend on AI.
When the data layer runs faster, you need less hardware to serve the same volume of requests. You pay for fewer tokens to achieve the same throughput. Your per-user costs come down. And because DDN Infinia eliminates I/O stalls, your hardware utilisation climbs — the expensive compute you have deployed actually does useful work instead of waiting for data.
Built for agentic AI and large context windows
If you are building agentic AI workflows — multi-step reasoning, tool calling, chain-of-thought — you need fast access to large amounts of context data. That is where the 27x KV cache improvement really shines.
Large context windows are one of the most powerful capabilities in modern AI. They are also demanding: pushing and pulling megabytes of data per request. DDN Infinia removes the friction that would otherwise make large-context applications sluggish or expensive.
For customers running Project MAGPiE — our sovereign LLM fine-tuned for Australian context — this means you can build rich, context-aware applications without compromising on speed or sending your data offshore.
What this means for your AI roadmap
Node 2 of our sovereign AI cloud, powered by DDN Infinia, is targeted to be operational by the end of 2026. That means Australian enterprises, government agencies, and research institutions will have access to an inference platform that is:
- Faster — sub-millisecond latency for responsive AI applications
- More efficient — 27x faster data access means better hardware utilisation
- Fully sovereign — all data, workloads, and infrastructure remain in Australia
- Ready for scale — designed for highconcurrency enterprise and agentic AI workloads
We are excited for you to try it
This partnership took a lot of work to put together, and we genuinely believe it changes what is possible for Australian organisations running AI at scale.
If you are already a customer, you will start seeing the benefits as we roll out Node 2. Same data residency, same sovereignty guarantees, better performance.
If you are evaluating SCX, we would love to walk you through what this means for your specific use case. Reply to this post or book a call — our team is ready to show you what sovereign AI performance looks like when the data layer is no longer the bottleneck.
Want to see the difference in your own workloads? Get in touch at [email protected] or book a technical session with our team.