Flexible data storage for a rapidly evolving AI world

• AI workloads are changing how data needs to live and move, turning cold archives into mission-critical infrastructure and exposing the limits of static storage tiering
  
  
• Supply chain volatility, shifting sovereignty requirements and unpredictable demand patterns are making traditional three-to-five-year storage refresh cycles increasingly difficult to plan around
  
  
• Cloud-adjacent dedicated storage offers a way to maintain sovereign control and governance while consuming cloud services, without locking into fixed capacity bets or absorbing punitive egress costs

Enterprise storage decisions used to be relatively straightforward: size for capacity, buy for performance, refresh every few years. But as AI workloads reshape data access patterns, supply chains grow less predictable and sovereignty requirements tighten across Australia and New Zealand, that approach is under pressure from multiple directions at once. 

We spoke to Mike Walls, Director of Cloud at Datacom, about why traditional storage planning is breaking down, how AI is changing the relationship between data tiers, and what a more flexible architecture looks like in practice.

In an unpredictable world, the only sane response is flexibility – and that’s exactly what we’ve tried to engineer into Datacom’s Dedicated Storage as a Service offerings.

I’ve seen a lot of rapid changes over the last 15 years of working on cloud infrastructure and transformation projects, but the complexity organisations are grappling with in the AI era is unprecedented.

For most of my career, storage decisions have boiled down to two variables: capacity and performance. You sized for three to five years, bought the hardware, deployed it, and hoped your forecasts weren’t too far off, but that playbook doesn’t work anymore.

There have been three fundamental changes that have changed the shape of an organisation’s storage needs. 

First, customer demand is now digital by default, which means every new product or channel is essentially a data problem.

Second, AI workloads stress storage differently, prioritising high-throughput access to large, often unstructured datasets rather than the input/output operations per second (IOPS) patterns traditional enterprise systems were designed for.

And third, global supply chains and component pricing have become so volatile that a quote you receive this week may not be honoured next week – and in some cases the vendor reserves the right to change the price until the gear is actually on the boat.

Add to that a geopolitical backdrop where cyber risk is rising, and governments are sharpening their stance on data residency and sovereignty, and you’ve got a perfect storm for anyone trying to lock in a storage architecture for the next five years.

Most organisations I work with across New Zealand and Australia are still relatively early in their AI journey, experimenting at the edges or tapping public AI services for small-scale use cases. But among customers who have moved from experimenting with AI to production at scale, we are seeing the implications from a data storage perspective. Essentially, AI changes the way data has to live and move. 

Datacom's 2026 data sovereignty research found 43% of New Zealand organisations said legacy systems, fragmented data platforms or unclear governance were slowing their ability to adopt AI safely. It’s a challenge that extends directly into how storage is architected.

Training, fine-tuning and retrieval-augmented generation (RAG) models need parallel, high-throughput access to large volumes of organisational data – much of which has historically sat in “cold” or archival tiers because it was rarely touched.

Those cold tiers are optimised for cheap retention, not fast inferencing, which means the data you thought you could file-and-forget suddenly needs to be promoted into performance tiers, potentially at scale and on short notice.

As AI moves closer to the edge – into robotics, manufacturing sites, and remote operations – data gravity (where massive datasets attract applications, services, and additional data and movement) becomes a critical architectural concern. Which datasets live where, how they traverse networks, and what latency profile your business can tolerate are now pressing issues.

If your storage design assumes predictable workloads and static tiering, AI will expose the weak points in that architecture. You either end up over‑provisioning expensive performance storage just in case you need it, or you constrain AI initiatives because the data pipelines simply can’t keep up.

Overlaying all of this is a growing focus on data sovereignty – particularly in regulated industries and government. Many organisations are still immature in data governance and classification. They don’t have a clear handle on what data sits where, how sensitive it is, or who should access it. As a result, keeping data resident and sovereign often becomes the default safety blanket.

Datacom's sovereignty research found that while 83% of New Zealand senior leaders are concerned about offshore data risks, only 35% have a formal policy governing where critical data should reside and how those decisions are made.

Public cloud continues to deliver enormous value, but it also introduces two challenges:

Cost: data egress and storage patterns can drive bills that are hard to predict and even harder to explain to a CFO once AI-scale datasets start flowing.

Control: jurisdictional exposure, including things like the US Cloud Act, is now a board-level conversation in many sectors.

This is where cloud-adjacent storage becomes powerful. By placing a dedicated storage platform you control alongside public cloud – rather than inside it – you can:

• Keep your data sovereign and within your compliance envelope while still consuming cloud services.

• Avoid punitive egress charges by moving data through your own storage environment rather than in and out of a hyperscaler system.

• Maintain a consistent operating and governance model across edge, private cloud and public cloud environments.

Dedicated Storage as a Service (DSTaaS) is designed to make that adjacency practical. The concept is at least 15 years old but has evolved to allow organisations to enjoy the agility and pay-as-you-go billing of public cloud, but with the high performance, security, and physical control of dedicated, hosted or co-located hardware.

If unpredictability is the defining characteristic of today’s infrastructure landscape, then the antidote has to be flexibility – in technology, in placement, in consumption, and in price. 

That’s the core design principle behind Dedicated Storage as a Service we offer in partnership with Dell Technologies, and which is tailored to the needs of customers across Australasia.

Here’s what that looks like in practice from my perspective.

Flexible consumption, not fixed bets: Instead of guessing your peak requirement five years out and sinking capital into hardware that may sit idle, you deploy dedicated storage capacity that you can scale and consume as needed, on a service subscription basis. That turns uncertain forecasts into a defensible spend profile aligned with real demand.

Tiering for AI and traditional workloads: DSTaaS supports multiple performance, capacity and archive tiers and is built to move data between them as use cases evolve. When an AI project suddenly needs high-throughput access to long-retained records, you can promote that data to a performance tier without redesigning your estate from scratch.

Placement across edge, private and cloud‑adjacent: You can deploy DSTaaS in your own data centre, in a hosted facility, at the edge, or adjacent to hyperscaler regions, and manage it through a single ecosystem. That hybridity gives you the freedom to co-locate storage near compute without locking yourself into one provider or one site.

Price stability in a volatile market: Behind the scenes, we absorb much of the supply-chain turbulence – extended lead times, component hoarding by large AI players, and week-to-week price swings. For you, that translates into a predictable, per-gigabyte consumption-based commercial model instead of line items that jump 200–400% between budget cycles.

From conversations with CIOs, CTOs and CFOs, the greatest appeal of the DSTaaS model is often psychological. They no longer have to bet the business on a single capacity forecast in a world where demand, regulation and technology are all moving targets.

When I sit down with organisations facing a major storage renewal or planning for AI-heavy workloads, I encourage them to pressure-test their architecture decisions with a few straightforward questions:
 

1. Does this platform support both AI and traditional workloads, with the differing performance, throughput and data patterns each requires? If deploying AI isn’t a key architectural consideration now, it will be soon.

2. Can the platform handle multiple use cases and performance tiers without locking you into assumptions that might be wrong in two years’ time?

3. Does it give you true hybridity – from edge to private to public cloud, so you can place storage where the compute and data actually are?

4. Can it meet your regulatory, compliance and sovereignty requirements, including the ability to maintain sovereign copies where needed?

5. Does it deliver the levels of availability and recoverability your business now needs in a world where ransomware attacks continue to cripple organisations and “assume breach” is the only realistic stance?

6. Finally, does the commercial model give you reasonable price certainty over time, without forcing you into large upfront capital bets based on inherently uncertain forecasts?
 

Dedicated Storage as a Service isn’t a silver bullet, and it’s not the only way to answer those questions. 

But in an AI-driven, sovereignty-conscious and highly volatile hardware market, it’s one of the few models I’ve seen that combines technical flexibility, operational control and commercial predictability in a way that actually matches the world we’re living in.

Find out more about Datacom Dedicated Storage as a Service offered in partnership with Dell Technologies and complete our ten-question self-assessment to receive personalised insights on how to best manage your data storage needs.