Australia's AI infrastructure boom is accelerating rapidly, with billions flowing into data centres and compute capacity scaling to meet demand. But behind the headlines, a more fundamental question is emerging: how will storage infrastructure keep pace with AI at scale?
According to Stefan Mandl, Vice President, Worldwide Sales and Marketing at WD, the answer lies in rethinking and reinventing the role of hard disk drives (HDDs).
"Data is exploding, and it's driving growth across every form of storage with it - from HDDs to SSDs, and even tape," said Mandl.
"In this rising tide, every boat is being lifted, and the momentum shows no signs of slowing," Mandl said. "As AI adoption grows, the bigger question is not just about how much compute is available, but how the data that AI systems continuously create, store and reuse will be managed at scale. Compute cycles, but data compounds."
While other storage technologies continue to evolve, Mandl believes that HDDs are poised to maintain their essential role well in the future.
"HDDs will continue to be the backbone of the data-driven AI economy. According to IDC, by 2029, nearly 80% of cloud storage will still be based on hard drives." (IDC Source: Worldwide Global StorageSphere Forecast, 2025-2029, Doc #US53561425).
That is driven by how AI workloads behave in practice.
"Every AI workload generates data that must be stored persistently, and that data compounds over time. Storage demand is driven by the accumulation, retention, and reuse of data – making it structural and durable, not cyclical. This compounding effect creates a tiered storage challenge that no single media type can solve alone," Mandl explained.
"From training datasets and model checkpoints to inference logs and telemetry, AI pipelines generate hot, warm and cold data. Flash alone cannot manage this economically nor sustainably. HDDs are indispensable to enabling AI at scale," Mandl explained.
Predictable economics and scalability to support AI boom
Australia's rapid data centre expansion is sharpening focus on cost and scalability. According to the Commonwealth Bank of Australia's Economic & Markets research data, the country is now the world's third-largest AI investment destination, with a pipeline nearing 6GW - or around AU$150 billion - and capacity potentially tripling by 2030.
"Our customers don't just buy drives," Mandl said. "They buy time-to-production capacity, reliability at scale and predictable economics."
HDD continues to play a key role here.
"For hyperscale cloud and enterprise businesses, total cost of ownership (TCO), predictability and scalability to meet data storage demands are critical. Capacity enterprise or nearline HDDs continue to hold an advantage. We're seeing a 6-10x premium for flash compared to HDDs, and in large-scale environments, the economics compound."
To meet these demands, WD is advancing a customer-centric storage roadmap across capacity, performance, power efficiency and reliability with the right storage economics and no disruption to business.
"At WD's Innovation Day this year, we announced critical innovations that reinvent the hard drive to meet the needs of AI," Mandl said. "This allows us to deliver more value per terabyte for our customers and meet increasing storage demands in the AI-driven data economy."
He stressed that customer-centricity is central to that approach. "Our customers' workloads are varied, and each requires a different mix of capacity, performance and power."
Choice and flexibility for transitions without disruption
According to Mandl, WD is progressing both Energy-Assisted Perpendicular Magnetic Recording (ePMR) and Heat-Assisted Magnetic Recording (HAMR) technologies in parallel - a strategy designed to support smoother transitions.
"Our world's new highest capacity 40TB UltraSMR ePMR HDD - is currently in customer qualification," Mandl said. "At the same time, HAMR capacities are scaling to 100TB and beyond with continued qualification momentum with hyperscale customers."
Importantly, these technologies are aligned. "ePMR and HAMR are built on a common architecture, enabling a smoother product transition with no infrastructure disruption. By advancing both in parallel, we help ensure our customers' roadmaps - not vendor constraints - define the pace of innovation," he said.
The technologies that WD used to bridge to HAMR, such as OptiNAND®, UltraSMR, triple stage actuator (TSA) and 11-disk platform support continued capacity gains.
HDDs reimagined for speed
WD is also addressing performance demands traditionally associated with flash. "We've introduced two industry-first innovations - High Bandwidth Drive and Dual Pivot technologies - that fundamentally reset HDD performance," Mandl explained.
High Bandwidth Drive Technology enables simultaneous read/write operations from multiple heads on multiple tracks, which delivers up to 2x bandwidth of conventional HDDs without power penalties, with a path to scale to 8x gains, and is already in customer hands for validation.
"Dual Pivot Technology enhances throughput further as it adds independently operating actuators on a separate pivot, delivering up to 2x sequential I/O gains within a 3.5-inch drive," he said.
"When combined, these technologies will enable WD to increase sequential I/O to 4x overall, delivering 100TB HDDs while maintaining the relative I/O per TB rate customers enjoy today. This reduces the need for customers to increase SSD deployment or rearchitect services as capacity scales."
Power-optimised HDDs at play
Meanwhile, energy is becoming a critical constraint as AI workloads grow.
Mandl explained that AI training and inference are generating massive volumes of cold data that still need to be accessed in seconds, not hours.
"That makes this data too active for tape, but often too cost-prohibitive for traditional capacity drives."
WD's response is power-optimised HDDs. These drives use 20% less power while maintaining sub-second access storage tier with the same 3.5-inch form factor. They trade minimal random I/O for higher capacity and significantly lower power.
He said this enables customers to reduce TCO, build lower-cost storage tiers and improve sustainability for AI-scale data.
Improving time-to-value by reducing complexity
Beyond hardware, WD is expanding into software to simplify deployment.
"Mid-scale customers face hyperscale challenges without hyperscale resources or economics, and that's why we're expanding our Platforms business to help reach a broader set of customers," Mandl said.
An intelligent software layer, built on an open API and expected in 2027, will help address this.
"It will enable organisations at 200+ petabyte scale to achieve hyperscale-level efficiency and economics, while reducing time-to-production and lowering qualification risk across storage tiers," he said.
HDDs reinvented, not replaced
Ultimately, staying relevant in an AI-driven world will come down to continuous innovation and the ability to anticipate what customers need next, Mandl says.
He emphasised that WD is a strategic storage infrastructure partner for the AI-driven data economy, adding that the company is driving innovation across storage architectures, platforms, and infrastructure solutions.
"As AI adoption accelerates, our customers require storage solutions that scale predictably and sustainably. We are delivering flexible, high-capacity storage solutions that can scale with dynamic AI workloads, empowering customers to build solutions that are seamlessly integrated, cost-efficient at scale, and future-ready," Mandl says.