By Hemant Harie, Group CTO at Data Management Professionals South Africa
Enterprises are being battered by a perfect storm in data storage. Artificial Intelligence (AI) workloads consume capacity at a pace no legacy architecture can sustain, component shortages choke global supply chains, and hardware costs and lead times have become erratic enough to derail even the best‑laid IT strategies.
AI workloads have moved far beyond the hype cycle and exploded into full‑blown operational demand, triggering an immediate and compounding strain on component availability. If we thought the supply‑chain chaos of the COVID lockdown era was severe, today’s reality is even more acute.
AI workloads consume not just storage capacity, but enormous amounts of compute, memory and power – placing unprecedented strain on the entire hardware supply chain. Storage, increasingly tied to these constrained platforms, is now feeling the knock-on effects.
As such, storage has become dramatically more expensive, procurement cycles have slowed to a crawl, and the risk profile has shifted. With AI accelerating at such a relentless pace, there is a real possibility that companies’ investments are already lagging by the time the hardware finally arrives.
Additionally, as cloud hyperscalers ramp up their AI‑as‑a‑Service (AIaaS) offerings, their data centres are now running vast AI workloads of their own, meaning they are competing directly with enterprises for the same hardware – the same chips, the same memory, the same SSDs. The result is a brutal supply‑and‑demand crunch that is pushing the market to breaking point. In some cases, we are seeing price increases of up to 400%, primarily concentrated in specific memory segments and high-performance storage components, driven by sheer scarcity and an overheated procurement landscape.
Risk of locking in resources
In this environment, buying storage means locking in scarce, increasingly expensive components in your own data centre and bearing the full risk of future price spikes, technology shifts and obsolescence. Ownership can still make sense for stable, long‑term workloads, but it ties up capex and commits organisations to rigid, multi‑year refresh cycles at a time when the market is anything but predictable.
Shifting from capex‑heavy storage purchases to opex‑based rental models reshapes both risk and budgeting flexibility. Instead of tying up capital in hardware with uncertain delivery dates and volatile pricing, an opex approach lets enterprises scale quickly and avoid supply‑chain shocks.
With AI driving sudden demand spikes and some components rising by triple‑digit percentages, renting capacity removes much of that exposure. Organisations do not have to gamble on hardware that may arrive late, cost far more than expected, or be outdated on delivery.
However, from a data‑management services provider’s perspective, priority should be given to stable, predictable workloads. Regulation alone demands it, as many organisations still require isolated, on‑premise storage to meet compliance and sovereignty obligations. And for latency‑sensitive systems, local storage is not optional but essential. These workloads need to sit close to the point of use and remain insulated from the wider infrastructure to deliver the speed and reliability they depend on.
Weighing up ownership vs rental
As a result, CIOs must weigh governance, security and performance carefully when deciding how much storage to own versus rent. Regulation is no longer a formality; every technology choice must align with the full set of compliance obligations.
In South Africa, the Protection of Personal Information Act (POPIA) remains a moving target, and the rise of high‑speed AI processing only heightens scrutiny. CIOs need clarity on where data may reside, who can access it, how it is protected, and how quickly it must be retrieved.
Latency‑sensitive, sovereignty‑bound or isolation‑critical workloads typically demand on‑premise storage, while more elastic or lower‑risk workloads can sit in the cloud. Ultimately, the decision is less about technology and more about governance and risk posture.
Multi‑cloud and hybrid storage architectures help shield businesses from price volatility and supply‑chain disruption by spreading risk across environments. Some workloads, due to regulation, sovereignty or strict latency requirements, must remain on‑premise, but that does not preclude the cloud. A hybrid model lets organisations keep sensitive or performance‑critical data close while using the cloud for elasticity and rapid scale.
Achieving greater agility, continuity and insulation
Because no single provider can meet every need, multi‑cloud adds resilience – if one platform faces shortages, price hikes or regional constraints, workloads can shift to another. The result is greater agility, continuity and insulation from market shocks.
The choices CIOs make now will determine how well their organisations withstand the volatility ahead. AI is outpacing traditional procurement, hardware scarcity is rewriting the economics of ownership, and regulatory pressure is only tightening. No single model is enough. The businesses that stay resilient will treat storage as a strategic mix, combining owned and rented capacity, using hybrid and multi‑cloud architectures, and aligning every decision with governance, performance and risk.
