Surge in SSD demand means it’s time to rethink AI hardware buying

The unprecedented growth of AI and data-intensive workloads has fundamentally changed how IT infrastructure is used, triggering an explosive demand for high-performance storage. For edge data centres and IT businesses, the focus is no longer just on GPUs and CPUs. In addition to the recent surge in demand for RAM, access to sufficient storage performance is increasingly what determines whether businesses can keep pace with ever-rising data demands.

High-performance flash storage – particularly SSDs – has become a critical enabler for AI workloads. But as demand hits a new high, supply is struggling to keep up. In the second half of 2025, prices for TLC NAND flash memory – a key component of many SSDs – roughly doubled, reflecting the extreme pressure organisations are putting on flash supply chains and component availability. With no easing of this pressure in sight, prices will continue to soar and procurement timelines are set to slow even more dramatically, creating significant challenges for organisations looking to scale AI-driven operations.

Beyond immediate supply shortages, the SSD crunch exposes a wider issue in AI infrastructure procurement. Prioritising the “latest, must-have” hardware often comes at a premium, despite it delivering only marginal performance gains. In many cases, this approach leads to longer lead times, inflated costs and infrastructure that exceeds the real needs of the workloads it’s intended to support – leaving budgets that could be invested elsewhere tied up in under-utilised storage capacity.

“New” isn’t always the right decision

AI workloads are far from uniform. While some demand top-tier performance, many run perfectly smoothly on reliable, enterprise-grade storage that isn’t the latest generation. Yet, driven by the perception that buying new is the best way to maximise performance, many businesses continue to overlook a smarter alternative: refurbished technology.

In times of constrained supply, the broader market rush to secure new SSDs can reinforce the idea that premium, next-generation components are the only viable option for growing businesses. In reality, buying new often means paying more, waiting longer for stock or compromising on capacity and performance to stay within budget.

Refurbished enterprise-grade SSDs and storage arrays – thoroughly tested and backed by robust warranties – are increasingly capable of meeting modern AI storage requirements. In many cases, they are previous-generation stock or perfectly healthy equipment acquired during routine data centre refresh cycles, where high-performing equipment is replaced after only a few years of use.

By aligning component selection with workload requirements – rather than defaulting to the new-only mindset – IT teams can deploy infrastructure that is genuinely fit-for-purpose, without committing to inflated budgets or extended lead times. For many data-heavy workloads, refurbished SSDs offer the performance needed now, avoiding uncertain gains promised by next-generation equipment that may not arrive for months.

Refurbished as a scalable strategy, not a compromise

In an increasingly component-scarce market, having access to SSDs and storage arrays isn’t just about overcoming short-term market constraints; it’s about helping your business keep up and scale responsibly in the long-term.

While AI adoption is accelerating, most businesses are still operating in the pilot stage – testing use cases, refining architectures and reassessing requirements before committing to full deployments. MIT’s ‘State of AI in Business 2025’ report highlights the low hit rate of such pilots: up to 95% fail to deliver returns, forcing organisations to go back to the drawing board.

With AI workloads still volatile in terms of performance demands, data volumes and emerging regulation, locking into a lengthy hardware commitment is increasingly risky. Infrastructure needs to be agile enough to support this experimentation phase, rather than forcing businesses to make solid decisions based on assumptions that might change later.

At the same time, we’re seeing data and AI sovereignty place growing pressure on businesses to reassess how and where they store and process data – whether in the cloud or on-prem – demanding even more flexibility.

The incremental and adaptable approach to scaling that refurbished offers means businesses can expand their physical capacity step-by-step, aligning with real workload demand rather than speculative future needs. With the rollout of edge data centres on the rise and data locality becoming more critical, maintaining this level of control is vital.

Changing perspective in today’s demanding AI market

The traditional view of refurbished tech as a last-resort option is outdated. Today’s enterprise-level SSDs are sourced from high-quality environments and are data wiped and tested to rigorous standards, with robust warranties that ensure reliable performance.

The key to maximising performance within budget is recognising that not every AI workload requires the latest SSD “on the market” (or not as the case may be). It does, however, require predictability, availability and cost-effective scalability.

As businesses across the globe scramble to navigate volatility in storage markets, those who are willing to look beyond the marketing hype – and consider refurbished as a viable option for hardware buying – will be best positioned to adapt, control costs and support sustainable growth in the long-run.