Cloud, AI and the cost of storage

Predictions are difficult, especially when they concern the future. That has never been truer than when planning infrastructure budgets in the age of AI. Storage prices are rising rapidly across DRAM, NAND flash, and even entire storage systems, and a global helium shortage, critical for semiconductor manufacturing, is adding further strain. The result is a perfect storm: soaring costs, constrained supply, and growing urgency for organisations to rethink how and where they store data. 

 

Li Peiying, CEO of DRAM giant Nanya Technology, stated in early March that the imbalance between supply and demand for storage has not improved. Prices not only continued to rise monthly this quarter but also experienced a sharp increase in the second quarter. Prices are expected to continue rising until the end of the year, especially now with the added strain of the helium shortage. A gradual improvement in the supply and demand imbalance is not expected until the second half of 2028 at the earliest. 

 

What’s emerging in response is a decisive shift toward multi-cloud. As storage becomes more expensive and harder to secure, companies are distributing workloads across multiple cloud providers to optimise costs, improve performance, and reduce reliance on any single vendor. But while multi-cloud offers flexibility, it also introduces new layers of complexity that organisations must learn to manage.

 

 

AI is driving unprecedented storage demand

The AI boom is generating mountains of data. From training large-scale models to running real-time inference workloads, organisations are producing and consuming data at a pace that traditional infrastructure simply cannot handle. This explosion is driving the rapid expansion of data centres, each requiring massive amounts of storage, processing power, and memory.

 

Prices for these components have risen dramatically, and there is little relief in sight. Industry leaders have warned that the imbalance between supply and demand for memory and storage will persist for years. NAND flash prices, for example, have surged sharply, with some manufacturers increasing prices by as much as 50% due to overwhelming demand from AI infrastructure expansion.

 

Adding to the pressure is a global helium shortage. Helium is essential in semiconductor manufacturing, particularly in the cooling and production processes for advanced chips. Supply constraints are now rippling through the entire hardware ecosystem, further increasing the cost and scarcity of storage components.

 

For enterprises, this creates a highly competitive environment. Businesses are now competing not only with each other but also with hyperscalers and AI giants for access to critical infrastructure resources. Storage, once a predictable line item, has become a volatile and strategic concern.

 

 

Multi-cloud as a strategic response

For IT architects, the cloud is becoming increasingly relevant, not just as an alternative, but as a necessity. Most AI models are born in the cloud and remain cloud bound. At the same time, organisations are adopting a best-of-breed approach, selecting the most suitable AI services and tools from different providers.

 

This naturally leads to multi-cloud strategies. Data and workloads are distributed across platforms to access specialised capabilities, optimise performance, and control costs. It also provides a hedge against rising storage prices, allowing organisations to shift workloads dynamically based on pricing and availability.

 

Resilience is another key driver. Outages among major cloud providers in recent years have prompted companies to reassess their reliance on single platforms. By diversifying across multiple clouds, organisations can reduce risk and improve business continuity.

 

Regulatory and sovereignty concerns are also accelerating this trend. European customers, in particular, want to store at least some of their data in a sovereign cloud to maintain control over data location, environment operation, and security, plus encryption key management. This will also drive further adoption of cloud infrastructure in multi-cloud environments. Eurostat reports that in 2025, approximately 52.74% of EU companies will use paid cloud services, and this figure is expected to rise. According to a report by Flexera, 89% of surveyed companies are already using multi-cloud strategies. This number is likely to increase further, as companies want to leverage the strengths of each cloud provider, from computing power to AI and analytics tools, to optimise performance and costs. 

 

 

Cloud proliferation creates new challenges

However, the rise of multi-cloud is not without its complications. As organisations expand across multiple platforms, managing these environments becomes increasingly difficult. Each cloud provider has its own tools, interfaces, and policy frameworks, leading to fragmentation.

 

Many organisations attempt to manage this complexity using multiple point solutions. The result is often a patchwork of tools that lack integration and visibility. Each tool sees only a portion of the environment, making it difficult to answer fundamental questions: Where is our data? Who can access it? Are we compliant?

 

This fragmentation is particularly problematic for AI workloads. Beyond traditional data, organisations must also manage and secure AI metadata—models, training datasets, pipelines, and outputs. These assets are critical to business operations and require the same level of protection and governance.

 

Without a unified approach, gaps in security and compliance are inevitable. Inconsistent policies across clouds can lead to misconfigurations, increased risk exposure, and operational inefficiencies.

 

 

The need for centralised control

To successfully manage multi-cloud environments, organisations need a centralised control layer that spans all platforms, whether data resides in the cloud or on-premises. A unified solution provides a single view of the entire data landscape, enabling consistent policy enforcement, improved visibility, and streamlined operations.

 

This centralised approach is essential for securing both traditional data and AI workloads. It allows organisations to protect sensitive information, monitor access, and ensure compliance across all environments. Advanced capabilities such as immutable storage, anomaly detection, threat scanning, and continuous backup analysis help identify and mitigate risks early.

 

Equally important is cost transparency. As storage prices continue to rise, organisations need clear insights into how resources are being used and where costs can be optimised. Understanding data classification, what data is critical, what can be archived, and what can be deleted, plays a key role in both cost control and regulatory compliance.

 

Ultimately, managing multi-cloud environments effectively requires more than just distributing workloads. It demands a cohesive strategy supported by powerful tools that provide visibility, control, and security from a single console.

 

 

Embracing multi-cloud in an AI age

The convergence of AI-driven data growth, rising storage costs, and supply chain constraints, including helium shortages, is reshaping the IT landscape.  

 

Organisations that embrace multi-cloud can gain flexibility, resilience, and cost advantages. But to fully realise these benefits, they must address the challenges of complexity and fragmentation with unified, centralised control.

 

---

 

Mark Molyneux is Field CTO at Commvault

 

Main image courtesy of iStockPhoto.com and tadamichi