Executive Summary

• Sam Walker, Vice President of Sales, EMEA and India of ProLabs explores the notion of NEM ‘alternatives’ and unlocking capacity within existing racks.
• With the DRAM supply exhausted and a 300% price surge on DDR%, operators are solving the problem with Network Equipment (NEM) alternative optics: standardised fibre optic modules manufactured by independent companies to replace expensive, brand-name offerings from the dominant market player.
• Global data centres are looking to consolidate existing infrastructure and scale new technologies, optics such as NEM-alternatives will be crucial in ensuring capacity is maximised, standing out as a cost-efficient way to unlock the full capability of infrastructure.

Across the data centre space, there is a growing misconception that the next wave of growth can only come from new builds. Capacity has become a significant issue for operators across the globe, with major data centre hubs across the Frankfurt, London, Amsterdam, Paris, and Dublin (FLAP-D) region pushed to their limits. Driven primarily by the surge in Artificial Intelligence (AI) workloads and cloud adoption, colocation vacancy rates across these hubs have now plummeted to roughly 6.3%, leaving limited room for expansion without substantial new developments.

With capacity increasingly constrained in these established markets, attention is already turning to emerging cities such as Madrid, Milan, and Warsaw to absorb this overspill. However, despite the increased investment and development of these secondary markets expected over the coming years, they are unlikely to close the gap in the short term. This leaves data centres operators facing a critical decision – either they continue to prioritise new builds, or look for ways to unlock capacity within their existing racks.

The capacity concerns

Since operators are already facing acute constraints thanks to the surge in AI workloads, optimising the available capacity within facilities has become just as important as adding new space. However, the challenge extends beyond the implementation of additional racks as AI workloads have become increasingly compute-intensive, placing greater demands across the entire infrastructure stack.

Consequently, they are establishing significantly higher requirements for both networking and power, creating new bottlenecks that can limit capacity even if physical space remains available. For networking, this includes constraints around bandwidth, latency, and switch capacity, while on the power side, limitations in power delivery and utility availability are becoming increasingly common.

This means that operators are increasingly focused on how efficiently their infrastructure resources are utilised. Given the networking constraints, it’s no surprise that the high-performance equipment capable of supporting these intensive compute environments is in demand. Yet attaining them – especially given the structurally elevated price of Dynamic Random Access Memory (DRAM) at present – is a challenge.

Navigating inflated DRAM costs

DRAM underpins AI system operations, providing the ultra-fast memory access required to feed data-hungry models. This utilisation of High Bandwidth Memory (HBM) pushes this even further: by stacking DRAM vertically and widening the memory bus, it forms the high-throughput memory layer within modern AI accelerators, drastically improving the efficiency of applications.

Unfortunately – in their attempts to ensure uninterrupted inference performance and eliminate system bottlenecks – hyperscalers and server Original Equipment Manufacturers (OEMs) have presently exhausted the available DRAM supply. The market is very scarce of essential HBM: between October and December 2025 alone, DDR5 prices surged by over 300%, evolving a procurement issue into a significant structural constraint.

Without affordable DRAM, operators cannot scale the right infrastructure, adding another layer of complexity to the equation. Taken together, these constraints reinforce the reality that AI capacity is determined by the availability of the underlying enabling components that support it. Even where data centre space, power, and networking infrastructure are in place, shortages in DRAM and other high-performance memory can prevent operators from fully populating and scaling compute clusters. For operators, the search for networking components that solve both the capacity and DRAM dilemma is on.

The NEM ‘alternative’ solution

Thankfully, many data centre operators have found their solution to the capacity and DRAM complexity in the form of Network Equipment (NEM) alternative optics. These solutions are standardised fibre optic modules manufactured by independent companies to replace expensive, brand-name offerings from the dominant market players. Interestingly, while they perform to the same level of NEM-branded units from companies like NVIDIA, Cisco, and Juniper Networks, these optics typically can be procured at 70% of the cost.

Because these products are engineered to work seamlessly across multi-vendor environments, operators are empowered to expand their available capacity without being locked into a single hardware roadmap or procurement cycle. This freedom to mix platforms and deploy new links where they’re needed also boosts the availability of bandwidth, relieving the increasing pressure on existing fabrics.

In turn, this is aided by the networking speeds offered by the extensive portfolio of solutions from NEM-alternative providers. ProLabs is just one company that has evolved these AI-ready transceivers to accommodate networking speeds from 400G to 1.6Tb, bringing the additional bandwidth density operators need to overcome today’s capacity-constrained AI fabrics. To overcome the issues of tomorrow, these NEM alternative vendors are also evolving their portfolios to deliver 3.2Tb connectivity in the future too.

The path ahead

Crucially, this approach enables organisations to scale network capacity more efficiently. By removing dependency on a single vendor ecosystem, operators can respond more quickly to changing workload demands while maintaining control over costs. Backed by a lifetime warranty and compliance with major trade standards, this frees operator budget to spend on HBM – crucial in such a time of market volatility.

With global data centres not only looking to consolidate existing infrastructure but scale new technologies, optics such as those offered by NEM-alternatives will be crucial in ensuring capacity is maximised, standing out as a cost-efficient way to unlock the full capability of infrastructure.