How Server Consolidation Can Create More Room for Enterprise AI

For IT teams managing data centre infrastructure, server consolidation offers a practical way to make existing resources work harder. The aim is to move workloads from a large number of ageing, less efficient machines onto fewer, more capable servers while maintaining performance and keeping critical applications running smoothly.

The benefits can extend across the entire server environment. Consolidation can release rack space and power capacity, reduce software costs under per-core licensing models and give organisations room to expand. That additional capacity can support growing business applications, analytics or AI infrastructure.

Matching Server Capabilities to Workload Needs
Successful consolidation requires more than reducing the number of machines. The replacement systems must handle existing workloads efficiently without introducing unnecessary migration complexity.

AMD EPYC™ processors combine high core density, strong performance per core and energy efficiency using the x86 architecture already deployed by enterprises. This allows IT teams to pursue consolidation without the additional application validation and migration considerations that can accompany a move to Arm®-based servers.

According to AMD, its 5th Gen EPYC processors power the most energy-efficient x86 servers available. The range includes several configurations to address different requirements.

• High-density processors with up to 192 cores to accommodate more workloads.
• High-frequency options reaching up to 5GHz to maximise performance per core and help reduce licensing costs.
• Balanced configurations for applications that need a combination of core density and per-core performance.

AMD also expects its forthcoming 6th Gen EPYC “Venice” processors to extend that density advantage, with configurations offering up to 256 cores.

These options allow organisations to select compute resources according to workload requirements while consolidating their infrastructure. The capacity released can also make room for AMD Instinct™ GPUs for AI inference and training, alongside AMD Pensando™ networking for predictable data movement.


What Consolidation Could Deliver
AMD illustrates the potential with a comparison involving 100 dual-processor Intel® Xeon® Platinum 8280 servers that are five to seven years old.

According to the company’s figures, 14 servers based on AMD EPYC 9965 processors can deliver the same total performance. This represents an 86% reduction in server count, with potential reductions of 69% in power consumption and 41% in three-year total cost of ownership.*

The core counts in this comparison are also lower. The original configuration comprises 100 servers with 28 cores per processor and two sockets, totalling 5,600 cores. The replacement configuration uses 14 servers with 192 cores per processor and two sockets, totalling 5,376 cores.

Reducing the physical server footprint can make rack space and power available for additional workloads within an existing data centre, without requiring more physical infrastructure.

Software licensing is another consideration. Many virtualisation and database products are licensed by socket or core. For these products, completing more work with fewer systems and cores can potentially save millions in licensing and maintenance fees.

The extent of those savings depends on the licensing model. Software charged by user, device or another measure does not benefit in the same way.

AMD reference supplied for this comparison: 9xx5TCO-001B.

How DNSE Securities Applied the Approach
Vietnamese online stock brokerage DNSE Securities needed the density and performance associated with cloud infrastructure while retaining its systems on premises to meet regulatory requirements. Its existing servers could not provide the level of consolidation needed to support further growth.

After benchmarking systems from AMD and Intel, DNSE deployed 3rd and 4th Gen AMD EPYC processors across its data centre. The company halved its rack count and brought power consumption down from 16–17 kilowatts per rack to 7–9 kilowatts per rack.

AMD EPYC processors now support 90% of DNSE’s workloads. The brokerage uses high-frequency processors for latency-sensitive trading systems and high-core-count processors for mixed containerised workloads. Matching processors to these different requirements enables the company to scale without overprovisioning.

DNSE chief technology officer Binh Nguyen said AMD EPYC-powered servers would allow the company to continue building its infrastructure and running its software in the same way, but at a much larger scale.

Making Capacity Available for AI
The immediate case for consolidation centres on measurable improvements in server count, rack usage, power consumption and software licensing costs. There is also a longer-term benefit. Running existing business applications more efficiently can free the power, space and cooling capacity needed for new infrastructure.

This is relevant to agentic AI, where each agent action can generate database queries, API calls, token generation and tool execution. Consolidating established workloads onto fewer, more efficient servers can help accommodate that additional demand alongside essential business applications.



Enterprise AI is also increasingly distributed across frontier models, cloud services, on-premises systems and AI-enabled client devices. Decisions about where a workload should run, and which model it should use, can depend on its complexity, data sensitivity, latency requirements and cost.

For organisations introducing AI, consolidation may create room for accelerators within existing data centre limits. It can also support gradual expansion, allowing IT teams to match processors, accelerators and models to suitable workloads without rebuilding their infrastructure around a single model or application.

A Measurable Priority for CIOs
Server consolidation may attract less attention than other technology initiatives, but its outcomes can be assessed through clear operational and financial measures. These include the number of servers in use, occupied rack space, power consumption, licensing expenditure and total cost of ownership.



AMD EPYC processors offer IT teams a way to pursue these improvements while maintaining performance and choosing configurations suited to different workloads. A more efficient server environment can meet current business needs while leaving capacity available for future growth.