High performance computing (HPC) data center is a fast-growing segment

High Performance Computing (HPC) is a field dedicated to computers aimed at performing a high volume of calculation in a short amount of time. To date, HPC users are mostly research organisms and large industrial groups, which are often concentrated in simulation-intensive industries such as the automotive, energy (including nuclear and O&G), aerospace and pharmaceuticals sectors.

Compared to regular computer, HPC have a higher electricity consumption due to more powerful processors and a higher density of servers, as well as more important cooling needs. Hence, HPC data centers are specifically designed to handle the need to host these high-density server racks.

High performance computing (HPC) data center is a fast-growing segment thanks a dynamic underlying market. Several drivers are pushing for an increased adoption of HPC systems, such as:

  • Initiatives at European and country-level (e.g. EuroHPC) fostering investment in HPC systems
  • Growing high performance computing needs (wide spreading of IA) and new use cases development (e.g., autonomous vehicle)
  • Standardization effort and lowering price of HPC equipment making it increasingly cost-effective for mainstream business including SME
  • Relatively short life cycle of HPC systems, due to the need for frequent upgrade of HPC capabilities

To date, HPC needs are mostly served through large and shared research / public HPC facilities (often being supercomputers), while on-premises private users’ facilities appear in general to be of moderate size (less than 1-2MW), with some exceptions. The number of off-premises HPC data centers is still relatively limited due to relatively strong entry barriers: CAPEX required, high power and cooling needs, and expertise needed

However, several off-premises HPC data centers are developing, especially in European Nordics countries (e.g., Norway, Sweden, Iceland, Finland) due to more favorable climate conditions and electricity cost. These off-premises HPC data centers are in general designed to handled large power capacity (up to several dozens of MW).

Moreover, several factors are fostering further outsourcing to off-premises HPC data centers:

  • Need for quicker access to HPC resources (versus being dependent on priority / queuing system of shared research HPC facilities)
  • Potential financial attractiveness for end-users versus both on-premises HPC systems or on-demand HPC resources in specific situations
  • Affordability for users that do not have the financial capability, knowledge nor strategic interest to invest in on-premises HPC server rooms
  • Possibility to benefit from up-to-date infrastructure / equipment and “best-in-class” services (e.g., security, PUE optimization to lower energy costs per rack)

Due to its growth dynamics, moderate competitiveness intensity to date and entry barriers, HPC data centers represent potentially attractive investment opportunities. However, several points need to be thoroughly assessed:

  • Asset profile and location: power availability, reliability and cost, climatic conditions, connectivity, security, and proximity to demand clusters
  • Addressable market: addressable market size and dynamics, willingness to outsource to off-premises facilities, key HPC customer segments, commercial opportunities, awareness and readiness from mainstream business clients
  • Fit with customer requirements: need for data sovereignty, preference for local players, sensitiveness to Green IT, technological requirements, need for multi-tier offering, etc.
  • Business model and offering: HPC collocation only or hybrid, dedicated ‘bare metal’ offering, on-demand Cloud offering
  • Pricing strategy: positioning of price per kW, pricing structure, relevance of multi-tier offering, discount to pre-committed client
  • Commercial strategy: go-to-market process and set-up, solidity of pre-commitments depending on project status, partnership agreements with value-added resellers, role of system integrators
  • Competition: competitive landscape and dynamics, comparison with standard colocation, Nordics HPC data centers, on-premises HPC system, on-demand Cloud HPC (AWS, Azure, etc.), other nearby assets in development