The case for building modular, repeatable data centers

Take heed to the article
5 min

This audio is auto-generated. Please tell us you probably have feedback.

Johnny Clemmons is international vice chairman and industries head at software program agency SAP. Opinions are the writer’s personal.

Synthetic intelligence is driving a large and unprecedented infrastructure enlargement to the tune of $40 billion in annual spending on data centers within the U.S. alone. 

The tempo of information middle development is accelerating the world over. There’s OpenAI’s $400 billion Stargate initiative, which can add 10 gigawatts of information middle capability with companions Oracle and SoftBank, alongside Amazon’s $100 billion global data center expansion effort. Moody’s tasks $3 trillion in global data center spending over the following 5 years.  

As hyperscalers and others make investments to satisfy AI’s rising knowledge capability urge for food, it’s vitally necessary that additionally they convey extra standardization and modularity to knowledge middle design and development. 

In the end, the architectural aesthetics with these tasks are secondary to effectivity, safety, security and repeatability. Which means that until a specific venture has distinctive environmental, siting or regulatory necessities, closely custom-made designs are pointless, even frivolous. 

For normal constructing design, the occupant expertise nearly all the time comes first. However for knowledge facilities, whereas the protection and luxury of those that work onsite ought to be a excessive precedence, so, too, is creating an optimum atmosphere for server racks and their contents. 

As a lot as stakeholders would possibly really feel inclined to deal with a knowledge middle like most different buildings, with customizations and distinctive design components to make the constructing extra interesting and the occupant expertise richer, these issues take a again seat to standardization, predictability, modularity, repeatability and sustainability. 

In truth, knowledge facilities could also be extra repeatable than most different sorts of constructing tasks. Lots of their vital parts — energy density, cooling methods, safety zones, structural hundreds and commissioning necessities — are likely to observe predictable patterns. As soon as these patterns are recognized — with the assistance of synthetic intelligence, after all — they can be utilized to tell design and development methods from the beginning.

That, in flip, creates a chance to develop a repeatable blueprint or method that knowledge middle builders can readily apply to future tasks. 

Extreme customization results in longer timelines, increased and fewer predictable prices, high quality inconsistencies and elevated threat. Modularization, then again, permits tasks to be designed, authorised, constructed and commissioned sooner, at a extra predictable and infrequently decrease price, with fewer surprises alongside the way in which.

Listed below are the particular benefits modularization brings:  

Pace: Modularization and standardization can considerably velocity up tasks with out compromising high quality. The power to develop one megawatt modules, for instance, with standardized HVAC configurations, supplies and labor necessities can allow builders to scale totally different parts after which assemble the venture as a complete. AI instruments might help design and configure the modules themselves. 

Off-site fabrication of modules, in addition to steady commissioning inside a venture, the place stakeholders can establish and deal with points as they come up, provide further time financial savings. 

Higher price management, if not price discount: Modularization brings higher venture price certainty and certain price reductions, as a result of fewer customizations and economies in procurement.

Tighter standardization for inside specs: Inside a knowledge middle, each tenth of a level in temperature and each millimeter of area counts. Miscalculations can result in costly inefficiencies and even tools failures. Repeatable, clever modeling and design attracts from previous efficiency knowledge and design specs to optimize knowledge middle inside configuration.

Alongside these strains, AI can construct and handle a reference library of greatest development practices particular to knowledge facilities. This simplifies the construct, significantly for these with much less expertise.

Predictability by way of tools and labor necessities: Repeatable designs enable stakeholders to develop commonplace necessities for the components, instruments, tools and expert labor wanted onsite, in addition to for the way lengthy. These elements can then be adjusted in response to a selected venture’s web site situations.