Keep cool and carry on: Mitigating the risks of retrofitting liquid cooling technologies
Insight shared by:
Article by
Immersion cooling may be touted as the key to unlocking cost and energy savings for data centres and hyperscalers, but its novelty also carries a high level of risk. We explore what these risks are, and why managing them from the outset is vital to avoiding delays, disruption, and outages.
The modern data centre is a hungry beast. Driven by power-intensive workloads such as artificial intelligence (AI) and cryptocurrency mining, data centre electricity usage is rising year-on-year. A report from the International Energy Agency (IEA) even suggests that data centres could consume more than 1,000TWh by 2026 in a worst-case scenario.
Cooling and ventilating the acres of servers, data storage drives, and network equipment needed to keep the world running contributes significantly to this. These systems account for 30-55% of a data centre’s power consumption, a figure that will inevitably increase as systems work harder and produce more heat.
Unfortunately, traditional air-cooled systems are not equipped for this. As such, more efficient and effective alternatives are needed to prevent powerful technologies and systems from overheating.
The good news is that such an alternative already exists in the form of liquid cooling methods such as immersion cooling. The bad news? These systems depart significantly from the ways in which existing data centres are planned, built, and structured.
What is immersion cooling?
Liquid immersion cooling involves submerging IT components in an inert, non-conductive fluid – also known as a ‘dielectric fluid’. This absorbs and dissipates the heat generated by components more than 20 times more efficiently than air.
Capable of accommodating entire racks, liquid immersion cooling systems transfer the heated fluid away from the hardware to a cooling system that extracts the heat and recirculates the cooled liquid into the immersion enclosure.
Unlike those equipped with traditional air-cooled systems, data centres that use immersion cooling can easily exceed 100kW. As such, the market has significant potential, with a predicted worth of more than $1.4bn by 2029.
Why is immersion cooling difficult to implement?
Despite its benefits, installing immersion cooling systems presents significant barriers, particularly for data centres that are not purpose-built to accommodate them. Implementing immersion cooling is not a case of swapping out fans and ventilation systems for immersion enclosures. It represents a wholesale change in a data centre’s structure and functionality.
The weight of cooling baths in which equipment is submerged, for example, requires greater structural floor loading than traditional air-cooling systems. Data centres must also move away from designs that keep water and other liquids out at all costs, to ones in which plumbing infrastructure and the flow of liquids are built in as a requirement.
Unsurprisingly, the extent of change required to retrofit liquid cooling capabilities will be expensive and time-consuming, not least because data centre operators will need to manage the resilience and continuity of the data centre’s systems while the new cooling technologies are being installed.
While it is tempting to push ahead with immersion cooling to demonstrate sustainability and greener credentials, the risks of outages and disruption during the process are high, particularly as this technology has yet to be adopted on an industry-wide scale.
Outages in particular are not a risk worth taking. According to research from the Uptime Institute, outages – while rarer – are proving more expensive. More than half of respondents reported that their most significant outage cost more than $100,000, while 16% reported costs of more than $1m. This can increase depending on any liabilities a data centre operator incurs following disruption to a pre-existing customer service agreement.
How can legal experts support the transition to liquid immersion cooling?
Regardless of the challenges inherent in adopting new technologies, data centre operators must be ready to invest in measures that boost efficiency and operational resilience. This will inevitably entail changes to a data centre’s design, location, capacity, connectivity, and interconnectivity over the long term.
Data centre projects are inherently multi-disciplined, requiring specialist guidance across investment, financing, real estate, construction, corporate mergers and acquisitions, commercial, and procurement. As such, they must be underpinned by complex, interlocking contracts, agreements and licences, with any relevant legal issues and risks managed via due diligence and robust documentation from the outset.
When considering retrofits, for example, data centre operators will first need to review how much freedom and flexibility existing leases, user licences, collocation agreements, outsourcing agreements, fibre access and ring agreements, right of use agreements, and power purchase agreements provide, and whether there is any room for early termination or re-negotiation.
Even when able to proceed, operators will need to manage projects carefully to ensure that disruption or outages are minimised, stakeholder expectations are managed, and risks are mitigated. For extensive retrofits involving immersion cooling systems, implementation will need to be gradual, accounting for a period in which old and new systems co-exist.
Other legal considerations
Minimising business disruption during the implementation of new, more efficient technologies is a key concern for hyperscalers. Here are a few ways in which risk can be managed and mitigated.
- Ensure hard-wired options in the documentation suite allow for power and space increases, as well as future renewal and refusal rights.
- Future-proof collocation agreements to ensure rights for upgrades and the assimilation of new technologies are built in.
- Review data centre provider contracts and consider conditions that lock in carriers for a long term, with termination resulting in an option to terminate transaction documents and claim for damages.
- Ensure indemnities for any power losses or delays are built into transaction documents. You may also wish to secure guarantees from new system designers and suppliers.
- Check that the risk of early termination of a service agreement is within reasonable limits.
- Negotiate strict clauses on minimum service levels throughout the agreement to ensure operations continue seamlessly.
- Manage potential disputes between owners, operators, and customers by including contractual protections such as indemnities, warranties waivers, cost containment measures, and liquidated damages.
- Negotiate step-in rights. This is particularly relevant for hyperscalers that are implementing novel technologies that exceed a provider’s current capabilities.
- Review insurance policies. What protections do they offer for events such as outages and business interruptions? Do they cover legal expenses for issues such as lost data, a failure to provide managed services, fires, data breaches, or auto accidents?
In the UK, a consultation on introducing mandatory minimum security and resilience requirements for data centre operators is currently underway. No such framework exists in the Middle East, meaning data centre owners and customers must manage and mitigate risks themselves through careful drafting and due diligence on the back of expert legal advice.
This approach, supported by experienced legal advisory teams, will help prevent disputes and ensure that any data centre will reap the rewards from the day the systems are switched on.
This article was co-written by paralegal, Katie Gill.
Gateley Plc is authorised and regulated by the SRA (Solicitors' Regulation Authority). Please visit the SRA website for details of the professional conduct rules which Gateley Legal must comply with.