Subsea vs Immersion Data Centers

Subsea Data Centers: we might not be the last step in solving data center challenges, but we are the next step.

In roughly 60% of the meetings I go to as the face and founder of Subsea Cloud, a subsea data center provider, I am asked this question: Is this better [and/or different] than an immersion data center? The answer is always yes. The fact that the question is asked raises two points in the moment:

1. Did I not explain our solution correctly (if I did, the question itself should not be raised)?
2. Do people (who should know based on their industry) not understand the challenges of the data centers of today?

The challenges, broadly speaking and in no particular order, are:

1. Land scarcity issues
2. Power costs
3. Latency
4. Cost to build
5. Scalability
6. Control points
7. Sustainability

Subsea vs Immersion Data Centers:

How We Are Similar

The data center environment is improved – it is steady and benefits from temperature optimization. The technology has been proven to be a low cost, highly reliable solution.

Also known as?liquid submersion cooling,?immersion cooling is the practice of?submerging full servers in a thermally conductive liquid (dielectric coolant). Liquid cooling is far more efficient than air cooling due to thermal conductivities. Compared to air, water has?a higher heat-carrying capacity. How high? about 3500 times as high.

There are two types of immersion cooling: single phase or two phase. For reasons just beyond the scope of this article, the market is geared toward single-phase coolants as a more viable option, so we will forgo the description of two phase cooling. A single phase coolant remains a liquid and is pumped to heat exchangers to remove the heat.

With single phase fluids, the data center environment is improved – it is steady and benefits from temperature optimization. The technology has been proven to be a low cost, highly reliable solution. The fluids are green and clean, eliminate the need for fans and reduce cooling equipment requirements. This results in an increase in the power usage effectiveness or PUE ratio, whilst lowering maintenance requirements, so no knocks there and we ourselves benefit in many of the same ways. But this is not a panacea. It doesn't help solve many of the challenges we face.

We can't only care about optimizing servers. We have to care about optimizing the whole data center for the buyers, the users and, crucially, for the climate.

How We Are Different

We submerge our servers in a non-conductive fluid, too. Technically speaking, we produce immersion centers. Although we don't require pumps to remove the heat as we benefit from free, passive cooling. In other words, we completely eliminate the need for electricity to be involved in the cooling process. But what about the other challenges listed??Here's how we solve for those one by one:

1. Land Scarcity - land shortages are a cannot be ignored or played down. In order to circumvent land shortages, so builders build up, but this doesn't solve for weight issues in high density centers. Moreover, land costs a lot. The seafloor, however, is virtually unlimited in its capacity to house data centers. What's notable about our environment is that you cannot buy or rent the seafloor – it's a permitting process.
2. Power Costs - data centers typically go where power is cheaper. However, power is not always cheaper closer to metropolitan areas, pushing facilities further and further out. We, too, have had to find ways to procure cheaper power. We benefit from longterm contracts like many data centers, and instead of being landlocked and further way from our user population for cheap power, we sit coastally, benefitting from cheaper power from municipalities and providers. We also benefit from low attenuation rates where renewables are concerned because of proximity (such as oceanic wind farms) which results in savings.
3. Latency - as discussed, data centers tend to go where power and land are cheaper, but that's not always where the users are. This results in latency. More than half the world's population lives coastally. Due to the rate that signals travel, being closer to the coast, rather than in-land, reduces latency. Ultimately, this means consumers get a boost in performance. We all demand speed, from households to enterprises and for some sectors, such as health and finance, it's critical.
4. Cost to build - We recently had a debate with an executive who was convinced that "doing anything in the water is infinitely more complex and more expensive than on land." This is simply not true. There are actually far less complexities for a lot of subsea engineering projects. We've been doing a lot down there, successfully, for a long time. Comparing apples to apples, we are 21% less expensive than an on-land data center with a 100MW capacity. We don't need to benefit from economies of scale to be cheaper –?1MW facility for 1MW facility, we are less expensive (CAPEX and OPEX considered), too. *NOTE: The failure of one company to do something effectively and cost efficiently is a company fail, not a category fail.*
5. Scalability – We are infinitely scalable near densely populated areas. We provide rapid development and deployment quickly and seamlessly - way beyond what is possible on land and more so than floating data centers (see providers like Nautilus). We are modular and we are location-agnostic and politically-agnostic. We benefit from a rapid infrastructure delivery model and we have no upper limit on deployment numbers.
6. Control points - We can squeeze a lot out of them but buildings, in the long term, are not agile. Even floating data centers suffer from limitations, including harsher permitting processes and more to consider in terms of weather and natural disaster zones. As data center providers, we should be the producers of outcomes, not simply physical infrastructure. Because of our modular design and agility, we are a support solution rather than a control point for what gets delivered. We are designed to let a business do what it needs to, when it needs to, where it needs to. We are not limited by space and have more opportunities in terms of where we can deploy than other solutions. Unlike traditional data centers, there are no sunk costs (no pun intended)... we can change with a business's needs quickly and without damaging costs.
7. Sustainability – most importantly for us is sustainability. Data center sustainability is our key theme and something we will post about in the future. Data centers are the largest power consuming asset class on earth; they use millions of gallons of water each day and produce astonishing amounts of CO2. We use zero water, zero refrigerants, zero harmful chemicals. Every pod we put in the water prevents 750t of CO2 from being emitted. Our not too distant roadmap items are to generate our own energy and benefit from other parts of the energy ecosystem, consuming renewable power exclusively. This is possible at sea and one of the benefits is low attenuation from all of the renewable powers.

There are other issues that we haven't gotten into that typically I do talk about on calls with prospective clients, partners and investors. Issues such as data residency and sovereignty. Other issues such as natural resources available in certain countries and lack of infrastructure come up, too. Usually, by the time I've explained how our solution solves or circumvents these issues the question "Is this better than an immersion data center?" is asked and I have to wonder why. The world doesn't only have to care about optimizing servers. It has to care about optimizing the whole data center for the users and the buyers and, most of all, for the climate.

What is Definitive

Conventional air cooling is no longer state of the art given increasingly dense racks and demanding server loads – air results in terribly ineffective thermal management.

Traditional immersion data centers are a brilliant step in advancing data centers and are certainly effective. They've made, and continue to make, a big impact on the carbon footprint of the world. However, they are not the last step in solving data center challenges and making them sustainable for a hyperconnected world. They are no longer the best in class in power density and power efficiency.

Subsea Cloud's position is that placing centers subsea leverages all of the benefits of immersion cooling, but addresses all the other challenges that they–and traditional–centers do not. Immersion data centers are an extension of traditional data centers. Subsea data centers are an evolution.

Looking 10, 20 or even 30 years out, we might not be the last step in solving data center challenges, but we are the next step.