Are data centers a fad and a bad place to have a career in?
Multifamily, industrial, retail, office, these are institutionally know, will stand the test of time and spending years or decades perfecting your knowledge in one could be rewarding and allow you to maybe do your own thing, or make a lot of money in corporate.
Although data centers seem to be the hot new thing, if AI is truly a bubble, or technology changes where these centers can be much much more efficient, or energy production changes, will these things still be relevant? The requirements are also so tough that I feel like at some point you just really can't build more of them so if you have an acquisitions role, will you hit a ceiling?
Mainly focused on whether this is a niche that has suitable long term trajectory or if it's safer to stick to the major food groups.
Data centers are far from being a fad and represent a promising long-term career trajectory. Based on the most helpful WSO content, here’s why:
Exponential Growth in Data Consumption: The demand for data centers is driven by the rapid growth in data consumption, cloud services, and the increasing reliance on digital infrastructure. Nearly all consumer and enterprise applications are moving to the cloud, making data centers critical for supporting this shift.
Institutional Investment and Stability: Data centers are increasingly being recognized as a stable asset class, with institutional investors actively participating in their development and management. Operators like Aligned, QTS, and Equinix are examples of major players in this space, often hiring talent from investment banking backgrounds (e.g., LevFin).
Technological Evolution and Adaptability: While technology may evolve to make data centers more efficient, this evolution is unlikely to render them obsolete. Instead, it will likely create opportunities for innovation in design, energy efficiency, and scalability. For example, advancements in 5G, IoT, and AI will only increase the demand for robust data infrastructure.
Barriers to Entry and Competitive Advantage: The high requirements for building and operating data centers (e.g., energy, cooling, and connectivity) create significant barriers to entry, which can protect the industry from oversaturation. This also means that professionals with expertise in this niche will remain in demand.
Career Growth and Diversification: Working in data centers offers diverse career paths, including acquisitions, development, operations, and strategy. While acquisitions roles may face limitations in certain markets, the broader industry offers opportunities to pivot into other areas, such as finance, strategy, or even transitioning to related sectors like telecom or cloud services.
In summary, data centers are not just a "hot new thing" but a critical component of the global digital economy. While traditional asset classes like multifamily, industrial, and office are well-established, data centers offer a unique opportunity to be part of a rapidly growing and evolving industry. If you’re concerned about long-term stability, the key is to stay adaptable and continuously build expertise in areas like energy efficiency, cloud infrastructure, and emerging technologies.
Sources: An Overview of Technology Media and Telecom (TMT) - Part 2 of 2, Interviewing For Infrastructure Investment Roles, Be wary of greener grass; there are always tradeoffs..., I'm long office properties, What's the catch in CRE?
Is there a way to turn off this AI swill?
These answers are generic and terrible
Which actually makes them a perfect response for this question!
Bump
I think data centers have a long runway. Society is becoming more online, not less and regardless of AI we're going to need more data storage. Everything else you mentioned won't affect the overall trajectory of the asset class.
Data centers in some iteration, are here to stay (at least until we can get them into space, then all bets are off).
Space is not a particularly viable place to put data centers.
Putting to the side that we still need something like 4 full generations of engineering advancement to make the technology viable, what happens when something breaks? Data centers, especially HPC data centers like the ones that run and train the LLMs, require a material amount of hardware maintenance. How does that get done in space? Space adds additional wear and tear that is not otherwise dealt with on Earth.
And space doesn't have people. You can't just roll a truck or have someone step out their office and put their hands on the system. You will have to send someone up to space to do it. Or, you have to basically have permanent stations up there and rotate the people.
Engineering is the easy part of space-based data centers.
I agree, and that's why I think it's a long runway, generations long.
The technology is already viable, what the fuck are you talking about? What do you think Starlink is? It is a networking backbone. A critical component of a data center.
You don't need people to service anything up there. The vast majority of failures in data centers are either 1) human caused or 2) caused by the air. Space is a vacum that doesn't have people bumping into shit and breaking it.
I think the answer is the same as it is with satellites...if something breaks, you don't fix it. It's just mothballed.
Now I am a data center developer (that believes humanity's relationship with AI inevitably ends the same way it did in the Duneverse and whose favorite historical figure is von Neumann), so I am by no means an unbiased commentor. But, some things worth keeping in mind.
TLDR: like in most bubbles, the specific company you go to work at is incredibly important for your long-term career prospects.
(1) Data centers are a fad in the sense that a tiny fraction of the announced data centers will be fully developed. And when they do get developed, the timelines are likely going to be materially longer than initially announced.
With that being said, data centers are a legit need. They will have good use cases over the long-term. It took Amazon ~10 years to get back to its Tech Bubble peak, but if you invested in Amazon at that Tech Bubble peak and held on to today, you would have made solid returns (~19% IRR). There will be a group of companies that make it through the bubble bursting that will follow a very similar trajectory.
The underlying assets will have long-term utility and there will be a group of LLM and LLM-adjacent companies that will be in a very strong position for long-term business success.
(2) There is a lot of LLM washing by the data center developers. A not insignificant portion of the "AI" data center spend that is being discussed (and by the hyperscalers in particular) are basically playing fast and loose with what gets counted as an "AI" data center as a decent chunk is just their normal cloud investment.
(3) Most data centers over ~100MW are really data center campuses. It will be multiple buildings on one property. This is important from a risk management perspective because this means what gets announced as a 1.2 GW facility may come online in 200 MW chunks. This means if shit hits the fan after 600 MW gets built, there is the ability to just stop. Now, the utility minimums will hurt as a 1.2 GW facility will likely have some minimum payments owed to the utility regardless of demand. But most large data centers are campuses that come on in stages which is a significant derisker (although the benefit should not be overstated).
There is a lot more flexibility in the system than it may first appear.
The thing about this development cycle that is fundamentally different than the 90s is the utilization. There is essentially zero dark comptue nodes anywhere right now. The 90s saw massive amounts of physical infrastrcture be developed that was never used. The last time I heard about a dark fiber discovery was in 2018, and there was a potential discovery in 2022 but I am not sure if that was a 90s era build out one that came in the mid 2000s with the O&G infrastructure boom. There was a shocking about of dark fiber built out then as well that was expected to serve the oil fields that later went bust.
Are data centers a fad? No. Are they a good place to have a career in? Likely also no. I think we are 7 - 10 years away from pretty much all new compute being done in space or on the edge. Data storage still be a massive space, but that will likely be able to be handled by the existing infrastrucutre for 15 - 20 years. It is tough to guage this, but I think earth bound data centers don't really make much long term sense for 90% of all industry and consumer compute. Unless you have something that needs microseconds of latency advantage or 100% physical destruction security, space will be an exponentially cheaper place to get your comptue done. Within 20 years 99% of all non direct interactive comptue will be done in orbit.
Space? In 10 years? And within 20 years 99% of all non-direct interactive compute being done in orbit?
Yes. Space removes massive amounts of ground based constraint, it has unlimited power and space, no pun intnended. The biggest constraint is cooling, but non pressure vessels are exponentially easier to cool in space than things like the ISS. Starlink is the networking backbone for a MEO data center constalation. Data centers are an interesting asset class, mostly because the operate much like a power plant in that the construction cost is a small component of the lifetime cost of the facility. If you remove the largest variable cost driver (power) you can radically transform the lifetime equity return for those deploying hardware. Sure we will have to replace the power cost with lift cost, but at a target $10/kg you can deploy far more compute into orbit than you can on earth with the same amount of capital. The costs to build the cooling and power infrastrucutre on earth in many cases is more than 50% of all of the deployment costs. That might seem low, until you realize that filling the data center with GPUs and other compute hardware is part of that cost.
It all comes down to $/compute unit. Based on the cost reduction curves for lifting mass into orbit, and the construction costs on each side I am betting on orbit being far more cost effective.
LOL you have no idea what you’re talking about
A couple of aspects that make me skeptical of the sustainability of the current data center market are: 1) that right now we have Google, Microsoft, Open AI, Anthropic, etc. pouring huge amounts of money into LLMs and I think there’s a good chance there is a consolidation akin to Google’s search dominance. Once the yahoos of the world realize they’ve lost, they’ll pivot elsewhere. 2) Data Centers (or at least the racks) are as much of a technology as they are real estate. Computing technology has a long track record of getting more efficient and requiring smaller form factors to store the same amount of data. Our computing needs are certainly still growing but I think it’s a matter of when not if that slope starts to flatten and requirements plateau. Someone will be left holding the bag and it’s a big bag.
Not really, while it is true that individual compute units (as much as you can define that) have gotten smaller and cheaper, the consumption of that compute far outstrips the ability to serve the compute needs purely on efficency compression. We are going to see a massive continued boom in compute construction. I just don't think long term it will be terestrial.
Private credit
youre missing a massive part of the equation - as cost of compute continues to decline (and not to mention the need for computer expected to increase 10-50x in the next 5 years) , your model requires the assumption that the cost of tokens decline at an equal or faster right. If your model is wrong, massive upside in profitability of these firms.
It’s paradise. Comp is insane. Beneifts are unheard of. I’m an Analyst that pivoted from IB. My base salary (not including bonus which tbf is a smaller percent than typical IB) is more than my base in IB and I’m working half the hours. Now for long term, I’m not sure. There’s a lot of really good points in the comments here.
What's your role? Strategic finance? Corp dev? etc
Project finance, with some M&A exposure
Honestly, calling data centers a "fad" feels a bit like calling the internet a trend back in the 90s. Even if the current AI hype cools down, the sheer volume of data we generate daily through cloud computing, streaming, and global logistics isn't going anywhere. From what I’ve seen, the "ceiling" people worry about actually creates a massive moat; because they are so difficult to power and permit, the ones that do get built are incredibly high-value assets. While the "major food groups" like industrial or multifamily are definitely safer bets for traditional career paths, the technical complexity of data centers means there's a lot less competition for top-tier roles. It’s definitely a niche, but as long as we aren’t reverting to paper files and landlines, it’s a sector with a massive runway.
I get so fed up with all the doomer talk. Is there probably long-term over investment in infrastructure (or somwhere in the AI value chain)? PROBABLY! Was that true in the 1996-2000 era? YES! Were there extreme profits to be made during and after that time in the internet? YES! AI is bigger than the internet, AI is the new internet.
Except the early internet was exciting and LLMs alternate between dystopian and painfully lame.
One thing people rarely mention is that data centers are slowly becoming less of a “real estate” business and more of an infrastructure + energy business. The winners long term may not be the people who only understand acquisitions, but the ones who understand power procurement, grid constraints, cooling systems, fiber connectivity, and government approvals. So the real question isn’t whether AI is a bubble. The question is whether digital infrastructure itself keeps expanding and I honestly don’t see the world becoming less dependent on data anytime soon. If I were entering the space today, I’d focus on learning the operational and energy side early, because that knowledge will probably stay valuable even if the market cycle changes.
As someone who doesn't build them and who is an AI skeptic, I think the answer is both yes and no.
Obviously data centers are a "fad" in the sense that currently there is an enormous amount of hype and volume in that space. Lots of capacity being announced won't be built, and even if it is, that kind of blistering pace isn't sustainable. On the other hand, it's equally obvious that data centers will continue to get built going forward, since they are a real need.
I think the answer to your question is that you shouldn't focus on this particular niche, not right now. As with every other gold rush type asset, people are pouring into this space without having any experience or expertise. Yields are going to compress as every bro who thinks development begins and ends with raising some friends & family money and then going on the conference circuit bids up prices. The time to enter a given space is not when everyone else is doing so, but when the tide goes out.
Look, the big challenges for data centers seem to be permitting and limited capacity for local utilities providers (again, I say this as someone who simply reads industry news, not an expert, so correct me if I'm wrong). Solving those challenges are something you'll learn about if you do literally any other kind of development, at least in part. As always in real estate, start getting reps in somewhere and worry about the long term trajectory of your career a little later.
I don’t think data centers are a fad. Entering the field opens up possibilities to move into adjacent sectors, some of which are in early innings.
Personally, if I’m a believer that digital immortality will one day be a mainstream need for people (from creating a video documentary, to digital twin, to full mind upload), we are going to need more data centers.
You could pull the plug on Grandma twice!
Considering the capital inflow here. Is this a space to get some good transaction reps in? Or is it too niche and likely wouldn’t apply to other sectors.
Also curious about this question… I see datacenter jobs open near me but afraid that I might be putting myself into a pigeonhole
Bumping this, especially curious at a more junior level
Ex MBB in tech here. I don’t think the main risk is that data centers are a fad. The main risk is that the sector stops being “special” faster than people expect.
Right now, data centers feel like a niche where you can build differentiated expertise. But if the asset class keeps maturing, the best sites get controlled by a small group of scaled players, hyperscaler relationships become more institutionalized, power access becomes the gating item, and capital keeps flowing in, then a lot of the acquisitions / underwriting work could become commoditized.
That is the career risk I’d worry about: will being a data center real estate person still command a premium once the market professionalizes, or will the premium accrue to people closer to power, development, grid strategy, customer relationships, and capital formation?
My two cents are that we've been building data centers for decades now, so calling it a fad is ignoring history to a certain extent. It is, however, very sexy right now, and that attracts capital and talent and "prestige" in a way that it hasn't before. My gut is that our demand for compute is going to probably be greater than our ability to supply it. That's a function of data centers yes, but also chips and other critical inputs. I'm becoming more of a believer that as the cost of AI compute trends downward, it's going to seep into just about everything. If humans don't need to be in the loop because a lot of this is agentic, that probably means there's no real ceiling on demand.
Space is interesting, and misunderstood. Calling it "data centers in space" is a misleading because people assume Elon is trying to build some massive hulking structure up in space. The reality is that these will be small satellites connected together with lasers transmitting data. It's basically Starlink, as others have said. Space compute will be better at agentic AI than inference AI, so I'm guessing that Earth bound data centers will always need to be a thing to some extent. Whether or not we overbuild here on Earth, or whether space compute negates some of the data centers we build here is anyone's guess.
I understand the AI skeptics. I'm sure we will overbuild at some point. But man, the level of technological advance I am seeing in my own AI use is astounding - I'll be honest, it's even a bit scary. It's getting so good at so many different things. I really hope that even if you are a skeptic, you are still using it.
Error accusamus quidem veniam architecto ipsum omnis labore. Quia libero porro laudantium porro necessitatibus numquam. Rerum est minus aut. Sed aliquam veniam qui commodi cumque.
See All Comments - 100% Free
WSO depends on everyone being able to pitch in when they know something. Unlock with your email and get bonus: 6 financial modeling lessons free ($199 value)
or Unlock with your social account...
Quia et odit magnam reiciendis voluptatem corrupti corporis. Est placeat numquam ipsam quae earum saepe qui. Sint asperiores praesentium est reiciendis ut fugiat. Inventore dolor numquam saepe nihil aut.
Magnam voluptate autem mollitia veritatis. Facilis enim odio nemo laudantium. Id unde voluptatum consectetur neque voluptatum.