The year 2023 is almost at its end. Compared to the previous three years, 2023 almost felt like a bit of a breather. And while there might not have been too many major events, there were still some very important changes in the world. Of course, the IT industry also had its fair share of challenges and events. Without a doubt the main one is artificial intelligence (AI) and its seemingly sudden and very fast entry into the world.
Of course, there’s nothing sudden about it. Companies and experts around the world have been working on various AI systems and projects for decades. About one year ago though AI finally had its moment and entered the mainstream thanks to OpenAI and its creation ChatGPT. For a while ChatGPT became the fastest growing new service in history, reaching 100 million users in a record short time. A record that was beaten by Meta’s new social network, Threads.
But while Threads is arguably a dud and has so far failed to catch on, ChatGPT has indeed caught on, even if it had a cool down period after the initial frenzy. One of the reasons is that people quickly realized that the algorithm isn’t as capable as they expected and it’s prone to mistakes – the so-called hallucinations. But that didn’t stop people from using it, albeit less and being more careful.
Meanwhile, OpenAI continues to develop and improve all its AI projects, and many other companies are also rushing to push their own competitor services or different approaches. Virtually all of the IT giants already have some sort of AI service. One of the big names missing from the unofficial AI club is seemingly Apple. It’s been reported that the company is also working on AI projects but will continue its traditional approach of keeping it a secret for as long as possible and will only talk about it at its reveal.
There were a lot of other news and events with, about or because of AI. Its user base continues to grow. Depending on the survey, it’s estimated that up to 79% of teenage and 31% of adult Internet users already use some types of generative AI tools and services.according to data by Britain’s Ofcom. The use of AI for work is growing to 33% of respondents, with 58% also using AI for entertainment, and 25% to aid in study.
Add to this the fact that currently there’s little to no regulation of AI. So, regulators all over the world have been scrambling to try and draft rules to govern this technology which skeptics like Elon Musk say can destroy humanity if not done correctly. Even OpenAI’s CEO and cofounder, Sam Altman, has raised concerns and called for regulation. Then there was the whole debacle about Altman’s tenure as CEO where he was surprisingly ousted from the company, but returned after virtually all employees said they will quit and follow him.
Just the tip of the iceberg
So, while all of this was happening on the surface, AI has also been boiling hot beneath the visual drama. It has changed a lot about data centers, too. The demand for data centers has jumped to record levels globally and it’s mostly due to AI and related technologies as they require massive amounts of hardware resources.
The analysts from JLL recently noted that the demand for data centers in the second quarter of 2023 in Europe alone has been an absolute record. In Q1, the demand for data centers in Europe was 51MW. In Q2 it skyrocketed to 114MW across the leading markets of Frankfurt, London, Amsterdam, Paris, and Dublin – the so-called FLAP-D.
London remains the largest market with a 35% share of tier one supply or a total of 902MW. Frankfurt is second with 656MW or 12%, and nets the biggest gain of 44MW compared to last year. JLL says that overall, the data center space increased with 73MW coming online in Q2 – that’s still not enough to cover demand and vacancy rates are falling fast. Frankfurt’s vacancy rate is down to 6%. London’s is down to 17%. Amsterdam is third, with 458MW capacity and a flat 18% vacancy, whilst Paris has 374MW and 17% vacancy. Finally, Dublin offers 199MW and 4% vacancy.
What’s driving the growth? Yes, you got it right – mostly AI. A lot of companies are training AI models. And currently, this doesn’t require low latency, but mostly power and storage. So, there’s benefit to build data centers closer to sites with renewable energy sources, cheaper land and water. JLL expects that this trend will continue all over Europe.
“The gold rush of AI continues to drive data centre growth even further and is opening an exciting new chapter for our industry. The second quarter of 2023 was another record-breaking period for data centre demand across Europe, with the activity showing no signs of abating,” Tom Glover, Head of EMEA Data Centre Transactions at JLL explained to DataCentreMagazine.
Gartner predicts that while in 2023 only 5% of enterprises are using Generative AI applications, that will soar to above 80% by 2026. Meanwhile, IDC predicts that by 2027, enterprises will spend $143 billion on Generative AI software and related infrastructure hardware and IT services. In 2023 it will be just $16 billion.
AI, AI, AI
As we can imagine, AI is not driving only data center demand. It’s forcing changes in the entire segment. Gartner predicts that by 2025, AI will be operating in half of all data centers in the world. And by operating, it doesn’t mean just crunching numbers for other services; it means that AI will take part in the daily operations of data centers.
And there are many areas where AI can help. It will help automate a lot of the daily maintenance tasks, it will find and fix inefficiencies, and can optimize resource usage. All of that will be needed, as data center workloads are expected to increase by 20% every year in the foreseeable future.
AI in data centers will be key to providing maximum utilization of resources and ensure everything is operating as well as possible. Especially considering that AI in general demands massive resources and it works much faster than regular apps and setups. So, it will need AI to help run all of the hardware properly and data center operators have no choice but to implement it fast.
Meta for example is already embracing AI into its data centers. The company has developed a new generation of data centers where AI has a key role. They are redesigned to consider the specific requirements of AI, but also to make sure the facilities are easier to adapt to new changes. “95 percent of our infrastructure today supports more traditional x86, storage readers, front end services – that’s not going to go away. Who knows where that will evolve, years and years from now? And so we know that we need that,” Alan Duong, Meta’s global director of data center engineering, said to DataCenterDynamics.
So, the company will use AI systems to optimize data centers and mix AI specific hardware with colocation and other services. “We can’t predict what’s going to happen and so that flexibility in our design allows us to do that. What if AI doesn’t move into the densities that we all predicted,” says Duong.
AI drives changes in data centers in all areas. But it’s not the only reason. Sustainability continues to be the second major topic and challenge for this segment. So, data center operators are working hard to improve the cooling. AI can help, by optimizing a lot of operations, but it can only do so much due to basic physics.
So, operators are also working hard to improve the overall cooling, too. Liquid cooling has been a big topic this year with a lot of developments in this area. One of them is the increased interest in submerged cooling and even various hybrid cooling combinations. One of them is direct-to-chip liquid cooling with heatsinks while the server is also submerged. Then air-cooling is minimal, left to just help dissipate possible excess heat.
The benefit of such an approach is a smaller footprint, much quieter overall operation and the option of keeping the server room temperature slightly higher, thus saving on overall cooling. The liquid cooling setups offer a lot of variations, depending on the needs of each data center. Meta has opted for a hybrid system which is slightly more expensive but offers greater flexibility and can accommodate a combination of direct-to-chip liquid cooling for the GPUs and air cooling for the traditional servers. It already sees benefits as it can keep the water it uses to cool the hardware to 30C degrees, and it even hopes it will serve as an example for other participants of the Open Compute Project (OCP) to adopt this temperature. Higher temperatures mean less energy needed to cool it down, thus these a few extra degrees can have a big impact in the energy bills and CO2 footprint of data centers.
Expect a lot more of the same
One thing is for sure. 2023 was just the preview and the start of all of these changes. The main challenge will be to keep adding enough data center capacity to fulfill the rising demand. General industry expectations are that won’t be the case. “Data center delivery has become increasingly difficult in recent years. This is due to a number of factors, including the need for larger amounts of power, competition from industrial developers, and an unreliable supply chain. This will lead to increased development timelines and likely higher lease rates, but also presents an opportunity for companies that can bring supply to the market quickly,” says David Liggett of datacenterHawk in a recent market overview, quoted by DataCenterFrontier.
Generative AI is also not going anywhere. In fact, it’s already passed the Rubicon point and it’s unstoppable now. The only question is which generaldirection will AI take, but in any case, it will always require a lot of resources. So, data centers will have to continue to take that into account.
Sustainability is also a topic that will continue to be a key part of data centers. There’s a lot of ground to cover in this area and many new opportunities to be explored. One thing is for sure, the data center segment had a very important year in 2023 and it’s only the start of the massive changes that are about to reshape the entire world, data centers included.