Build now or wait for adoption to take hold? That is one of the key questions at NorthC Datacenters regarding AI at the moment. However, NorthC’s answer to this question is crystal clear. It is firmly committed to the first option and aims to grow rapidly in order to provide a robust and regional foundation for AI. We discussed NorthC’s plans with Alexandra Schless, the company’s CEO.
The ambitions of NorthC mentioned in the introduction are more than just words. The company, which has already grown rapidly in recent years from 10 to 25 employees, was recently acquired by investor Antin. This didn’t happen by chance, Schless explains: “We want to grow rapidly, and the new investor has capital available.” NorthC can now accelerate its growth and invest in new data centers. This could involve acquisitions of existing data centers, such as the six Colt data centers it recently acquired, but certainly also new construction.
Schless realizes that data center providers like NorthC play an extremely important role in the adoption of AI. “We are the builders of data centers,” she notes. In other words, NorthC is at the foundation of organizations’ AI ambitions, particularly for those that do not want their workloads to leave the region. “We believe in the new segment [the AI market, ed.], so we want to build faster and have more capacity to meet demand,” Schless explains.
Inferencing is the driving force
Schless expects a specific AI workload to take off, including in the region: inferencing. “That’s going to make the difference,” she states unequivocally. In any case, for the vast majority of organizations, inferencing is far more interesting than training AI models themselves. In practice, inferencing workloads will have a much greater impact on the day-to-day operations of organizations than any training workloads. Think of automatically answering questions, performing real-time analyses, and evaluating data to make quick decisions.
Inferencing consists largely of real-time data fed into trained models. When you say real-time, you’re essentially saying you want the AI and the data as close together as possible. Latency plays a key role here. Additionally, there’s naturally the sovereignty issue, which is becoming increasingly important for many organizations. In that case, sending data to the public cloud isn’t an option. Hosting everything in the public cloud is certainly possible, but then you’re faced with the sovereignty issue.
Keep in mind, it won’t all be just about inference. Especially as models become smaller and/or organizations want and are allowed to train models with sensitive data, training will certainly take place in NorthC data centers as well. In fact, that’s already happening. About ten months ago, we wrote about Juvoly, which installed the first Nvidia DGX B200 in the Netherlands at a NorthC data center in Rotterdam. It develops and hosts AI models there for use in healthcare.
Two key words: sovereign and regional
Schless observes that sovereignty is playing an increasingly significant role. “We knew it was a topical issue, and we also see that it’s a key factor in choosing a data center partner,” she says. That’s naturally beneficial for NorthC as a European company. In addition, NorthC’s regional approach in the Netherlands, Germany, and Switzerland enables it to offer a complete ecosystem of sovereign services to its customers in collaboration with regional IT partners.
The terms “regional” and “sovereign” are, of course, closely linked, but Schless still wants to emphasize “regional” without the sovereign undertones. “Many companies want to manage part of their IT environment themselves, and in that case, it’s nice to have the data center nearby,” she notes. This may be a very basic reason, but it’s one that comes into play much more often than you might think. It’s the case in the Netherlands (Schless specifically mentions the Eindhoven region), but even more so in Germany and Switzerland. Companies in those countries have an even more regionally focused mindset than in the Netherlands.
How does NorthC know what to build?
Schless expects inferencing to take off in a big way within two to three years. So NorthC needs to have the capacity for that. However, the data center of the future will look quite different from what we have today. It’s not just about more capacity.
Consider, for example, the power density per rack. Schless sees this shifting from around 4.5 to 8 kW to as much as 30 to 50 kW per rack. Such a shift brings new requirements for a data center. NorthC is certainly aware of this. Inferencing is therefore already more than just theory for the company, Schless tells us. The first use cases are already here, she notes. And discussions are underway with potential customers coming from the inferencing sector. According to her, the design for new NorthC data centers is already fully based on the requirements that this development entails
Modular construction means flexible construction
NorthC naturally already has a clear vision of what the data center of the future should look like. Yet they, too, depend on what other players in the market are developing when it comes to how they build their data centers. We spoke with someone from Schneider Electric a while back, who admitted that in the area of power supply, they unexpectedly had to skip a step and go straight from 400V DC to 800V DC.
Of course, the example above is quite specific, and NorthC Datacenters is just a step further along in the development cycle. Nevertheless, they too will inevitably encounter challenges from time to time when developing their own data centers. The modular construction method that NorthC has established as one of its core principles offers a solution. Consider, for example, data center cooling. “We’re already building data centers capable of both liquid and air cooling,” Schless notes by way of illustration. When adding a new module to a data center, it can be built exactly as needed.
The elephant in the room isn’t such a big deal for NorthC
We can’t have a conversation about data centers in 2025/2026 without discussing AI, but certainly not without discussing the energy supply for data centers either. According to Schless, the entire data center industry is affected by this. “However, there is a difference between regions and countries. And because we request less capacity than, for example, the hyperscalers, there are still opportunities for us in the various regions. Additionally, we can grow in stages, because we build modularly,” she states. She does, however, call on the government and grid operators to continue investing and, preferably, to significantly accelerate this process.
Grid congestion is primarily a problem in the Netherlands and certain cities in Germany, Schless adds, further qualifying the potential issue for NorthC. According to her, there is little congestion in Switzerland. “In Germany, we will see the strongest growth in the short term in terms of new data center construction. This makes sense, because they are catching up digitally there,” she notes. Additionally, virtually every region in Germany has a strong economy, so there are many opportunities for NorthC there.
Finally, regarding energy supply, it’s worth noting that NorthC is currently exploring the possibility of connecting directly to a wind farm. This would allow for greater use of renewable energy.
The ideal mix for the future
In our conversation with Schless and in this article, we’ve focused primarily on how NorthC plans to build data centers capable of serving the AI inference market. With several new construction projects underway in various countries, she confirms that the company is ready for future demand.
However, this does not mean that NorthC is building data centers solely to meet demand from the AI market. “We will always have a diverse customer base,” in Schless’s words. Those existing customers, who lease space in NorthC’s data centers for all kinds of workloads that aren’t necessarily related to AI, are and will remain the company’s “core business,” she states. “We want to continue serving the many customers we have.”
With the current plans for the next five to ten years discussed in this article, coupled with the solid foundation NorthC Datacenters already has, the future looks bright for the company.
Of course, this currently applies to virtually every player in the data center business, since without data centers, AI cannot exist or be used. Based on the relatively rapid yet well-considered growth over the past six years—from 10 to 25 data centers in what Schless calls Phase 1 of NorthC’s growth trajectory—we know the company isn’t shying away from the challenges involved. To be continued, no doubt.
