It’s been a remarkable year in the data center industry, with growing digitization influencing how critical infrastructure providers enable artificial intelligence, organizations of all types strategizing on how to use AI, and people realizing how data centers are part of everyday life.
The momentum of AI adoption increased, amplifying the data center space availability/demand imbalance. In response, hyperscalers incrementally increased capacity, and a record number of data centers are under construction1 to provide colocation for specialized high-density workloads and ongoing digital transformation.
What’s notable about the trend is the uptick of speculative building in markets previously not considered priorities – for example, Tulsa, Oklahoma and Madison County, Mississippi – where land and power are available at lower costs.
As secondary markets gain customers, cloud regions and AI compute clusters are becoming more and more distributed. Hybrid cloud is proving to be the infrastructure architecture of choice for enterprises and digital platforms alike, enabling the flexibility to locate physical and virtual resources where it makes most sense from a business process and industry perspective.
Density and Demand
AI isn’t driving transformation only for data centers. It’s compelling a wave of entrants into the “AI ecosystem” and a surge of new product offerings. You can now access bare-metal GPU servers on demand from startups like CoreWeave or rent cloud GPU computing services offered by companies that have been in business for decades, such as Oracle.
Data generation, analysis, storage and transport all happen in or through data centers, and of course the level for each of those has elevated. Also, the volume of data generated at endpoints outside of data centers is rising beyond imagination, whether that’s smartphones, industrial machinery, sensors monitoring heartbeats, public safety videos or the millions of other connected devices. Data from those sources eventually traverses both wireless and wireline networks and reach data centers – that’s why data centers are part of everyday life more than ever.
Enterprises recognize that their on-prem data centers don’t offer the power densities and scale required for the high-performance compute needed for AI and machine learning nor the high-speed interconnection to public clouds and entities in their digital ecosystem that colocation data centers offer. Consequently, colocation is continuing to be seen as a sweet spot for enterprise-grade private AI workloads as well as large language model inferencing, secure and sovereign data storage.
As for business results, we experienced record absorption in our New York, Chicago and Virginia campuses, and continue to see low vacancies across our major markets. Clearly, it’s precious to have large swaths of contiguous capacity in the biggest markets, and the ability to customize deployments at retail and wholesale scale.
Construction and Interconnection
If you’ve been watching data center industry press, you know that CoreSite has construction projects under way in Silicon Valley, Denver and New York. These facilities are purpose-built for modern workloads, incorporating what we’ve learned by supporting all types of applications during 20+ years in the industry, what our clients tell us they need now, and in anticipation of future data center requirements.
Several important AI-relevant interconnection technology enhancements were achieved in 2024. For example, we:
- Added AWS Direct Connect capabilities to our Chicago data center campus
- Partnered with Amazon Web Services to launch native 400 Gbps (400G) Dedicated Connections in markets located near AWS availability zones
- Launched the first Oracle FastConnect 50G multicloud networking capability in Los Angeles, Chicago and Northern Virginia
I’m proud to add being certified for the NVIDIA DGX-Ready Data Center program to 2024 accomplishments. It’s one more way we enable the ultra-high-density power requirements customers need while making it easier for them to deploy advanced technologies and bring their innovations to market.
2025: More of the Same … a Lot More
Looking ahead, broad trends and developments specific to CoreSite point to more of what we saw in 2024. There will be continued pressure on data center providers to create availability, support higher density workloads, enable the processing of massive data volumes and work with industry partners to push the limits on data transport speed and capacity.
What that means is further campus development which, at CoreSite, we think of as meaningful scale; it’s important to expand where it’s good for all stakeholders. We expect the data traffic traversing from CoreSite data centers to the AI compute regions to grow exponentially – and we have been preparing for that.
Earlier, I talked about AI creating new services. As it becomes easier to build AI capabilities into all applications, I think that we will see an even broader spectrum of density requirements and, in turn, “smart” location of AI services. I think it’s important to keep in mind that not every AI model needs to be trained at public cloud scale. Not every workload requires 100kW power density per rack. My point is that the type of services a given provider offers will define their niche. CoreSite has decades of success executing a multi-tenant business model that accommodates deployments ranging from a single rack to a full computer room, or even building across a wider power density spectrum.
The proliferation of AI has brought a cultural shift within the general public. Data centers are suddenly in the limelight, instead of enigmatic facilities most people never notice, and the recognition isn’t always positive. Clearly, there’s a need for education as to what data centers are, how they are essential digital infrastructure underpinning our digitized existence, and how "not all data centers are created equal."
In 2025, we will add content to our website that points out what data centers enable, which includes everything from the digital infrastructure facilitating voice communication and smartphone applications to essential services such as public safety, healthcare, education and transportation.
Ready for What’s Next for AI
I’ve remarked in blogs and articles during the last two years how excited I am to be involved in the data center industry. That feeling hasn’t waned at all. Imagine the future of entertainment, when you could build your own programs or movies based on your favorite shows, plots and actors. Think of the possibilities algorithmic quant trading arms might create, enabling financial advisors to grow investments and manage risk at unprecedented levels. Making purchasing suggestions based on your history is already here; consider the advantages for your doctor if they could suggest treatment informed by near-instant individual chromosome analysis. That’s why the impact of AI excites me.
I’m also confident that we are ready to empower AI with purpose-built data centers delivering the operating environment and interconnection infrastructure that enables near- and long-term proliferation of AI applications. The platform, and the underlying infrastructure we have developed, will remain the same but will be augmented for a much larger scale, with much higher power density, more ubiquitous interconnection and higher connectivity speeds.
I look forward to the rest of 2025 and beyond with high hopes – for CoreSite, for the data center industry and for AI. But mostly, I have high hopes for how our mission of empowering a more connected future will help improve everyday lives.
Know More
Juan Font is featured in several industry forums and publications, providing insights on data center industry trends and how data centers affect the way we work, live and play.
- InterGlobix Magazine (4th Anniversary Issue cover story)
- Forbes Technology Council (series)
- Investment Reports (content provider for Newsweek)
References
1. North America Data Center Trends H1 2024, CBRE (source)