In November 2022, OpenAI released ChatGPT, which has amazed countless people. I said back then: The collective enthusiasm around generative AI is almost as overhyped as the metaverse concept in 2021. Looking at 2023-2024, it is clear that AI technology has achieved revolutionary breakthroughs and is having a huge impact on the construction and deployment of data centers and the overall network architecture design.
The reality is that the infinite expansion of our data consumption and the need for cloud-based services, combined with the bright future of generative AI, mean that the need for robust and secure data storage, faster data transfer, greater compute intensity, and greater data center efficiency has never been more urgent. I've said since the first move to the cloud: if there's one thing we're sure of, it's that our dependence on data centers will only increase, not decrease.
In 2024, the following four key trends will continue to impact data center operations.
1. There is no doubt that generative AI will actively promote the expansion of data centers
According to Synergy Research Group, analysts expect hyperscale data centers to triple in size over the next six years to meet the demand for generative AI. Just think about the billions of dollars that tech giants have invested in developing more advanced AI capabilities in recent years, and you can see how much the industry values the potential of AI.
Speed to market, i.e., how quickly hyperscalers, enterprises, and operators can get their data centers up and running, will remain a challenge and competitive advantage. Driven by massive data growth and the significant increase in computing power for generative AI applications and workloads, organizations must rethink how they plan, design, and build new (or renovate existing facilities) to meet today's growing demand, as well as anticipate demand in the months and years to come.
2. Labor and power supply will hinder the construction of data centers
One significant issue facing data center operations in 2023 is supply chain delays, which have led to shortages of chips and other basic products and raw materials. While these challenges have been largely mitigated, they are now replaced by labor shortages and power supply issues, so to speak, like not seeing a glimmer of light at the end of a fiber optic cable.
Overall macro market conditions have prompted central banks around the world to raise interest rates, which has fundamentally changed the economic state of data center construction and further pushed up the already shortage of construction labor costs. While operating labor—the technical work required to manage and maintain a data center—remains a challenge, it's nowhere near as urgent as the dozens of jobs and hundreds of workers required for a 100,000-square-foot facility—from land and raw materials to ribbon cutting.
In new AI deployments, the power consumption per rack has increased significantly. This has undoubtedly increased the challenge of accommodating AI clusters in existing data centers and made it more difficult to find new locations that can support the added power consumption. Building new power plants for these more energy-intensive data centers and providing the infrastructure to enable and manage the power supply has undoubtedly increased the challenge of finding new sites and obtaining regulatory approvals.
While increased productivity and efficiency are hallmarks of generative AI, it can also backfire on the building/labor and power sides. As the use and density of generative AI increases, the labor and power required to keep up with the physical installation and maintenance of critical components, such as new chips, servers, and cables, will increase significantly.
3. Sustainability will determine the success or failure of data center deployment
As corporate boards become more aware of climate impact, it is essential for companies to recognize and adopt efficient, sustainable practices at all three levels of the data center lifecycle – siting, construction, and operations—while avoiding increased material costs for day-to-day operations. Any device that consumes milliwatts of power must be optimized – from chips, servers, switches, cooling systems, to vending machines in break rooms. This challenge is exacerbated by compute-intensive generative AI workloads.
In fact, you may have heard some recent reports of metro areas expressing concern about the prospects of new data center proposals and all the power, space, and water impacts that may come with them. The current environmental philosophy is to see ecology and social responsibility as equally important as economic prosperity. While many companies have already invested resources to address their direct carbon footprint, 2024 will continue to see increased scrutiny of indirect carbon footprints – from supply chain impacts all the way to carbon claims for individual products and materials purchased for data center operations.
Still, current macroeconomic headwinds are putting significant pressure on business leaders to carry out the difficult task of adopting sustainable practices and processes without increasing costs or compromising profitability, which is most important. According to PwC's 2021 Global Investor ESG Survey, 81% of investors only accept a reduction in investment returns by no more than one percentage point in pursuit of ESG goals, while nearly half (49%) are unwilling to accept any reduction in returns.
4. The global data center regulatory game will intensify
We will continue to witness the game between lawmakers and enterprises around two key sticking points in the data center – sustainability (the land and energy needed to operate) and data sovereignty (where and how data is stored).
Some jurisdictions have issued decrees to suspend or suspend the construction of new data centers, while many others have imposed restrictions, including areas where protests are in the news, such as London, Amsterdam and Singapore.
On the one hand, delay demand and data sovereignty laws will force data centers to move more into local areas. On the other hand, while data centers play a necessary role in avoiding data sharing and processing abroad, they may face backlash from elected officials and distrust from local communities. Of course, one factor that is fueling this debate now and in the future is AI, especially as more and more businesses work to develop and deploy this technology into their own operational processes, while still being highly vigilant about the integrity of the data used to train their models.
The inevitable AI regulations will undoubtedly pose further challenges to this area, and these regulations will continue to gain momentum in Washington, including the recent executive order of US President Joe Biden, as well as the Artificial Intelligence Act in the European Union, elsewhere, national, and multinational jurisdictions. For now, enterprises and data center operators are still waiting to see how the safeguards in these plans will work and what impact they will have on AI exploration, development, and deployment.
Mystic Weapons: High energy efficiency is the only way out
In the broader technology landscape, we can hardly stumble upon a perfect custom solution, but all paths seem to point to an intrinsic solution – energy efficiency.
On the labor side, a more efficient data center means less labor is required from a design, power usage, and power density perspective, reducing the construction and maintenance burden. More efficient "plug-and-play" infrastructure products and architectures are easier to install, so they take less time and rely less on highly skilled labor.
On the labor side, a more efficient data center means less labor is required from a design, power usage, and power density perspective, reducing the construction and maintenance burden. More efficient "plug-and-play" infrastructure products and architectures are easier to install, so they take less time and rely less on highly skilled labor.
Finally, in a vast ecosystem of regulatory systems, high energy efficiency of data centers (characterized by minimized physical, carbon and energy footprints) coupled with transparent data sovereignty compliance helps reduce social and political resistance to data center deployment. With existing multi-tenant or colocation physical infrastructure, physical energy efficiency can reduce overhead and alleviate anxiety for local officials and communities who may not want to see heavy machinery break ground on a large, self-contained new complex.
Given the constraints imposed by competitive technological innovation, labor, sustainability, macroeconomic conditions, and regulatory pressures on hyperscale facilities, enterprises, and operators, continued global demand must be balanced with a strong commitment to energy efficiency throughout the siting/discovery, construction/design, and operation/management phases to drive data center growth and expansion.
