Typography
  • Smaller Small Medium Big Bigger
  • Default Helvetica Segoe Georgia Times

Building Artificial Intelligence (AI) data centers presents several challenges such as increased power consumption, the need for efficient cooling, and complex software management.

As AI adoption grows, addressing these issues becomes crucial for sustainable and effective data center infrastructure.

In addition, as the world embraces AI and Machine Learning (ML), the demand for data centers to support these technologies is skyrocketing. However, building and maintaining AI data centers comes with its own set of challenges and goals.

“Most of the AI workloads today live in the cloud, so any amount of AI growth will inevitably fuel further cloud growth, which we are seeing in projected cloud revenue as well as the amount of data center capacity cloud providers seem to be planning to add,” explained 2024 Trends in Data Center Services & Infrastructure by S&P Global.

However, AI data centers are still moving forward with a revenue market projected to reach $121.40 billion dollars in 2024, where network infrastructure dominates the market with a projected market volume of $56.15 billion dollars this year, according to Statista Consulting data.

Also, revenue is expected to show an annual growth rate of 6.04%, resulting in a market volume of $153.50bn by 2028.

In global comparison, most revenue will be generated in the United States with $99.16 billion dollars in 2024.

Scalability and Flexibility

As AI adoption grows, data centers must scale rapidly to meet increasing demand and this requires flexible infrastructure design.

Predicting future requirements and building adaptable systems is critical for long-term success because AI data centers play a pivotal role in supporting the AI ecosystem.

AI is partly responsible for the space demand data center markets. For example, Northern Virginia and Phoenix had an extremely strong demand for data center spaces; for this reason, providers and hyperscale firms are buying large amounts of land seemingly in advance of potential requirements, highlights S&P Global.

Moreover, real estate funds and other property investors have been entering and expanding in the data center sector, said the S&P Global research.

“This adds competition in some locations, but it can also offer data center providers the opportunity to sell facilities with predictable revenue that appeals to investors, while freeing up capital for the data center provider to build new facilities in locations that may entail more risk but offer growth potential.”

S&P Global added that the business models become more specialized in managing capital and designing, building and operating facilities, as well as serving specific customer segments such as AI Data Centers.

“Ideally, if you want to construct a massive data center, you'd choose a location with affordable real estate costs, easy access to energy, low natural disaster risks, and the ability to connect to high-capacity network infrastructure. Unfortunately, sites that meet those criteria have already been used up in many cases,” commented a Data Center Knowledge analysis.

The recommendation for this challenge is to think creatively about where to locate AI data centers, considering places where the data center can grow, without leaving aside access to resources such as water and energy.

Power Consumption

AI workloads are power-hungry and can consume massive amounts of electricity, resulting in a significant carbon footprint. Also, AI applications are becoming ubiquitous, from personalized recommendations to autonomous vehicles.

AI represents 4.5 GW of power consumption today and projects this to grow at a CAGR of 25% to 33%, resulting in a total consumption of 14 GW to 18.7 GW by 2028. This growth is two to three times that of overall data center power demand CAGR of 10%, estimated the Energy Management Research Center by Schneider Electric.

Additionally, the International Energy Agency (IEA) predicts that global electricity demand, fueled by AI growth, will double by 2026. This surge in power consumption poses significant challenges for data center operators.

Furthermore, chips and central processing units optimized for AI consume up to three times more power than previous generations. For these reasons, data centers must find ways to meet the increased power demands of AI clusters while minimizing their impact on the environment.

To address the energy consumption challenge requires a holistic approach by embracing sustainable practices and investing in efficient cooling solutions. Data centers can support the AI revolution while minimizing their environmental impact.

Cooling Solutions

AI workloads generate substantial heat and traditional air cooling may not suffice for high-density AI servers.

On average, 20% to 30% of servers in large data centers are unused or obsolete but still consume electricity, according to a Siemens analysis.

“These ghost or zombie servers not only add costs to the electrical bill, but they also create excess heat and drive higher demand for cooling throughout the data center. As a result of all these factors, most data centers are overcooled, wasting valuable resources,” detailed the analysis.

A solution for this challenge is free cooling that uses the outside temperature to cool a medium so that this medium can reduce the load and the heat generated by the computer center that generates heat all day because the computers process data, explained Isaac Jiménez, Director of Operations in Stulz, Mexico.

Free cooling technology helps data centers to obtain their ROI (Return on Investment) in a period of three years, said Jiménez.

This type of technology implemented in data centers can have energy savings of up to 30% and 40% annually.

Sustainability

Concerns about the data centers' sustainability focus on the large amount of electricity consumption and how efficient they are.

“Concerns about the challenges of getting energy to data center sites  are accompanied by concerns about water use, noise, diesel generator pollution, the number of jobs created, and the facilities’ large size and their arguably unattractive external appearances,” explained S&P Global study.

Data centers will need to look beyond their individual sustainability efforts to examine how their infrastructure is impacting the community.

However, AI can predict cooling requirements, optimize airflow, and identify energy-saving opportunities. By proactively managing resources, data centers can reduce overall energy consumption and carbon emissions.

AI-based robots could handle tasks like maintenance, monitoring, and equipment optimization, leading to greener operations

Also, AI-driven insights enable data centers to make informed decisions about energy usage, from load balancing to renewable energy integration, AI contributes to sustainable practices.

North America Opportunity

Strong demand and developer appetite continues to drive new construction of data centers in North America, according to a CBRE research.

Preleasing activity in primary markets is strong, with 73.1% of the 2,287.6 MW under construction preleased, and hyperscalers are securing spaces 24 to 36 months in advance of delivery.

“However, a lack of readily available power and extended lead times for key pieces of electrical infrastructure are delaying construction timelines.”

In the region, data center operators are prioritizing power availability, rather than selecting markets based on location, connectivity, and water and land pricing, detailed a CBRE analysis.

U.S. data center operators will have the major challenge of decreasing Scope 1, 2 and 3 emissions for carbon reduction mandates while also overcoming supply chain delays and power shortages.

“We expect operators to expand in markets with the most renewable power availability, such as Texas, Central Washington, Montreal, Quebec, Des Moines, Iowa and Umatilla, Oregon,” said CBRE.

And Reno, Nevada and Charlotte, North Carolina’s power availability will continue attracting development.

Mexico and Latin America Outlook

Low supply, construction delays and power capacity are some of the challenges for data centers in Mexico and Latin America.

The main challenges for building data centers in the region include the following: the shortage of energy capacity (17.73%), creating an environmentally sustainable construction project (15.76%), lack of qualified labor (15.52%), government licenses (13.05%), delay in the delivery of materials (10.10%), connectivity infrastructure (9.85%), site selection (9.36%), absence of tax incentives (8. 37%) and others (0.25%), revealed the Market Research on the Construction of Data Centers in Latin America by Data Center Dynamics (DCD).

During first quarter of 2023, the inventory was 672 MW in Latin America, primarily across Brazil, Mexico, Chile and Colombia, according to CBRE data. Queretaro, Mexico, for example, has only 1.2 MW available for lease, explains a CBRE research.

In terms of average vacancy rate, the percentage in Latin America has declined from 12.2% in the first quarter of 2022 to 8.6% in the first quarter of 2023.

“This trend is most notable in Santiago, Chile where vacancy has moved from 11.7% in the first quarter of 2022 down to 3% in the first quarter of 2023. Queretaro, Mexico is also showing very tight vacancy at 3.1%. Most of the region’s new inventory has been preleased to hyperscalers, with persistent high demand,” highlights the research.

The increase of data center services demand impacts the rental rates that are rising in the region.

“Pricing is notably higher in Latin America compared with more mature Western markets in North America and Europe. Limited supply, high fees and taxes are important factors causing rental rate increases.”