Demand for AI will Create Challenges and Opportunities

The demand for AI (Artificial Intelligence) will challenge our existing power grid and drastically impact commercial real estate.

In recent years, growth in AI companies has been exponential, driven by technological advancements and a surge in venture capital. In Pittsburgh alone, a section of East Liberty was just coined “AI Avenue”. AI Avenue is a one-mile corridor home to 20+ emerging artificial intelligence companies who are tapping into the vibrant tech ecosystem of skilled AI engineers from Carnegie Mellon University.

Our client, Strategy Robot, recently signed a lease at Bakery Square, which is located along AI Avenue. Strategy Robot is led by serial entrepreneur Dr. Tuomas Sandholm, Angel Jordan University Professor, and Co-Director of CMU AI . This DoD-focused AI software product company is pioneering AI defense solutions and growing its Pittsburgh presence.

With the growth of AI companies in Pittsburgh and across the country comes an unprecedented demand for AI data centers. This has caused a massive spike in power demand, which hasn’t been seen in the US since the early 2000s, and AI data center requirements are beginning to have a significant impact on commercial real estate development and infrastructure.

“US utilities will need to invest around $50 billion in new generation capacity just to support data centers alone.”  Goldman Sachs

An AI data center is typically located in a warehouse and relies on a high-performance computing (HPC) infrastructure to train and deploy complex machine learning models, algorithms, and data processing. To meet these needs, data centers need to have a reliable, and sufficient power supply, strong network connectivity, and be located in a stable climate.

The considerable processing and storage need for AI have thrust data center development into the spotlight, with hyperscale sites expected to double globally by 2024.  iOptimize Realty.

On average, a ChatGPT query needs nearly 10 times as much electricity than a Google search. The average power draw of a full stack of servers in an AI data center can reach 50 kilowatts or more per rack, compared to a traditional data center which requires closer to 8 kilowatts per rack. With more power comes more heat which requires cooling. The growing power demand will force utility companies to upgrade their infrastructure and improve their energy efficiency because of the increase of greenhouse gas emissions levels. Utility companies may also need to integrate renewable energy and consider the use of fossil fuels to keep up with the demand.

Nuclear power plants, known for their capacity to provide large-scale, steady, and low-carbon electricity, are also well-positioned to meet the demand for AI data centers. The constant and high output of nuclear plants helps to prevent power fluctuations and outages that could disrupt operations.

Pennsylvania is the 2^nd largest nuclear capacity state, with nine nuclear reactors and five nuclear power plants.

Pennsylvania is also home to Amazon’s recently purchased Cumulus Data Center which neighbors the 2.5 gigawatt Susquehanna Nuclear Power Station, one of the nation’s largest nuclear power plants. This carbon-free AWS data center was purchased for $650 million and not only helps Amazon meet its emissions goals but has the ability to expand to 960 megawatts. Their strategy enables the tech giant to capitalize on the explosion in artificial intelligence with a reliable power source next door.

960 megawatts is equivalent to powering 800,000 US homes.

Here are 15 ways AI data centers can impact commercial real estate…

1. Instead of sexy, amenitized trophy office buildings, the hottest properties in commercial real estate are power-hungry data centers.
2. Real estate is no longer ‘location, location, location’ its now ‘land, water, and power’.
3. With growth in Edge Computing, we could see smaller more localized data centers pop up near tech centric areas in order to be closer to where the data is being generated.
4. Leasing rates are not per square foot but based on power consumption (kw).
5. The battle exists between advancements in technology and infrastructure constraints.
6. Speculative data center development demand is significant with properties often pre-leased before they are even out of the ground.
7. Data center locations will be driven by the availability of power in that area.
8. Data center hubs could attract other tech businesses, fostering a tech ecosystem.
9. Data centers will stimulate investments in renewable energy projects such as solar and wind farms.
10. Local and federal regulations will play a crucial role in the development and operation of data centers through zoning, tax incentives, and environmental regulations.
11. New data centers and ESG initiatives are colliding as companies balance power consumption with efficient power and water usage efficiencies (PUE and WUE).
12. AI data centers require 100+ megawatts, compared to a traditional data center requirement of 20 megawatts.
13. Colocation operators may need to procure renewable power, which can be expensive.
14. Existing data centers running near peak could become obsolete or require a massive power overhaul if existing customers shift to deploy AI.
15. Electricity shortages will pose a significant challenge in prime locations.

As AI becomes more and more complex, not only will it continue to transform the technology sector but also significantly impact the commercial real estate industry, infrastructure development, local economies, and tech hubs such as AI Avenue. I believe that Generative AI and LLM’s (large language models) will reshape our lives. But, is our nation’s infrastructure ready?