Data Centers May Prove Just as Valuable as Chips in the AI Arms Race
In a stock market fueled by excitement over artificial intelligence (AI) and its potential, much attention is paid to the advanced microchips used to power these ever-evolving technologies. However, the unsung heroes of AI’s rapid evolution are the data centers that form the backbone of this industry.
As big tech companies such as Microsoft, Amazon, Alphabet, META, and others race to dominate the AI landscape, the importance of data centers has become increasingly evident, arguably surpassing microchips in certain aspects.
To subscribe to the MalcolmOnMoney newsletter and receive more content like this, click here.
For instance, an AI model used for image recognition must analyze and learn from millions of images to accurately identify objects. This process demands not only sophisticated microchips capable of running multiple complex equations simultaneously, but also massive compute power that only well-equipped data centers can provide.
Data centers are a crucial component of the further development of AI because they provide the infrastructure, cloud storage, and processing power that AI systems require. Artificial intelligence algorithms learn from large datasets, and the training process involves filtering this data to identify patterns and make quick decisions.
These data centers should have both the physical capacity to house large servers and storage units and the advanced networking capabilities to handle the immense data flow involved in AI training and execution across global networks. For example, a user in Europe accessing an AI-powered service will experience a faster and more reliable performance if the data is processed in a nearby data center rather than one located halfway across the globe.
The world grossly underestimates how much the demand for compute power capable of handling AI-related tasks will expand the global market for data centers over the next decade. Amazon plans to spend approximately $150 billion to build data centers over the next 15 years providing the infrastructure needed to handle the increased workload. Microsoft is rumored to be working on a massive data center project that could cost more than $100 billion over the next five years alone.
Still, these big tech companies and their competitors will need to rely on partners to provide additional storage capacity to meet the needs of their sprawling AI projects. Real Estate Investment Trusts (REITs) specializing in data centers (like Digital Realty, Equinix, and others) spearhead the global data center infrastructure expansion.
Data center companies have been investing heavily via strategic acquisitions and development projects around the world, which signifies the growing influence and importance of data center REITs in supporting the expansion of AI capabilities. In fact, many of the big tech companies they serve are now requesting entire campuses for themselves.
Further propelling the demand for new data centers is the fact that older ones are difficult to revamp. Many of the existing data centers find themselves grappling with the daunting challenge of adapting to the demanding requirements of AI workloads.
Artificial Intelligence applications—particularly those involving deep learning—require extensive computational power and energy resources, which can exceed the capabilities of older, traditional data centers. The existing infrastructure that supplies electricity to current data centers unfortunately cannot support expansions.
The rapid growth of data centers has led to some challenges, particularly at the local government level. With their extensive cooling and power requirements, data centers can strain local power grids and raise several questions about sustainability. Additionally, the physical size of these facilities can lead to land use conflicts, where local governments and homeowners voice their resentment over their development in or near residential areas. Many communities have expressed concerns about the environmental impact, energy consumption, and potential changes in the local landscape.
In some European countries, for example, there have been instances where local governments have pushed back against the construction of new data centers due to fears of excessive water usage and energy consumption. In response, data center REITs are increasingly focusing on enhancing their energy efficiency, investing in renewable energy sources, and engaging more with local communities to mitigate these concerns.
But these concerns could be short lived. There have been academic breakthroughs, like the one sponsored by the U.S. National Science Foundation and published in Nature Scientific Reports, that highlights the way that a new, ultra-energy-efficient method of chip production compensates for the temperature variations that typically cause legacy data centers to eat up approximately 2% of all electricity used in the U.S. annually.
Modern data centers are designed with advanced cooling systems and energy-efficient infrastructures to mitigate these demands, which not only reduces operational costs but also aligns with the growing emphasis on sustainable technology practices. And with the number of companies like Amazon, Alphabet, and Microsoft currently pouring billions of dollars into designing and producing their own chips capable of powering data loads at this level, it is also likely that in the near future, we will not need to consume nearly as much energy as we currently do.
While the graphics processing units (GPUs) designed by Nvidia dominate the AI landscape today, they were never intended to power generative AI the way they do now. While inefficient, they are currently the best solution. But with the number of startups currently pursuing a microchip powerful enough to rival Nvidia’s GPUs, with less demand on the power grid, other tech companies will soon find themselves able to run their full suite of AI products at a fraction of the cost.
Nvidia’s CEO Jensen Huang, estimates that $250 billion will be spent on data center equipment annually, and believes that is where the more durable investment opportunity lies. Executives from Blackstone, a leading asset management firm specializing in real estate, believe that data centers represent a generational buying opportunity in commercial real estate.
While microchips are the brain of AI systems that enable the rapid processing of information, data centers are the body that sustains and supports these operations. They encompass the necessary hardware, energy, and networking that make such advanced computations possible. As AI technologies continue to evolve and demand more from the underlying infrastructure, the role of data centers will only grow in importance, possibly even outpacing the innovations in microchip technology in terms of their criticality to the AI arms race.
*********************************************
Malcolm Ethridge, CFP® is the Managing Partner of Capital Area Planning Group based in Washington, DC. He is also the Managing Partner of Capital Area Tax Consultants.
Malcolm’s areas of expertise include retirement planning, investment portfolio development, tax planning, insurance, equity compensation and other executive benefits.
Disclosures:
The information provided is for educational and informational purposes only, does not constitute investment advice, and should not be relied upon as such. Be sure to consult with your legal advisors before taking any action that could have tax and legal consequences.
Investments in securities and insurance products are:
NOT FDIC-INSURED | NOT BANK-GUARANTEED | MAY LOSE VALUE