What if we’re wrong about data center power demand?

Last week, I wrote about the investment opportunities that the AI and data center boom brings across infrastructure, energy and software. Before delving further into the details of these opportunities, I wanted to address a question that we think about a lot at Smart Society Ventures: “What if we’re wrong about data center power demand?”

At this point most people know that the AI data center boom is forecast to bring astronomical power demand growth. Goldman Sachs Research forecasts that global power demand from data centers will grow 50% by 2027 and 165% by 2030. A single AI data center will consume the power of 800,000 US homes. By 2028, AI data centers could consume as much as 12% of US power. There will be approximately 122 GW of data center capacity online by 2030, with 70% provided by hyperscalers and wholesale operators. Given the more complex workloads demanded by AI, the density of power use in data centers is also forecast to grow significantly. At the same time, we have the rest of the economy to power. In short, we need a lot more electricity. Fast.

As discussed last week, the AI data center boom is creating many exciting investment opportunities across infrastructure, energy and software. As we evaluate these opportunities, we constantly ask ourselves (and so do investors) what would happen to our investment thesis if this AI data center boom doesn’t materialize as forecast. What if the power demand numbers end up being half of the forecast? Or less? Innovations in power and processing efficiencies like those announced by China’s DeepSeek, are a very real possibility.

While it feels like nearly everyone is talking about the power required by AI data centers, the fascinating thing about the power market is that global power demand was already forecast to grow significantly before anyone was talking about AI data centers. In fact, the vast majority of global power demand growth forecast is not from AI data centers; it’s from electrification and the growth of the middle class in emerging markets. When we layer on the power demand from AI data centers, the opportunity becomes exponential - and much more poignant as these data centers take electrons away from electricity-hungry industries and consumers. Moreover, the power demand puzzle becomes increasingly uneven, as areas where a large amount of power-hungry AI data center capacity is being built demand large amounts of local power and further risk imbalancing the power supply for other parts of the economy. As data sovereignty becomes increasingly important, AI data centers are being announced throughout the world furthering the global power demand picture and extending questions of regional concentrations. Put simply, there was already a huge global power demand growth forecast. Then we layered on the needs from AI data centers. And now it’s not just a big investment opportunity; it’s massive.

McKinsey predicts that global power demand will grow by a 3.5% CAGR through 2050 driven primarily by the electrification of industry and buildings. Looking at the puzzle however, AI data center power demand growth represents a relatively small portion of this forecasted growth, while the electrification of industry and buildings represent a significant share.

McKinsey: Forecast power demand by sector through 2050

On a global basis, both in scenarios with AI power demand and without it, non-US regions are forecast to drive a significant amount of global power demand growth. This is driven by two factors. Firstly, a large portion of the global industry and buildings that are being electrified (manufacturing, processing, etc) are located outside of the US. Secondly, 100 million people in emerging markets are entering the middle class each year, with 88% of this in Asia. According to Oxford Economics, by 2034, the emerging markets middle class will be 687 million households. As families enter the middle class, they demand more and more electricity as they purchase refrigerators, washing machines, TVs, hair dryers, vacuums, computers, mobile phones and much more.

McKinsey: Forecast power demand by region with and without AI

At SSV, we believe that AI is a transformational force that will become increasingly foundational to our daily lives. Looking ahead, we forecast that agentic AI will underpin everything in our economy and daily lives, much like the internet and mobile phones do today. We expect that the AI data center boom will come to fruition as forecast, and will require an increasing amount of electricity, making an already-complex power demand puzzle even harder to solve and increasing the investments needed in infrastructure, energy and the associated software “brain” that can help generate and orchestrate electrons for data centers and industry. Amid this growth, we believe there will also be strong investment opportunities in processing, cooling and energy management that will allow AI data centers to operate more efficiently than forecast. Software can play a big role in energy efficiency across AI data centers, industry, buildings, households and grids reducing overall power demand and loads, while other investment opportunities generate more electricity to satisfy demand. We’re excited about the energy and data center efficiency software as a part of the AI infrastructure and energy investment landscape, and we’ll write more about this in the coming weeks.

Given the global macro fundamentals of the power market, we see significant opportunities for investments in AI-enabled energy systems that can both serve data centers and industry. Providing onsite reliable power and microgrids for data centers is of course both a critical and massive investment opportunity, but so are investments in onsite power and microgrids for industries and households that demand more power as the emerging markets middle class grows, as industry electrifies, and as power is increasingly soaked up by data centers creating instability for consumers and industry worldwide. Our investment in Caban Energy, a company founded in Silicon Valley and providing onsite power for both emerging markets telecommunication infrastructure and backup power for developed world networks, is a good example of this. So is our investment in Voltiris, a company founded in Switzerland which provides onsite patented solar power systems for greenhouses advancing food security in developed and developing economies. The holy grail of AI-enabled energy systems is of course companies that can provide modular, scalable and onsite power on a global basis to data centers, industry and households, across developed and developing markets on both sides of the meter – or in some combination of these. Bloom Energy’s natural gas fuel cell technology, Mainspring Energy’s linear generator and Fourier’s modular onsite hydrogen system are all nice examples of this.

A third big investment opportunity we see is in grid infrastructure and software that drives greater efficiency, flexibility and resilience across existing grid infrastructure that serves data centers, industry and consumers throughout the world. We’ll dive deeper into grid technology and software with global application in subsequent weeks.

The fundamentals of global power demand create significant investment opportunities across infrastructure, energy and software - with the potential for strong investment returns. This is true with robust AI data center power demand forecasts - and with more conservative ones. Because the fundamentals are in fact driven by three factors: the electrification of industry, the growth of emerging markets and the boom in AI data centers. We are excited about investment opportunities that are fueled by one of these large macro trends. But we are most excited about investment opportunities that are fueled by all of these and have global applicability as they scale.

We’ll dive deeper on specific investment sub-sectors in the coming weeks.

Onward,

Brynne

Managing Partner, Smart Society Ventures

• 🇫🇷 🏗️ Ardian raised $20 billion for its record Infrastructure Fund VI to accelerate Europe’s energy, transport, and digital infrastructure expansion.

• 🇺🇸 🔥 Arbor Energy raised $55 million in a Series A to develop next-generation carbon-free gas turbines and scale clean, on-demand power for the AI economy.

• 🇩🇪 🤖 etalytics extended its Series A to €16 million, with Microsoft’s M12 joining, to scale its AI-powered industrial energy optimization platform across Europe.

If you’re new this week, you can catch up on our previous editions: “How to Break Down the AI, Energy & Infrastructure Market” and “AI-Enabled Power Companies.”