Beyond the hype- what is really happening with AI?
Society is rapidly adopting new Generative Artificial Intelligence (GenAI) tools which can create new content such as video, text, music and graphics.
Tools such as MS Copilot and Google Gemini are revolutionizing business and personal productivity. AI is making it easier to create emails, documents and marketing materials. AI is also enabling more complex business productivity tasks such as writing code and providing virtual “chatbot” interfaces with customers.
Life changing advancements in science, medicine and engineering are also beginning to be realized through GenAI. With expert human guidance, AI algorithms can find patterns in large sets of technically based data, suggesting new relationships, efficiencies and theories that may have been overlooked previously. This is already benefiting us through the rapid development of new molecules, materials and complex predictive models.
One of many advancement examples can be seen with researchers from Harvard Medical School who are using advanced GenAI tools to greatly accelerate the time to develop new, more effective vaccine designs. By being able to analyze extensive biological, structural and evolutionary data using AI, they are able to better target the evolutionary patterns of viruses, thus accelerating the development of vaccines for targeting COVID and other viruses that can rapidly mutate.
AI and Data Centers drive electricity demand
GenAI applications require large amounts of data and large-scale computing power. The AI surge is manifesting in a construction boom of the regional data centers which house the data processing, storage and networking infrastructure needed for AI. A recent report from Accenture estimates that data center power use will increase tenfold to 612 terawatt-hours (TWh) by 2030, which is comparable to the energy draw of Canada. In the US, the percentage of overall electricity consumption from data centers is anticipated to rise from today's 4% to between 6-18% by 2030. Energy demand will be significantly higher in data center rich regions, such as Virginia and Texas.
While rapid data center expansion has its environmental challenges due to increases in water and energy use, it will also provide new opportunities for accelerating clean energy growth.
Data center providers including Microsoft, Google and Amazon AWS are creating strong clean energy demand signals in the regions that they are deploying. This is often in the form of purchase agreements with renewable energy, energy storage and other clean energy providers. These commitments can be gamechangers, retired nuclear plants in Pennsylvania and Iowa, for example, are being revitalized to provide clean electricity to data centers.
The big AI tech firms are also creating much needed demand signals for earlier stage low carbon technologies through partnership agreements. Google is one firm making big bets on the next generation of clean energy technologies by partnering with geothermal, advanced nuclear, fusion and energy storage companies.
While many data centers will be fueled by natural gas plants in the short term, traditional gas suppliers Exxon, Chevron and Equinor are making announcements for plans to deploy carbon capture technologies and infrastructure to greatly reduce their carbon footprint. According to a recent purchase agreement announcement made by Google, we could see such infrastructure in the ground by the early 2030’s. Google’s regional data center energy demand would be met by electricity from a Midwestern natural gas plant equipped with carbon capture and storage (CCS) technology, which is estimated to reduce 90% of the plant’s CO2 emissions.
Beyond clean energy, low carbon steel and cement are needed to reduce the carbon footprint for new buildings such as data centers. Microsoft recently announced a partnership with Stegra, a Swedish green steel provider, for near-zero emission steel to be utilized in its data centers, sending a global demand signal to clean steel manufacturers.
Using AI to accelerate the energy transition
Another facet of this rapidly evolving story is how AI is advancing clean technology innovation. One landmark study estimates that by 2035 up to 5.3 Gigatons of global CO2 emissions annually can be avoided by using AI based methods to accelerate low carbon technology adoption rates and efficiency. Achieving such a reduction would be roughly equivalent to taking the annual emissions footprint of the US off the table.
How is this possible? For efficiency, the large sums of data becoming available through advanced sensors and monitoring are already enabling AI tools to optimize energy use from buildings, transportation and large-scale industrial processes. In an interesting twist, AI is also making its use of the electricity grid, the energy delivery system that data centers depend upon, more efficient. AI tools are helping to balance data center energy use during peak demand times, which can both optimize local grid flexibility and make it easier to integrate clean technologies into the grid, such as renewables.
AI can also accelerate the innovation cycle of earlier stage clean technologies such as energy storage and fusion. Commonwealth Fusion Systems (CFS), based in Massachusetts, recently announced a partnership with Google to utilize their DeepMind AI tool to optimize the performance of CFS's SPARC fusion reactor. By training on the operational scenarios of the reactor which involve complex plasma physics, the AI based learnings can be used to guide SPARC’s control system, making the reactor more cost effective to run.
Overall- AI is a much needed clean technology accelerator
The relationship between AI, data centers and the energy transition is rapidly evolving. The energy demand requirements for data centers are significant, and that demand must be met affordably and reliably. Optimistically, data center companies are at the forefront of creating the demand signal that many low carbon technology providers need to grow to full-scale. AI itself is also proving to be a powerful accelerant in bringing clean technologies and practices to the market sooner. We will all want to watch closely as the situation evolves.