AI Challenging Climate Goals

In This Issue:

• Artificial Intelligence’s rise is not without challenges
• AI showing great promise for the global economy
• AI’s enormous energy appetite is going to challenge climate goals of policymakers
• AI’s future will depend on more energy

One of the hallmarks of the US economy for over a century has been its ability to renew itself through a dynamic process of creative destruction. New companies seemingly come out of nowhere and over time go on to become globally dominant. Whether it was Ford or IBM in the early part of the 20th century or Microsoft, Apple, Google, and Amazon in the 1980s and beyond – entrepreneurial ingenuity has allowed for the founding of entirely new industries. The innovations developed by many great companies helped to provide a continuous supply of fuel for economic growth. In recent years, it has been the arrival of Artificial Intelligence applications that are expected to provide the next great economic leap.

Though the concept of Artificial Intelligence (AI) has been in use since the 1950s, in today’s technologically advanced world it is a term that refers to the ability of machines, guided by computer programs, to simulate human intelligence by not only processing and analysing large amounts of data quickly but also to learn to use the information as a person would while applying a level of reasoning and logic. As AI developments continue, machines will be able to increasingly interact with human language, perform speech recognition, and have vision that would allow for self-driving vehicles, amongst other advancements.

The implications for the future of the global economy are difficult to foresee with certainty but on the positive side, forecasts would indicate that AI will allow for the automation of routine tasks that would allow workers to focus on more complex activities that in turn would enhance productivity. Another example of how AI is being used already is in call centers – where AI powered chatbots and virtual assistants are able to provide faster and more efficient customer service.

Companies such as Amazon are investing huge sums in AI in order detect fraud, improve inventory management and optimize supply chains. On the customer side of its business, it is using AI to enhance its relationship with its customers. Amazon uses AI to analyze the immense data its customer interactions generate each day. It uses this data to proactively reach out to customers with recommendations for products based on past purchases.

As the largest apparel retailer in the world, Amazon uses AI to learn millions of details on apparel items, such as style, size, and customer reviews. Amazon has been able to determine through AI that its customers are more likely to purchase and keep an item when a size or particular attribute is recommended to them. In this way, Amazon is able to reach hundred of millions of customers across the world quickly and cost effectively. From a bottom-line perspective, the impact on profits is positive since AI is able to induce customers to spend more on products that are a better match for their style and size preference. In turn, this minimizes the volume of merchandise that is returned back to Amazon – reducing costs for the company.

While it is too early to say what the future implications will be with any great precision, the potential for AI to be the “next big thing” is certainly possible. But AI is posing some challenges that could limit its growth, at least in the short to intermediate term. One of those challenges relates to generating enough energy to power the incredible growth of AI. Specifically, AI requires a tremendous amount of computational power that is done in data centers.

Data centers are large warehouse sized facilities that house computer systems and communication and storage systems. In turn, these facilities allow for the storing, processing and sharing of data and computer applications across wide areas. Many of the world’s largest institutions and corporations rely on data centers for their day-to-day operations.

Data centers are classified as “critical infrastructure.” If a data centre were to go offline due to a malfunction or due to a power shortage, it can cost companies millions of dollars per minute. Therefore, data centers must be running every minute of every day. To safeguard against this risk, data centers are intertwined with critical support systems (infrastructure).

These data centers require very high levels of power that must be consistent and able to increase in capacity. Not only do the computers require power but the cooling systems that keep the data center from overheating require a strong and consistent supply of power too.

Data Centers’ Risk to Net Zero

A long-standing goal of global policymakers has been to generate a continuous rise in power generation from renewable sources of energy such as wind, solar and hydroelectric. The global community of nations has pushed for “net zero” by 2050 which refers to achieving a balance between the amount of greenhouse gases emitted into the atmosphere and the amount removed from it through carbon capture, storage and reforestation. Some scientists have stated that achieving net zero emissions is essential to limit global temperature rises to levels that avoid the severest consequences of climate change.

The challenge, global electricity consumption has been rising steadily over the past 20 years; driven by population growth, industrialization, urbanization, and technological advances that have led to the proliferation of electronic devices, smart technologies and more recently – the rise of data centers.

Servers require extraordinary amounts of electrical power to perform the technological services that they are designed for, but must be kept at specific temperatures at all times. A site must have access to a power grid that can keep the servers running and the cooling systems working around the clock. Data centers alone are consuming over 3% of global electricity consumption. As the backbone of the digital economy, they play an indispensable role in storing, processing, and delivering the data that powers the economy and all aspects of society. However, this reliance comes with an immense energy cost, making the sustainability and scalability of our energy sources critical.

As the chart above shows, the US has more data centers than any other nation. It houses 5375 data centers – almost 50 percent of the global total. Data centers represent about 4 percent of all of the electricity consumed in the US. By the end of the current decade, data center power consumption in the US is expected to reach 35 gigawatts (GW). This is double the level of power consumption of the level of 2022.

The US data center capacity is centered in Virginia which has the largest concentration of data centers in the world. Earlier this year, Google announced a $1billion expansion of its three data centers in northern Virginia. Virginia’s leading power provider, Dominion Energy, states that 20 percent of its power sales in the state are to data centers and it expects that by 2040, power consumption by data centers will quadruple.

Environmental Constraints

The Virginia chapter of the Sierra Club, one of the largest environmental organizations in the US, is not pleased with this trend. It believes that at the current pace, 40 percent of Dominion Energy’s power generation will be directed to data centers. The Sierra Club believes that to meet demand, new power plants fueled by fossil fuels are going to have to be built with the costs borne by Virginia households. Virginia’s governor, Glenn Youngkin, states that critics of datacenters are not taking into account the over $2 billion in wages paid to Virginia based employees in the data center sector and the over $1 billion in tax revenue that has been generated to fund government services.

Energy Shortfall Hinderance to AI

It is not only in Virginia that the need for data centers and the environment are in conflict. Ireland is facing the same issues. Ireland has 82 data centers in service, with 14 more under construction and another 40 approved for planning. Ireland is a favoured destination of the technology industry because of its favourable taxes, educated workforce and its weather. Specifically, its rainy and often cool weather is helpful for a data center since it lowers the energy costs for cooling systems.

Ireland’s Central Statistics Office states that data centers account for almost 20 percent of its electricity consumption. This is equal to the power consumed by nearly all Irish urban households. The Irish government has stated that it will miss its 2032 greenhouse gas emissions targets by a wide margin. Furthermore, if all of the datacenters that are being proposed for Ireland are built, experts estimate that up to 70 percent of the entire country’s electricity supply will be consumed by datacenters.

Irish energy regulators have pushed back against data center owners such as Microsoft, Amazon and Meta by requiring data centers to have backup generators so that if they are asked to reduce their energy pull from the national grid, they can keep running during times of peak consumption. Without this measure, Ireland would be plunged into rolling blackouts.

Iceland: The Ideal Data Center Host

Iceland has also entered the competition for data centers. Iceland might be the perfect place for a data center given its cool temperatures and cheap and environmentally friendly power supply since all of its power is renewable. Iceland’s president, Guoni Johannesson was quoted in a data center industry publication earlier this year as saying “Data centers are here to stay. When I was studying in England in the late ’80s, there was no Internet. My mom sent me newspapers that arrived 14 days later…Last night, I attended a sporting event. I took loads of videos, listened to Spotify on the way back, downloaded necessary stuff, unnecessary stuff. But I didn’t pay any attention to the energy, because it’s just the cloud – I don’t see smoke billowing up out of my laptop or phone.”

Nuclear Powered Internet

While large economies cannot rely on renewable power the way Iceland does, technology companies and even some environmental groups have stated that nuclear power is the next best solution. It is able to provide a reliable source of power and it does not emit greenhouse gases. Some environmental groups have stated that if greenhouse gas emissions targets are going to be met, then nuclear power has to be part of the solution. The Biden Administration is broadly supportive of nuclear power to meet the energy needs of the US and views nuclear energy as the only proven clean and reliable source of power.

In the short to intermediate term, even if the commitment is made to increase nuclear energy capacity, plants can take over a decade to move from permitting to production. Meanwhile, hundreds of billions of investment in new data centers is being planned and data centers can be up and running in under two years. Given the time gap between the immediate term demands of the AI industry and more reliable power supply coming on stream, it is likely that fossil fuel use will have to rise to fill the gap. This would mean in the US and other nations, natural gas demand  for power plants would rise. Currently, about 43 percent of US electricity demand is met by natural gas plants.

A further limitation on energy options comes from the fact that under-investment in nuclear plants has caused the US  and western nations to now lag China by 10 to 15 years in rolling out next generation nuclear reactors.

If the AI boom is to be sustained and enhanced and in turn provide the benefits being touted for commerce, science and the broader economy, then energy production will have to rise. The future growth of AI is going to have to require a strong contribution from old economy sectors such as uranium, natural gas and utilities. In many parts of the US and around the world, data centers are putting a strain on power grids. Even a relatively simple AI application such as ChatGPT requires about ten times the energy for a query that Google needs in a traditional non-AI search.

It is perhaps ironic that as world leaders finally came to a consensus on the need to reduce greenhouse gas emissions, they are being confronted with the potential for their economies to face severe energy shortages from advances in computing technology. Under that scenario, they would likely have to consider abandoning net-zero emissions goals.