The rapid growth of artificial intelligence (A.I.) has created an unprecedented demand for energy, as companies pour huge sums of money into developing and deploying these advanced technologies. Jared Cohen, the President of Global Affairs at Goldman Sachs and former CEO of Jigsaw, discusses the diplomatic challenges and energy requirements surrounding the A.I. boom.
The Challenge of Meeting A.I.'s Energy Demands
Cohen explains that there is a significant "known unknown" around the energy requirements of generative A.I. models, compared to smaller A.I. models. The question is whether the use cases for these large-scale A.I. systems will justify the immense energy expenditure required to power them.
The more pressing issue, according to Cohen, is whether the United States has the capacity to meet the growing demand for energy to power A.I. workloads. The U.S. does not have enough powered land, data centers, or base load power to exclusively meet this demand. This is a global challenge, as many other countries face similar constraints.
Choosing Locations for A.I. Data Centers
Cohen outlines three potential options for where to locate the data centers required to power A.I. systems:
Keeping them in the "democratic world": Countries like Canada, Australia, the Nordic countries, and France, which are well-suited for liquid cooling.
Placing them in the "global south": Countries like Malaysia and Indonesia, where energy is cheap, but this raises national security concerns as the power could go to China.
Locating them in the Middle East: These countries have cheap energy, the ability to build massive infrastructure at scale, and access to the sea for liquid cooling. However, they are geopolitical "swing states," so keeping them aligned with the U.S. is a challenge.
The Role of Export Controls and Geopolitical Considerations
Cohen explains that the U.S. government's export controls on critical technologies, such as advanced chips, give it significant leverage in determining which countries can access these capabilities. This is a key policy issue that the new administration will have to navigate.
The "geopolitical swing states" that Cohen refers to are countries that benefit from sustained tensions between the U.S. and China, as they have a differentiated position in the supply chain. These countries cannot simply do as they please when it comes to A.I., as they ultimately need the U.S. government's approval to access the necessary technology.
The Challenges of Balancing Energy Needs and ESG Concerns
Cohen acknowledges the inherent contradiction between the energy-intensive nature of A.I. workloads and the environmental, social, and governance (ESG) goals that many countries and companies have adopted. The need for base load power, such as from fossil fuels, to power A.I. data centers conflicts with the push for renewable energy sources.
This dilemma is not unique to A.I.; it also applies to the production of critical minerals required for electric vehicles, where the "dirty" processing often takes place in China. Cohen suggests that there is no easy solution to this challenge, as the urgent need to meet the energy demands of A.I. cannot be easily reconciled with the long-term sustainability goals.
The Importance of Domestic Capacity and Geopolitical Alignment
Cohen emphasizes that the U.S. government's export controls give it significant leverage in determining which countries can access the critical technologies required for A.I. development. This means that the geopolitical alignment of countries hosting A.I. data centers is of paramount importance.
While the U.S. has the potential to meet some of the energy demands domestically, the scale of the challenge is so urgent that it will likely require a global approach. Balancing the energy needs of A.I. with ESG concerns is a complex issue that will require careful policy decisions and diplomatic maneuvering in the years to come.
Part 1/6:
The Urgent Need for Energy to Power the A.I. Boom
The rapid growth of artificial intelligence (A.I.) has created an unprecedented demand for energy, as companies pour huge sums of money into developing and deploying these advanced technologies. Jared Cohen, the President of Global Affairs at Goldman Sachs and former CEO of Jigsaw, discusses the diplomatic challenges and energy requirements surrounding the A.I. boom.
The Challenge of Meeting A.I.'s Energy Demands
Cohen explains that there is a significant "known unknown" around the energy requirements of generative A.I. models, compared to smaller A.I. models. The question is whether the use cases for these large-scale A.I. systems will justify the immense energy expenditure required to power them.
Part 2/6:
The more pressing issue, according to Cohen, is whether the United States has the capacity to meet the growing demand for energy to power A.I. workloads. The U.S. does not have enough powered land, data centers, or base load power to exclusively meet this demand. This is a global challenge, as many other countries face similar constraints.
Choosing Locations for A.I. Data Centers
Cohen outlines three potential options for where to locate the data centers required to power A.I. systems:
Part 3/6:
Placing them in the "global south": Countries like Malaysia and Indonesia, where energy is cheap, but this raises national security concerns as the power could go to China.
Locating them in the Middle East: These countries have cheap energy, the ability to build massive infrastructure at scale, and access to the sea for liquid cooling. However, they are geopolitical "swing states," so keeping them aligned with the U.S. is a challenge.
The Role of Export Controls and Geopolitical Considerations
Cohen explains that the U.S. government's export controls on critical technologies, such as advanced chips, give it significant leverage in determining which countries can access these capabilities. This is a key policy issue that the new administration will have to navigate.
Part 4/6:
The "geopolitical swing states" that Cohen refers to are countries that benefit from sustained tensions between the U.S. and China, as they have a differentiated position in the supply chain. These countries cannot simply do as they please when it comes to A.I., as they ultimately need the U.S. government's approval to access the necessary technology.
The Challenges of Balancing Energy Needs and ESG Concerns
Cohen acknowledges the inherent contradiction between the energy-intensive nature of A.I. workloads and the environmental, social, and governance (ESG) goals that many countries and companies have adopted. The need for base load power, such as from fossil fuels, to power A.I. data centers conflicts with the push for renewable energy sources.
Part 5/6:
This dilemma is not unique to A.I.; it also applies to the production of critical minerals required for electric vehicles, where the "dirty" processing often takes place in China. Cohen suggests that there is no easy solution to this challenge, as the urgent need to meet the energy demands of A.I. cannot be easily reconciled with the long-term sustainability goals.
The Importance of Domestic Capacity and Geopolitical Alignment
Cohen emphasizes that the U.S. government's export controls give it significant leverage in determining which countries can access the critical technologies required for A.I. development. This means that the geopolitical alignment of countries hosting A.I. data centers is of paramount importance.
Part 6/6:
While the U.S. has the potential to meet some of the energy demands domestically, the scale of the challenge is so urgent that it will likely require a global approach. Balancing the energy needs of A.I. with ESG concerns is a complex issue that will require careful policy decisions and diplomatic maneuvering in the years to come.