Episode 2211 - July 2 - Tiếng Anh - Phỏng vấn với Brian Janous - Vina Technology at AI time - Lê Quang Văn | Podcast - Nhac.vn

Interview with Brian Janous

Brian Janous is Co-founder of Cloverleaf Infrastructure, which develops strategies to help utilities unlock new grid capacity. Previously, he was Vice President of Energy at Microsoft. Below, he argues that US power infrastructure is not prepared for the coming surge in power demand from AI and other sources, setting up for a painful power crunch in the coming years.

The views stated herein are those of the interviewee and do not necessarily reflect those of Goldman Sachs.

Jenny Grimberg: Power demand is surging across parts of the US, and utilities and grid operators have significantly raised their estimates of US electricity demand growth over the next five years. What role are advances in AI technology playing in the US’ growing hunger for electricity, and how do data centers fit into that?

Brian Janous: Cloud data centers have grown rapidly since the advent of cloud computing around 2010. However, global data center electricity consumption barely budged over the subsequent decade as these data centers cannibalized on-prem workloads, which used multiples more electricity per unit of compute than cloud data centers. So, the migration of data to the cloud resulted in a significant increase in computation with almost no rise in electricity usage. But as the cloud data center capacity of the three large hyperscalers of Microsoft, Amazon, and Google grew from a few hundred megawatts in the early 2010s to a few gigawatts by the end of the decade, power consumption began to rise. And the release of ChatGPT 3.5 in November 2022 ushered in a new layer of AI-related demand, which will likely require adding hundreds of megawatts—if not gigawatts—of data center capacity annually. So, power demand is set to continue surging over the coming years.

Jenny Grimberg: How much does electric grid capacity have to expand to meet this surge?

Brian Janous: That’s the million-dollar question. Utilities are fielding hundreds of requests for huge amounts of power as everyone chases the AI wave, but only a fraction of that demand will ultimately be realized. AEP, one of the largest US electric utility companies, has reportedly received 80-90 gigawatts (GW) of load requests. Only 15 GW of that is likely real because many of the AI projects that companies are currently envisioning will never actually see the light of day.

But 15 GW is still massive given that AEP currently owns/operates around 23 GW of generating capacity in the US. And even if overall grid capacity grows by only 2% annually— which seems like a reasonable forecast—utilities would still need to add well in excess of 100 GW of peak capacity to a system that currently handles around 800 GW at peak. The increase in power demand will also likely be hyperlocalized, with Northern Virginia, for example, potentially requiring a doubling of grid capacity over the next decade given the concentration of data centers in the area. So, grid capacity will need to expand substantially across the US, and likely even more in certain regions.

Jenny Grimberg: Are utility companies and the underlying power infrastructure equipped to meet the rapid surge in power demand?

Brian Janous: No. Utilities have not experienced a period of load growth in almost two decades and are not prepared for— or even capable of matching—the speed at which AI technology is developing. Only six months elapsed between the release of ChatGPT 3.5 and ChatGPT 4.0, which featured a massive improvement in capabilities. But the amount of time required to build the power infrastructure to support such improvements is measured in years. And AI technology isn’t developing in a vacuum—electrification of transportation and buildings, onshoring of manufacturing driven partly by the Inflation Reduction Act and CHIPS Act, and potential development of a hydrogen economy are also increasing the demands on an already aged power grid.

Regulatory lags and interconnection and supply chain constraints are also impediments