Energy for the AI Infrastructure Boom I didn’t use AI to…
0 reactions
Energy for the AI Infrastructure Boom
I didn’t use AI to generate the text of this post, but I did use it for research.
My premise is pretty simple: The only way large-scale data centers are going to get built is if they bring their own power. I see four main energy sources to power the AI boom:
1. ~~Natural gas turbines~~ (discussed at end)
2. ~~Diesel generators~~ (discussed at end)
3. Fuel cells
4. Solar + batteries
I think natural gas turbines and diesel generators will not supply power to data centers for reasons discussed at the bottom of this post. Fuel cells are a great option, but they are dramatically supply constrained and rely on a single company. Solar with battery storage sounds far-fetched until you see that they meet or exceed on almost every metric (except land area).
Data from ERCOT https://www.ercot.com/gridmktinfo/dashboards/fuelmix shows significant penetration of solar batteries into the Texas electricity market.
* Fuel Cells ($BE https://stocktwits.com/symbol/BE ) -
I think this sector is overhyped and overbought.
* Fuel cells are quieter and pollute less than natural gas turbines. They require pipeline connections, which limits their siting. Bloom Energy is the dominant company in this market, and it’s not even close. $BE’s forward EV/EBITA is 83x, but they have big increases in operating revenue to show for it. $BE is expanding manufacturing capacity, but the waitlist for new orders stretches into 2029. $BE estimates its fuel cells generate electricity from natural gas at for \~10¢/kWh.
* Solar + Batteries ($MWH https://stocktwits.com/symbol/MWH ), ($PWR https://stocktwits.com/symbol/PWR ), ($MTZ https://stocktwits.com/symbol/MTZ ) - I believe the AI boom will boost this sector for three reasons: 1) speed to power; 2) siting flexibility; 3) cost. These companies don’t make the solar panels or batteries they install; they install what is available and that speeds up projects. Because they don’t require natural gas pipelines, these projects and their data centers can be sited in distant locations with less local opposition. $MWH estimates its solar + battery systems deliver power for \~4-8¢/kWh, the other companies' estimates top out at 11¢/kWh . These companies are all cheaper on a EV/EBITDA basis than $BE ($MWH is 16x, $PWR is 24x, and MTZ is 14x). The biggest constraint to these projects is land area, which they require the most of compared to all other energy sources. A second barrier is higher upfront cost (considerable in a high interest rate environment), which is mitigated by lower operating costs and uncertainty.
Here are my reasons for skepticism on natural gas and diesel:
* Natural Gas Turbines ($GEV https://stocktwits.com/symbol/GEV ), ($CAT https://stocktwits.com/symbol/CAT ) - Natural gas is cheap and plentiful in the U.S., but is dependent on a pipeline connection (available, but this constrains geography) as well as gas turbine availability (constrained) and air quality permits (likely waived at the federal level, and possibly waived at the state/local level).
* Diesel Generators ($CMI https://stocktwits.com/symbol/CMI ), ($GNRC https://stocktwits.com/symbol/GNRC ), ($CAT) - Diesel fuel is not cheap, and may not be for a long time. Like natural gas, diesel is dependent on equipment availability and air quality permits. A third constraint is noise, which will make local permitting even more difficult.
Let me know what you think.
#industry
source
I didn’t use AI to generate the text of this post, but I did use it for research.
My premise is pretty simple: The only way large-scale data centers are going to get built is if they bring their own power. I see four main energy sources to power the AI boom:
1. ~~Natural gas turbines~~ (discussed at end)
2. ~~Diesel generators~~ (discussed at end)
3. Fuel cells
4. Solar + batteries
I think natural gas turbines and diesel generators will not supply power to data centers for reasons discussed at the bottom of this post. Fuel cells are a great option, but they are dramatically supply constrained and rely on a single company. Solar with battery storage sounds far-fetched until you see that they meet or exceed on almost every metric (except land area).
Data from ERCOT https://www.ercot.com/gridmktinfo/dashboards/fuelmix shows significant penetration of solar batteries into the Texas electricity market.
* Fuel Cells ($BE https://stocktwits.com/symbol/BE ) -
I think this sector is overhyped and overbought.
* Fuel cells are quieter and pollute less than natural gas turbines. They require pipeline connections, which limits their siting. Bloom Energy is the dominant company in this market, and it’s not even close. $BE’s forward EV/EBITA is 83x, but they have big increases in operating revenue to show for it. $BE is expanding manufacturing capacity, but the waitlist for new orders stretches into 2029. $BE estimates its fuel cells generate electricity from natural gas at for \~10¢/kWh.
* Solar + Batteries ($MWH https://stocktwits.com/symbol/MWH ), ($PWR https://stocktwits.com/symbol/PWR ), ($MTZ https://stocktwits.com/symbol/MTZ ) - I believe the AI boom will boost this sector for three reasons: 1) speed to power; 2) siting flexibility; 3) cost. These companies don’t make the solar panels or batteries they install; they install what is available and that speeds up projects. Because they don’t require natural gas pipelines, these projects and their data centers can be sited in distant locations with less local opposition. $MWH estimates its solar + battery systems deliver power for \~4-8¢/kWh, the other companies' estimates top out at 11¢/kWh . These companies are all cheaper on a EV/EBITDA basis than $BE ($MWH is 16x, $PWR is 24x, and MTZ is 14x). The biggest constraint to these projects is land area, which they require the most of compared to all other energy sources. A second barrier is higher upfront cost (considerable in a high interest rate environment), which is mitigated by lower operating costs and uncertainty.
Here are my reasons for skepticism on natural gas and diesel:
* Natural Gas Turbines ($GEV https://stocktwits.com/symbol/GEV ), ($CAT https://stocktwits.com/symbol/CAT ) - Natural gas is cheap and plentiful in the U.S., but is dependent on a pipeline connection (available, but this constrains geography) as well as gas turbine availability (constrained) and air quality permits (likely waived at the federal level, and possibly waived at the state/local level).
* Diesel Generators ($CMI https://stocktwits.com/symbol/CMI ), ($GNRC https://stocktwits.com/symbol/GNRC ), ($CAT) - Diesel fuel is not cheap, and may not be for a long time. Like natural gas, diesel is dependent on equipment availability and air quality permits. A third constraint is noise, which will make local permitting even more difficult.
Let me know what you think.
#industry
source
| 4,000 mlx | |
| 5 | |
| 0 |