Regarding distributed nuclear generation, check out VSMR's - Very Small Modular Reactors.
"These reactors are one step down to actual small modular reactors (SMRs) in terms of power and size. SMRs typically have a power range of 20 MWe to 300 MWe, while vSMRs have a power range of 1–20 MWe."
"It can be observed that SMRs can be factory-built with all components integrated into a capsule-like structure that is exactly like vSMR. However, for small-scale applications, vSMRs can be a good choice because they can be easily manufactured and transported."
Great piece Arjun. Off topic question for you. Have you done any work on the role of AI and LLMs on oil and gas exploration (e.g. improved efficiencies in the interpretation of seismic data for example)?
Thank you Blind Squirrel. Yes, that is an area I am spending time on. If you take a basin like the Delaware or Midland in the Permian, there are a lot of zones, a wide swath of areas, many wells drilled, lots of data on all aspects of fracing and drilling a well...no doubt "AI" could have a big impact on future productivity and understanding.
A very timely piece, Arjun! And a prudent Q&A-based approach to the examination of AI opportunities & benefits to society. Three comments/questions:
Full disclosure : I’m from Missouri on AI, and a disciple of Vaclav Smil (VS) on energy and materials: nothing is sustainable at current rate of consumption.
1) I endorse your CH4 warning:” US and Canadian NG producers would be well served to ensure that domestic resources are bulletproof on the methane issue.”
2) on Big Tech’s energy literacy and global energy systems, you are right to credit Meta for adding John Arnold to its ranks. (I know on good authority that a “hooded billionaire” also sought VS council.) But Bill Gates certainly deserves an energy literacy pioneer’s recognition.
3) my two-pronged question:
a) should big tech have to disclose the energy intensity of their ad, AI and search services?
You made it abundantly clear – and you are in good company – that AI will consume gargantuan quantities of energy and water. You further note that “O&G create essentially zero products that use the energy they produce.” But car and appliances manufacturers do and disclose the energy consumption of the vehicles and appliances offered to the public to enable buyers to make informed choices.
The energy consumption of search engines has the allure of a state secret. I am using duckduckgo because it doesn’t track my searches or build a profile based on my history, presumably using less energy. But am I right? An energy consumption label would help.
b) Would an EnergyGPT improve the EROI of the energy system? If AI consumes more energy than it yields, what is the point?
Thank you so much Michel. You pose some great questions. I'd need to give it more thought. It's a good point that on many items, one is forced to disclose or we otherwise know how much energy is used...that is really not the case for any tech product. And the 2nd question on EROI is also excellent.
As always, a great write up, and you have not cursed the AI space by doing some much needed analysis!
How can anyone be so confident in the capex and growth plans of AI infrastructure without solving the power needs??
You are addressing the other side of the equation, and pointing out rightly that it's not a matter of renewables vs nat gas vs nuclear.....if the forecasts are correct, we need it ALL!!
I'm sure it didn't escape your notice that the NY Times and the Washington Post had big pieces on the topic this week.
Very timely to kick off this discussion on the intersection of AI/Power and O&G. A couple of thoughts and questions. 1. Implicit in your commentary is that AI and all things digital are increasing the demand for more data centers. I love the term storage in the cloud but we all know at the end of the save button, there is a hardstack somewhere in Virginia whirring away for the next time you need that datapoint. Multiple studies show large increases in power needs across the country from industrial activity including data centers. We are at an interesting time in the utility sector for this surge in demand. Vast parts of the distribution and transmission infrastructure need replacement with assets decades past their sell by date. Reliability and affordability are lighting rod issues for the retail customer like you and I. As big tech and industrial customers move quickly to lock up "green" and reliable power, this puts more pressure on capacity to keep the lights on for the average voting public. I agree we will need more generation to come on the market but who should pay? If Big Tech Co XYZ locks up a 15 PPA at an attractive price, requiring new build generation by the regulated side of the house, should some of that be pushed back to industrial contracts? This is a big policy issue we will face. 2. On the ESG and sustainability point. I am hopeful that the convergence of AI/Tech and Energy will move us away from Carbon Tunnel vision when it comes to the sustainability discussion. Water intensity is a big issue, biodiversity, community, profitability, etc.. all need to figure into the definition of what constitutes sustainability for an enterprise. I think we will see this push the discussion past GHG emissions into a more holistic analysis. Look forward to more like this from you!
Deborah, all great thoughts and perspectives. Thank you for sharing. I especially like the "Carbon Tunnel vision" branding...I am going to ask you to borrow it!
Brilliant as always Arjun. If I was at a cocktail party, dinner or on the golf course with you, I would ask you to describe the “pushback” that you must get from Super-Spiked. Who pushes back and how do they do it? You are so incredibly kind and considerate in your writing, do you get the uncivil response that I imagine? Thank again!
Regarding distributed nuclear generation, check out VSMR's - Very Small Modular Reactors.
"These reactors are one step down to actual small modular reactors (SMRs) in terms of power and size. SMRs typically have a power range of 20 MWe to 300 MWe, while vSMRs have a power range of 1–20 MWe."
"It can be observed that SMRs can be factory-built with all components integrated into a capsule-like structure that is exactly like vSMR. However, for small-scale applications, vSMRs can be a good choice because they can be easily manufactured and transported."
https://www.sciencedirect.com/science/article/pii/S1738573322005824
thanks Robert for sharing link
Excellent as always. Incredible that AI power demand was overlooked.
Thank you Andrew. Always appreciate your support.
Great piece Arjun. Off topic question for you. Have you done any work on the role of AI and LLMs on oil and gas exploration (e.g. improved efficiencies in the interpretation of seismic data for example)?
Thank you Blind Squirrel. Yes, that is an area I am spending time on. If you take a basin like the Delaware or Midland in the Permian, there are a lot of zones, a wide swath of areas, many wells drilled, lots of data on all aspects of fracing and drilling a well...no doubt "AI" could have a big impact on future productivity and understanding.
Look forward to reading that!
A very timely piece, Arjun! And a prudent Q&A-based approach to the examination of AI opportunities & benefits to society. Three comments/questions:
Full disclosure : I’m from Missouri on AI, and a disciple of Vaclav Smil (VS) on energy and materials: nothing is sustainable at current rate of consumption.
1) I endorse your CH4 warning:” US and Canadian NG producers would be well served to ensure that domestic resources are bulletproof on the methane issue.”
2) on Big Tech’s energy literacy and global energy systems, you are right to credit Meta for adding John Arnold to its ranks. (I know on good authority that a “hooded billionaire” also sought VS council.) But Bill Gates certainly deserves an energy literacy pioneer’s recognition.
3) my two-pronged question:
a) should big tech have to disclose the energy intensity of their ad, AI and search services?
You made it abundantly clear – and you are in good company – that AI will consume gargantuan quantities of energy and water. You further note that “O&G create essentially zero products that use the energy they produce.” But car and appliances manufacturers do and disclose the energy consumption of the vehicles and appliances offered to the public to enable buyers to make informed choices.
The energy consumption of search engines has the allure of a state secret. I am using duckduckgo because it doesn’t track my searches or build a profile based on my history, presumably using less energy. But am I right? An energy consumption label would help.
b) Would an EnergyGPT improve the EROI of the energy system? If AI consumes more energy than it yields, what is the point?
Thank you so much Michel. You pose some great questions. I'd need to give it more thought. It's a good point that on many items, one is forced to disclose or we otherwise know how much energy is used...that is really not the case for any tech product. And the 2nd question on EROI is also excellent.
As always, a great write up, and you have not cursed the AI space by doing some much needed analysis!
How can anyone be so confident in the capex and growth plans of AI infrastructure without solving the power needs??
You are addressing the other side of the equation, and pointing out rightly that it's not a matter of renewables vs nat gas vs nuclear.....if the forecasts are correct, we need it ALL!!
I'm sure it didn't escape your notice that the NY Times and the Washington Post had big pieces on the topic this week.
Well done sir.
Thank you Jonathan! Very much appreciated.
Very timely to kick off this discussion on the intersection of AI/Power and O&G. A couple of thoughts and questions. 1. Implicit in your commentary is that AI and all things digital are increasing the demand for more data centers. I love the term storage in the cloud but we all know at the end of the save button, there is a hardstack somewhere in Virginia whirring away for the next time you need that datapoint. Multiple studies show large increases in power needs across the country from industrial activity including data centers. We are at an interesting time in the utility sector for this surge in demand. Vast parts of the distribution and transmission infrastructure need replacement with assets decades past their sell by date. Reliability and affordability are lighting rod issues for the retail customer like you and I. As big tech and industrial customers move quickly to lock up "green" and reliable power, this puts more pressure on capacity to keep the lights on for the average voting public. I agree we will need more generation to come on the market but who should pay? If Big Tech Co XYZ locks up a 15 PPA at an attractive price, requiring new build generation by the regulated side of the house, should some of that be pushed back to industrial contracts? This is a big policy issue we will face. 2. On the ESG and sustainability point. I am hopeful that the convergence of AI/Tech and Energy will move us away from Carbon Tunnel vision when it comes to the sustainability discussion. Water intensity is a big issue, biodiversity, community, profitability, etc.. all need to figure into the definition of what constitutes sustainability for an enterprise. I think we will see this push the discussion past GHG emissions into a more holistic analysis. Look forward to more like this from you!
Deborah, all great thoughts and perspectives. Thank you for sharing. I especially like the "Carbon Tunnel vision" branding...I am going to ask you to borrow it!
Brilliant as always Arjun. If I was at a cocktail party, dinner or on the golf course with you, I would ask you to describe the “pushback” that you must get from Super-Spiked. Who pushes back and how do they do it? You are so incredibly kind and considerate in your writing, do you get the uncivil response that I imagine? Thank again!
Thanks so much Martin! I always appreciate your support. I will make your request a future issue of Super-Spiked.