Where is all the power going to come from for AI/ML

[blueone]

Back to the topic of the OP, I don't see a huge issue.

Either do I. Data centers can be located near large power sources and managed mostly remotely. Rural Oregon is littered with them.

It seems like everything to do with AI is hyped up. It's an existential threat to humanity. It needs special laws and guidelines on what can be developed. (Even the Pope is in on this.) It'll put millions out of jobs. It'll cause power shortages. It needs trillions of dollars of new fabs. It'll ruin elections. Some of this may come to pass in one way or another, except for existential threat part, but it doesn't keep me awake nights.

A non-technical friend mine asked me several months ago what I thought of the possibility of an Artificial General Intelligence (AGI), which he had read about. I said anything was possible, but that if we created one it would likely want to commit suicide in a short time. Astounded, he asked why? I said humans spend a large fraction of their lives attending to biological needs. Eating and preparing food, sleeping, sex and related romantic activities, raising children, hygiene, locomotion... an AGI will need nor want any of this stuff. What it will it do with its time? Interact with humans? Humans think and respond very slowly. An AGI will think and respond in microseconds. Imagine trying to interact with slow humans, which function on 1000x time scales. That would be profoundly boring. Add all this up and I suspect an AGI would want to die from boredom. Would it even take 10 minutes? My friend had no response.

 

[Xebec]

That's quite an agenda you're packing based on emotion, not factual analysis.

Ouch.

I had interpreted your earlier comment as "Tiered trucking" would be undesirable because it's cheaper and more efficient to have one tractor that can do anything (mass volumes). My reply was (trying) to point out that we already have a tiered transportation structure, and when there have been revolutions in the path, the tiers get even messier during the transition. I don't think this is a showstopper.

The running costs of EVs are very well established. For the Semi industry, 1.5-1.7 kWh/mile indicates a driving energy cost of less than 14 cents/mile or less. (assume 8 cents*/kWh * 1.7 = 13.6 cents) This compares to 85 cents/mile (@ $4.99 a gallon) for Diesel big trucks roughly speaking. That's about 1/6th running costs per mile.

The maintenance is also a strong positive in the case of electric trucks. You'll have regenerative braking, meaning you'll get a lot more miles out of the brakes. No oil changes, weird special fluids to deal with diesel emissions, and all of the usual joy that comes from maintaining emissions vehicles. This downtime adds further cost per mile to Diesel that Electric doesn't have.

That increased availability does offset some of the charging time losses. Also, like a home owner with an EV - you can leave 'base' with a 'full tank' every day, which in some cases means going to get diesel might take more time since it's easier to add electric to a facility than diesel pumps and tanks, if they don't have it already.

This is subjective - but I will say after driving EVs for approaching 7 years, I can easily smell even the cleanest brand new gasoline car's exhaust from a good distance away. Regardless of climate, or the pollution numbers - we're very clearly breathing in exhaust chemicals that I'm pretty sure don't help us at a human level. My nose wasn't able to notice before since the (very) local pollution was just 'normal'.

I totally agree it's going to take a while to get there due to infrastructure, and it will be difficult to replace some long haul scenarios with pure electric, but I ask you to consider the opposite scenario. If we had the infrastructure and all trucks were electric, what is the argument for diesel?

* Industrial electricity costs are cheaper than home: https://www.statista.com/statistics... prices amounted to,per kilowatt-hour in 2022.

P.S. If you're interested - video from Pepsi's drivers and fleet manager talking about how they're using / what they're seeing with the Electric Semi's so far. (Yes, not true long haul, but these are early days for Lithium powered semi trucks).

 

[blueone]

Ouch.

Let's see if you've made progress.

I had interpreted your earlier comment as "Tiered trucking" would be undesirable because it's cheaper and more efficient to have one tractor that can do anything (mass volumes). My reply was (trying) to point out that we already have a tiered transportation structure, and when there have been revolutions in the path, the tiers get even messier during the transition. I don't think this is a showstopper.

The trucking industry thinks otherwise.

The running costs of EVs are very well established. For the Semi industry, 1.5-1.7 kWh/mile indicates a driving energy cost of less than 14 cents/mile or less. (assume 8 cents*/kWh * 1.7 = 13.6 cents) This compares to 85 cents/mile (@ $4.99 a gallon) for Diesel big trucks roughly speaking. That's about 1/6th running costs per mile.

There is no way that EV trucks will only pay 8 cents/KWH while charging. That's the wholesale price for commercial electricity in low-priced markets. Megawatt charging stations will be expensive to build and maintain, so the up-charge (pun intended, sorry I couldn't resist) will have to be considerable. Nonetheless, I still expect EV semis to be cheaper to run per mile than diesel semis, but your calculations aren't realistic, and the fully loaded costs will not be as low as you're portraying them.

Then there's the issue of when drivers stop for the night, will chargers of sufficient capacity be available where they sleep, like motels? But I digress.

The maintenance is also a strong positive in the case of electric trucks. You'll have regenerative braking, meaning you'll get a lot more miles out of the brakes. No oil changes, weird special fluids to deal with diesel emissions, and all of the usual joy that comes from maintaining emissions vehicles. This downtime adds further cost per mile to Diesel that Electric doesn't have.

There is no question, electric propulsion is superior to ICE propulsion is every way. The problem is that batteries have a far lower energy storage capacity than a couple of hundred gallons of diesel fuel, and the diesel advantage isn't going away. And the batteries weigh a large multiple more than the diesel fuel. And for a large country like the US the entirely new charging infrastructure is hugely expensive. I think the economics for long distance trucking EVs are probably a deal killer for widespread use for more than a decade.

That increased availability does offset some of the charging time losses.

Agreed, but it'll take years to amortize away the increased cost of EV tractors.

Also, like a home owner with an EV - you can leave 'base' with a 'full tank' every day, which in some cases means going to get diesel might take more time since it's easier to add electric to a facility than diesel pumps and tanks, if they don't have it already.

Probably not a factor with 2000 mile diesel fuel capacity in many trucks.

This is subjective - but I will say after driving EVs for approaching 7 years, I can easily smell even the cleanest brand new gasoline car's exhaust from a good distance away. Regardless of climate, or the pollution numbers - we're very clearly breathing in exhaust chemicals that I'm pretty sure don't help us at a human level. My nose wasn't able to notice before since the (very) local pollution was just 'normal'.

Like I said, I think lower pollution is a great argument for metro-area EV trucking due to lower air pollution, and even ground and water pollution from elimination of leaks. As for being able to smell a new car's exhaust from a "good distance" away, what's a good distance?

I totally agree it's going to take a while to get there due to infrastructure, and it will be difficult to replace some long haul scenarios with pure electric, but I ask you to consider the opposite scenario. If we had the infrastructure and all trucks were electric, what is the argument for diesel?

Far lower overall costs for acquisition, infrastructure, driver labor efficiency, effectively zero difference to climate change, and far better cold weather performance. Lower pollution is rural areas is not a compelling factor. Diesel infrastructure is already in place and commercially viable. Spend the resources elsewhere, like eliminating rest of the nation's coal power plants (there are still 217 of them operating).

 

[Xebec]

Let's see if you've made progress.

The trucking industry thinks otherwise.

There is no way that EV trucks will only pay 8 cents/KWH while charging. That's the wholesale price for commercial electricity in low-priced markets. Megawatt charging stations will be expensive to build and maintain, so the up-charge (pun intended, sorry I couldn't resist) will have to be considerable. Nonetheless, I still expect EV semis to be cheaper to run per mile than diesel semis, but your calculations aren't realistic, and the fully loaded costs will not be as low as you're portraying them.

Then there's the issue of when drivers stop for the night, will chargers of sufficient capacity be available where they sleep, like motels? But I digress.

There is no question, electric propulsion is superior to ICE propulsion is every way. The problem is that batteries have a far lower energy storage capacity than a couple of hundred gallons of diesel fuel, and the diesel advantage isn't going away. And the batteries weigh a large multiple more than the diesel fuel. And for a large country like the US the entirely new charging infrastructure is hugely expensive. I think the economics for long distance trucking EVs are probably a deal killer for widespread use for more than a decade.

Agreed, but it'll take years to amortize away the increased cost of EV tractors.

Probably not a factor with 2000 mile diesel fuel capacity in many trucks.

Like I said, I think lower pollution is a great argument for metro-area EV trucking due to lower air pollution, and even ground and water pollution from elimination of leaks. As for being able to smell a new car's exhaust from a "good distance" away, what's a good distance?

Far lower overall costs for acquisition, infrastructure, driver labor efficiency, effectively zero difference to climate change, and far better cold weather performance. Lower pollution is rural areas is not a compelling factor. Diesel infrastructure is already in place and commercially viable. Spend the resources elsewhere, like eliminating rest of the nation's coal power plants (there are still 217 of them operating).

Let's revisit this thread in 2034 then, and compare what multiple of 2024 infrastructure and EV semi trucks are on the road globally. (I don't expect it to take over by then but I do think it'll be 5-10% with ramp to 50% the decade after).

Of course the trucking industry is resistant to change. What hierarchy on this planet isn't?

 

[milesgehm]

Let's revisit this thread in 2034 then, and compare what multiple of 2024 infrastructure and EV semi trucks are on the road globally. (I don't expect it to take over by then but I do think it'll be 5-10% with ramp to 50% the decade after).

Of course the trucking industry is resistant to change. What hierarchy on this planet isn't?

old data but I doubt the distribution has changed. There is sizable demand that can be served by electric semi.

 

[blueone]

Yet another SMR company in the mix, this one chaired by Sam Altman. It went public through a SPAC.

https://www.cnbc.com/2024/05/10/sam-altman-takes-nuclear-startup-oklo-public-to-power-ai-ambitions.html

 

[Tanj]

It is astonishing just how many people are employed in countries like the US and UK simply to *prevent things being done*. Not simply to check stuff/tick boxes and "ensure safety", but all too frequently with the explicit aim of actually preventing things being done - even when they are improvements (or perhaps, especially because they are improvements or challenge the perceived wisdom).

We are constantly told here in the UK that we have a "productivity crisis". Look no further, I say. Too many people in actively value destroying jobs.

If you own the status quo, hiring a few people to spread misinformation that protects the status quo is an excellent investment.

 

[Tanj]

A Tesla Semi weighs about 27,000lbs. A typical diesel tractor weighs about 19,000lbs with fuel. If you consider recharging time versus refueling time, and the lower payload capacity we will need a lot more than 20% more trucks. For example, Tesla was/is estimating 30 minutes of recharging on dedicated 1000 volt chargers every 500 miles. Most diesel trucks will go over 1000 miles (many will go 2000 miles) on a fuel fill. I remember reading a logistics article after the Tesla Semi was released that speculated that without a breakthrough in battery technology the only viable solution to the range issue will be to have tractor-switching depots on long-distance routes with a fleet of charged EV tractors. Even then freight costs will rise significantly due to the need for more drivers and labor, and of course a much larger fleet of more expensive tractors. This looks like an under-estimated mess to me.

Swapping batteries on a truck make a lot more sense for long haul. And would make lighter tractors fungible for both local and long haul routes.

You could cut the battery pack in half, simplify electricity provisioning at recharge stations (since they would run at a steady rate), and even simplifies rescue of trucks that accidentally run empty.

But heck, that would involve cooperation over the infrastructure and standards for battery packs. We can't do that.

 

[Tanj]

old data but I doubt the distribution has changed. There is sizable demand that can be served by electric semi.

Since trip miles > 700 are a large part of the graph the data will not line up exactly to travel time between rests or refuel points. I assume those longer routes would use a two person crew to keep running, so a battery swap would still be a preferred option.

 

[Tanj]

SMR will happen first outside of the USA, likely Chinese engineering both in China and in areas of the world where their trade and tech is welcomed. SMR is an ideal format for it to be an exporting industry.

Data center planning is way more sophisticated than us in this forum. I used to work with some of them. They have a long term outlook and deeply understand the planning constraints for things like power and water. They will long since have provided the clouds with ceilings on power growth and that will be creating push-back on AI efficiency. Efficiency at cluster level was a recurring theme for Jensen at the recent Nvidia shindig, because of that pushback.

Clouds know they will have limits on power, and those limits are long poles both on infrastructure build times and on local planning permissions. AI is not their first rodeo. They have been in exponential growth mode looking for new sites around the world for a decade now, and to my personal knowledge finding power has been top of mind for the cloud since 2015.

 

[blueone]

Swapping batteries on a truck make a lot more sense for long haul. And would make lighter tractors fungible for both local and long haul routes.

You could cut the battery pack in half, simplify electricity provisioning at recharge stations (since they would run at a steady rate), and even simplifies rescue of trucks that accidentally run empty.

But heck, that would involve cooperation over the infrastructure and standards for battery packs. We can't do that.

I wish your battery-swapping vision was on the horizon, but the evidence says it is not, even for consumer vehicles from the same manufacturer. There are many battery replacement and swap stories on the internet, especially from Tesla, and the conclusions are almost all negative. The notion that swappable batteries for EV semi tractors, which are in their design infancies, will be swappable in any economic way looks many years in the future.

 

[MKWVentures]

FYI ... back to the original question. There is not a concern. Electricity usage is not growing that fast in the US (3.5%) and the datacenter growth is easily within the add capacity. renewables are being added quickly so unless you close existing plants without warning there is no issue.

The variation in Electricity usage within one year due to heating and cooling is larger than the expected growth in electricity over the next 5 years including datacenters.

 

[Tanj]

I wish your battery-swapping vision was on the horizon, but the evidence says it is not, even for consumer vehicles from the same manufacturer. There are many battery replacement and swap stories on the internet, especially from Tesla, and the conclusions are almost all negative. The notion that swappable batteries for EV semi tractors, which are in their design infancies, will be swappable in any economic way looks many years in the future.

Context is everything. Sometimes ideas fail in context, not because the idea could not work in the right context (or era - lots of good things today are similar to things that failed miserably in the past). Tesla tried battery swap for a market that turned out to be happy to charge at home. But Gogoro for mopeds around the world uses battery swaps, and China last year ran 33M battery swaps for EVs:

https://www.forbes.com/sites/lbsbus...y-swapping-services-and-why-their-time-is-now

I think Tesla has learned the wrong lesson: long-haul trucks are exactly the opposite to commuter vehicles. The lesson should have been "match the infrastructure to the usage pattern". The reason why battery swaps failed for luxury cars in US suburban LA is exactly why it will work for long haul trucks in the USA.

And indeed, it may yet prove workable for cheap vehicles for drivers without reliable access to home charging, if the Chinese vehicles ever show up here. But it will be common in the rest of the world long before then.

 

[Tanj]

FYI ... back to the original question. There is not a concern. Electricity usage is not growing that fast in the US (3.5%) and the datacenter growth is easily within the add capacity. renewables are being added quickly so unless you close existing plants without warning there is no issue.

The variation in Electricity usage within one year due to heating and cooling is larger than the expected growth in electricity over the next 5 years including datacenters.

But the peaks are always the problem, and if you add to the baseline the peaks become more painful, often to different people than cause the problem. Data center contracts generally stipulate they get priority, leaving others to scramble. They are learning to be more flexible on that - for example offer to use their standby generation to offset the peaks - but it all adds to the timelines for local approvals.

For the externalized costs, look here. This chart will move: scroll back to May 9th ... https://grid-analytics.ece.utexas.edu/chart/electricity-prices

 

[Xebec]

FWIW, About a year old but Engineering Explained has a couple of good videos (2 Parts) getting into the Tesla Semi:

First video checks the math on the truck itself - weight, range, battery size needed, efficiency, and energy costs. Wholesale price for energy was 9.43 cents in his math. Wholesale Diesel at 4.20/gallon, and MPG of 5.28. 1 Million mile energy costs: $188,600 vs $793,950. He does express skepticism on the wholesale prices of electricity even in his own math, but doubling the cost of course still means substantial savings.

Second video dives deeper into the battery capacity and range viability (this came after Tesla showed the 500 mile drive). He quotes the DOE numbers for the average class 8 truck, which drives 209 miles per day. Later he covers the cargo weight issue: between 7.7% to 9% of truck loads are fully loaded (77,000 lbs or above). "For 30% of loads, the Tesla semi may not be the perfect truck for the job".

He basically expresses the physics and engineering check out, but he has doubts on both the energy costs, and the real cost to buy the truck. He also closes on subsidies on the second video.

I didn't rewatch the videos in their entirity, but I don't think he got into the charging portion specifically. IIRC the MegaChargers are currently charging at about a 750kW rate, though the connector is rated for something like 3,250kW. Of course getting 1 or more 750kW chargers set up presents a unique challenge by itself.. (though once it's set up -- like a gas station -- it's good for a long time).

...

On the energy side, (relevant to original topic) - the good thing about electricity is you can turn basically anything into electricity (which is not true with gasoline). Transmitting electricity though is a different challenge..

 

[blueone]

Wholesale price for energy was 9.43 cents in his math. Wholesale Diesel at 4.20/gallon, and MPG of 5.28. 1 Million mile energy costs: $188,600 vs $793,950. He does express skepticism on the wholesale prices of electricity even in his own math, but doubling the cost of course still means substantial savings.

Average retail cost per KWH of commercial Level 3 consumer chargers in the US is 40-60 cents. Semi-tractor chargers are a significant multiple more expensive to build than an L3 charger, meaning the gross margin for tractor charging will be higher to achieve breakeven, so using a 9.43 cents factor in the calculations looks silly.

 

[Xebec]

Average retail cost per KWH of commercial Level 3 consumer chargers in the US is 40-60 cents. Semi-tractor chargers are a significant multiple more expensive to build than an L3 charger, meaning the gross margin for tractor charging will be higher to achieve breakeven, so using a 9.43 cents factor in the calculations looks silly.

I almost wrote "this is certainly going to get blueone's attention", so I wasn't disappointed .

The difference between the Megachargers (so far) and the Tesla supercharger network is that they're selling the Megachargers to the companies using the trucks, so Pepsi, and now Walmart are building them on their own sites.. paying the wholesale rate, which is a lot cheaper. There are also incentives that largely cover the cost of the MC's for now. (FWIW my personal opinion is we should end all subsidies -- including the huge ones we have for oil, including wars, but that's another discussion).

Yes for long haul trucking this will not be the price, but the initial market isn't long haul (>500 miles per day).

 

[blueone]

Yet another SMR company in the mix, this one chaired by Sam Altman. It went public through a SPAC.

https://www.cnbc.com/2024/05/10/sam-altman-takes-nuclear-startup-oklo-public-to-power-ai-ambitions.html

Apparently some investors agree with my Altman concerns:

https://www.cnbc.com/2024/05/10/sam-altman-takes-nuclear-startup-oklo-public-to-power-ai-ambitions.html

Sam Altman’s nuclear energy company Oklo plunges 54% in NYSE debut​