So often in work I want to give clear short answers. But the context of engineering is often so complex with many subtleties to be considered that it is a near impossibility to be as succinct as I wish.
@dwarkesh_sp It seem like intelligent and power are things that can achieve goals.
You can attain more power with intelligence. If you have a lot of power you can coerce intelligent people to work on your behalf.
@Stphn_Lacey That's right, otherwise second order infastructure costs are placed driven by a new system peak load which does get passed on to rate payers.
This is politically fraught. It seems good for utilities to be load agnostic and simply treat all customers the same. Particularly given thier monopoly status.
Some common sense carve outs make sense, such as never tripping a hospital before a suburb. But keeping this life and death should be where the line is.
The cost of operating these microgrids will be so high. Part of what the utilities bring is institutional knowledge of operating large scale power systems.
When I worked at a utility, as a courtesy we would solve engineer problems for our large power customers.
If you have a microgrid you are on your own.
Yes exactly.
The only way data centers increase cost would be if they forced new infrastructure to be built because they push the peak demand of the system up causing overloading of second order systems.
New data center must pay for, first order new infastructure such as building a new substation on site.
But utilities traditionally pay for second order upgradea such as reconductoring a transmission line that becomes overloaded when the new load is served.
If the data center pay for all the infrastructure upgrades, including second order ones than it's no issue. But that would be an unusual treatment of new load.
@ShanuMathew93
Great post.
Adding some subtlety on power for 1GW utility-fed loads: Pure kWh energy rates often drop to 3-7Β’ for steady high load-factor customers.
This is because utilities also charge for power itself in the form of demand charge for industrial power users.
Demand charges can be $10-20+/kW per month.
At 850 MW that adds about $150 million in annual power charges at $15/kW monthly.
Than another $400 millions for energy charges assuming a little over 5 cents per kwh.
Total cost is likely closer to $550 million per year for power/energy.
@parveen__tyagi I've always heard that writing online is one of the best things you can do to increase luck. Opportunities start developing from things you wrote years ago.
Trying to put that into practice.
Directionally I agree.
But 2 point triggered dissonance.
1. It is advantageous for 1 company to own design, construction, and project management. That dramatically lowers coordination costs.
2. Starting to build early, particularly civil works, can be a smart strategy. If you wait for a complete design, all the way through SCADA and more abstract technical design, you may delay the project by years.
@Robotbeat It's good to have the smartest people in the world have to deal with power as a bottleneck on their path to agi.
The attention and investment in power generation is going to do incredible things for the average person a decade from now.
@ChrisBarnardDL I can't get my head around why the market hasn't solved the transformer supply problem. It seems like China would have dozens of mega factories, pumping out these massive transformers and flooding the market.
Are they really more complex than cars for example.
@JigarShahDC Not sure if this is the right solution. But the identification of the problem is insightful. Should the grid treat all loads equally.
The status quo is a market distortion in a way I didn't notice before.
@BooshCarron@ShanuMathew93@xiaowang1984 No, energy is an input to almost everything. No one likes the market forcing higher prices. Low cost energy is a virtue.
@ShanuMathew93 Particularly if the data centers are improving grid infrastructure, adding transmission lines, substations, energy storage, and generation.
This is a boon for the grid.
@saeverley Really great point.
A caveot is that demand for intelligence is infinite. If consumers keep increasing demand, the data center bar may dominate soon.
Still, surprising that it doesn't already dominate electricity growth.