The large transformer shortage has been a problem for years. Large transformer making is a craft, where the winding supports are made of hardwood, like furniture, and wound by hand. Then the windings go into a case that's an oil tank.
The build teams aren't that big - 30-50 people. The main barrier to entry is that it takes people who know how to hand-build big transformers. Utility buyers want a supplier who's going to be around half a century from now, since these things last that long.
Here's a summary of the market, from a transformer maker in China.[1]
Here's an AI-generated fake video of large transformer manufacturing. It's about half wrong.[2] But right enough to be worth watching. I'd like to see the prompts for this.
Virginia Transformer is the US's biggest maker of large transformers.[3] They advertise their "short lead times" of two years. The margins are low, and makers don't want to go idle between orders. This is a problem with much heavy machinery. It could be built faster, but when you catch up, everybody gets laid off and the factory sits idle. There goes your profit margin.
That's a good point. The AI slop video is inaccurate, but entertaining.
For comparison, here's the real deal - transformer winding at Virginia Transformer in the US.[1]
That video provides a good sense of why these things take so long to make.
All those wooden parts. All that slowly and carefully hand wound heavy wire. As they point out, if that wire can move at all, as the magnetic fields pushes and pulls on it, the vibration will, over time, wear out the transformer. It's a very fussy job to get the position and tension right, with wire firmly supported against movement in all directions. That's the difference between a lifetime of a few years and many decades.
It's a boring video.
Here's the whole manufacturing process at ETD in the Czech Republic.[2]
This shows roughly the same sequence of steps as the fake video, but it's real.
Big industrial bay with lots of transformers and overhead cranes. Sheets of lamination steel.
Winding. The moving and shipping of the big transformer.
All that is in both the real video and the AI slop.
This is the real video from the manufacturer, and it assumes that if you're watching, you know what you're looking at. There's little narration.
It's a confusing video.
Here's a small open frame transformer.[3] If you've done much electrical or electronics work, you've seen one, and may have replaced or installed one. When you see the big ones being built, the process makes sense. Same concept, with a laminated core, windings, insulation, and lead wires. The big ones have the same key parts, just much bigger. But if you don't know a transformer from a transistor, the manufacturer videos are just wallpaper.
And there's the problem. The AI slop version will give the average viewer a general idea of the process. The accurate videos from manufacturers require more background knowledge to comprehend.
The target audience is different. The manufacturers don't make those videos for the general public.
As far as I can tell, the video itself is 100% fake. A bad fake. I particularly love the part where the worker levitates a large coil of steel with his hands. The narration sounds OK, so just turn off your monitor.
It's a generic problem with flat demand in heavy industry. Shipbuilding, bridges, nuclear reactors - when the production backlog runs down and the factory goes idle, the factory dies. So do the companies that feed specialized parts into the process.
This would be a great opportunity for the government to get involved.. Tell them to just make two of every order they have now and the government will buy the second one at whatever price the customer is paying. Put the spares in a strategic repository and sell them at “cost” to whoever wants them. Would be a much better use of a few billion dollars than some asinine Star Wars II or another half a trillion into the war maw.
> You'd think if it's causing this much of a problem, there would be money available.
There’s plenty of economic solutions if companies are really that desperate. They can pay a premium to encourage more investment. They can invest themselves, or enter into partnerships, acquire their suppliers or even open their own facilities.
Companies often complain about shortages, but it usually comes with the caveat ‘at the price we’re willing to pay’
I worked in a machine shop of a large shipyard and 20 years ago they had CNC Winders for rebuilding armatures for Navy ships.
They replaced older versions that were NC winders.
So this "hand wound" story is just that.
Windings going into Oil Tank? I think you mean varnish tank... After the rotating assembly is balanced they go into a protective coating tank that is a varnish.
Afaik utility scale Transformers operate in cooling tanks. Maybe that's what they meant about "afterward" "oil" covers many things. I believe the constant cycling and thermal load can make heinous PCBs.
You can’t “join up transformers”, a single-phase transformer is an iron core with two sets of copper or aluminum windings around it. As far as I know, the only way to increase the volt-amp rating is more iron and copper/aluminum. Wiring them in series doesn’t increase the volt-amp rating but you can use (3) paralleled single-phase transformers for a three-phase circuit, I occasionally see a set of three medium-voltage to 480V ‘pineapple’ transformers (with visible live parts!) at the service entrance for older buildings.
Why can’t we mass manufacture aircraft carriers by making a lot of small boats and joining them up?
It’s the same kind of problem.
(And notably, it’s not that it’s actually completely impossible to do it that way - just impractical compared to the alternative. You could actually make something that kind of sorta worked for an aircraft carrier by joining tens of thousands of small pontoons and support ships. Operationally, it would just suck compared to the alternative.)
> Here's an AI-generated fake video of large transformer manufacturing. It's about half wrong.[2] But right enough to be worth watching. I'd like to see the prompts for this.
I'd like some sort of shared blocklist support for YouTube and Instagram. I'm sick and tired of content thieves and AI slop farms.
> So how did we get to a point where one component can hold trillion-dollar industries hostage? Turns out, a quirk of history made the entire world’s electricity systems reliant on transformers.
> At the end of the 19th century, when electricity was just starting to become a commercial source of energy, two businessmen fought to control its future in what came to be known as “the war of the currents.” Thomas Edison promoted the use of direct current (DC) and George Westinghouse, inventor and industrialist, was convinced that alternating current (AC) would prove more practical.
> In a clash of personality, finance and some genuine technical advantages, Westinghouse won out and the world has been mostly stuck with using AC as a means of generating and transmitting electricity. Transformers are necessary to make the AC system work.
This entire section is a glaring load of nonsense and needs to be removed. We had to start with AC for a variety of technical reasons, the main one being that boosting DC voltage pre-switching technology was impossible. DC cant pass through a transformer unless it is converted to some form of AC, usually in the form of PWM square waves these days. Before the invention of the mercury arc rectifier (And later valve) in 1902 you had boost DC using mechanical methods: generators. The problem there is physical, they did not have the ability to insulate the generator windings at high voltage potentials. They also had problems with DC voltages over 2000 volts on commutators [1] citing excessive arcing. Commutators are also a limiting factor in machine size as beyond several MW they dissipate too much power. So with all this the highest practical voltage for a DC grid using early electrical machinery is around 2 kV. Now imagine all that mechanical complexity on the distribution end. Meanwhile, early AC transmission was already in the tens of kilovolts: 11/22/33 kV (multiples of the early Edison 110 volt standard.)
As for the whole war of currents, I feel it is vastly overstated and was more a public spectacle than serious scientific dispute. It was already known from early on that AC was the future thanks to its ability to easily be transformed to higher voltages for transmission and back again with no moving parts. The "war" was likely Edison marketing to sell off the remaining inventory less desirable DC machinery.
Yes this is the most glaring issue. There also two disconnects later in the article: at the end it laments how china has been increasing transformer manufacturing but the US government has done nothing. Then in the next sentence its mentions trumps tariffs have increased transformer costs, I. E. Government action to increase domestic production. It also glosses over the new DOE rule on how transformers are made…so maybe there is a larger story there relevant to the lack of supply.
Tariffs don't help onshore manufacturing when they apply to the materials that the manufacturing needs and might evaporate before the manufacturing capability is actually created. Tariffs needs to be applied carefully and consistently to actually encourage this.
We had targeted policies under Biden to increase US production of grid components. This entailed invoking the DPA and setting aside millions for manufacturing improvements. Trump paused all that and created blanket tariffs that don’t seem like they’re designed to onshore US manufacturing of these very specific components but do increase all the material costs. This is not an easy thing to fix with dumb tariffs, and it’s really easy to make everything worse.
Also it is quite a leap to the conclusion that HVDC eqipment will not have long lead times as well especially since there are quite a few less companies making it.
The early limit was because high voltage DC required producing it at the generator, whereas you could produce high voltage AC by generating at a lower voltage and then stepping it up with a transformer for long distance transmission.
The rules are changing because of switchmode voltage conversion, using transistors to switch the voltage at a high frequency, where the magnetics (transformers, inductors) can be much smaller and more efficient, then converting back to DC. This is how virtually all smaller power supplies have been made for years, the only question (which I don't know) being how far along we are at reaching the voltage levels of long distance transmission in this way.
I'd think that hustling us towards DC with electronic voltage conversion would be a reasonable strategic goal for dealing with the transformer problem, worthy of support by a government.
That link talks about 5MW 35kv AC / 800v DC converters.. completely different thing, they try to sell a single-source PV invertor-to-35KV AC solution first, then 35KV to 800V DC second, to have a sorta complete solution of PV-to-datacenter. And it's only 5MW. And only 35KV AC. For moving 100MW even over a few km you would need 110KV at least. I think. An overhead wire can handle about 600A of current, that's the physical limit and the reason for kilovolts there.
Consider also that there is nothing existing in transmission and switching gear certified for HVDC it being rare one-off projects so far, while AC is ubiquitious, more-or-less mass-produced and many people are trained in its maintenance.
> If a 100MW PV farm and a data center are separated by 1km (20 Olympic pools) - is there a way to avoid AC?
Not in any economical sort of way. A rectifier and two transformers is cheaper than directly switching HVDC. If you step up the voltage to 115kV, a 100MW three-phase AC circuit is only 500 amps.
I mean if you wanna leapfrog china just throw more money into switched mode inverters and rectifiers. You don't have to use transformers. As long as you have a black cheque, there are plenty of options.
I think the article has things backwards. It's the shortage of stable demand that is holding back the building of transformers. A transformer factory that can make reliable, efficient, large transformers takes a long time to create because a lot of it relies on institutional memory. But it can be destroyed much more quickly by adverse market conditions and impatient investors.
Remember that the product has a typical lifetime measured in decades, there are huge numbers of large power transformers that have been in near continuous operation for over half a century. When one of those fails it is often more economical to repair it than replace it with a new one but that depends on there being institutions that understand what was done fifty years ago. All this requires the opposite of modern move fast and break things investing.
We had a decent domestic transformer manufacturing ecosystem for decades, and all the major manufacturers shut down their NA plants and moved them to South America to make a little bit more money. The problem is, as it often is, the perverse incentives of the religion of quarterly earning reports.
Capitalism is a fire - if you tend it well and regulate it, it serves a useful function. Let it burn out of control and it will consume everything.
We should learn something from the Cuban heavy transformer manufacturing enterprise. I mean I get that critical theory is that easy, but the step everyone always seem to
miss is a credible explanation how a centrally managed command economy is better. There is also basically no free market mechanisms functioning in power delivery in the United States. You can’t stick a solar panel in your backyard and sell the output.
Possibly the easiest way to bring any metropolitan area or region into the Stone Age for unknowable amounts of time is simply to destroy large, bespoke transmission (rather than distribution) transformers. Crazy people shooting out the cooling systems have done this several times.
Meaningful grid security means these items need rapid, standardized, domestic production capacity and cold spares distributed offsite and ready to be deployed should anything happen to ones in use. These are critical items that must not be neglected to reactive actions disaster recovery.
I live near Palo Alto and a transformer blew at the closest PG&E substation and it took six days to ship in a replacement and they cut electricity to the whole area for that time, I want to say last year or the year prior. So it’s not like they had to wait years as described in the article for new builds but I wonder where they got it from and how many do they have.
It might be easier for DC transmission components to be standardized. Sure, anything with complex controls has a lot more opportunity to fail to interoperate, but DC gear can often be configured for different voltage ratios and can much more directly control how much current flows where.
Maybe the grid needs a multi-source agreement for equipment like the network industry has for optics.
I've been wondering for awhile about the economics of the AC vs DC grid thing. Historically, AC made a lot more sense because transformers are simple and relatively straightforward to make. But now we have amazing capabilities to handle enormous amounts of power with modern IGBTs and similar power-switching transistors. (A modern high-end EV motor controller, for instance, might be able to handle a megawatt of power. Not continuously, but still.) Is a DC-DC converter now more economically viable than an equivalent transformer? The former is more techincally complicated, but the latter is bulky and requires large quantities of expensive input materials like copper.
A big problem with solid state electronics is fault current handling. The grid would become extremely brittle if it was purely a DC conversion setup. Semiconductors don't do too well outside their happy zone. All it takes is some wind and tree to fire up a very large arc welder. If you can't momentarily handle 10x+ the rated system capacity, you are gonna have a really bad time. Ordinary transformers in oil bath can take a hammering for many cycles. A semiconductor wouldn't make it through one.
That seems like it's within striking distance of competitive, no? You get some major advantages in size and production automation. Perhaps it's ok for it to die sooner if you can get it built now and then replace it later.
If you are not ready to lock yourself in a bunker after reading the article and watching that short, I strongly suggest you consider the inclined plane.
You’d better do it now. Very few locks work in the absence of transformers, springs and inclined planes.
transformers are infrastructure, 100% duty cycle with a significant overload capacity that can be 3 times name plate, right there with dams and bridges, and if one developes a fault, realy bad things happen and you get a crater.
There very nature makes them imensely heavy and very compact, all of the equipment used to form the parts is gargantuan, and materials to build them come in units that must be moved by house sized forklifts, consider changing a tire on such a forklift.
Remember that the largest transformers travel on the heaviest rail cars, specials, these things are way heavier than anything else per ft³.
Which gets us to cold,warm, or hot idle, or decomisioning, which are your choices when a huge factory has no work, hot idle means limited production, warm means some of the guys hang out and tinker with stuff, cold means, locked up,no employees but security, as decomisioning something like this has strategic considerations, or should.
That's in the ballpark of the Heathrow transformer that blew, I think.
I understand they will be not cheap, with tariffs and all, but nothing the Magnificent 7 or Heathrow could not afford.
It seems to me that (as the article points out) that production facilities are pretty old and production COULD be much more automated, and products improved if there was a will.
However, "Now those firms are seeing a rise in demand for transformers alongside the buildout of data centers for AI, but remain unsure if the trend will continue, says Gonzalez Isla. “Transformer companies aren’t going to open new plants only to shut it down after 10 years of business,” she says."
And THAT seems to be the crucial difference here between the transformer industry and, say, NVIDIA.
Standards/regulations that the transformers almost certainly don't meet, produced in factories that don't follow any standards or regulations, and then add in the cost and delay in shipping something like that.
China has the most sophisticated grid in the world, and is spending $100B a year on expanding and upgrading. They have a uniquely high share generated by renewables. It runs 800kV and will go higher after the upgrade. The first Small Modular Reactor will come online this year. If you think that’s all just being built in random factories without standards you’re very much mistaken.
"Transformers are necessary to make the AC system work."
This isn't quite wrong but the motivation is backwards: AC is necessary to make transformers work.
1. All grids need to move energy at high voltage and low current to minimize losses.
2. This requires a mechanism to step voltages up and down for transmission.
3. In 1890 the only such mechanism was the transformer.
4. Transformers only work on AC, not DC.
Hence our legacy grid is AC.
Nowadays we have an additional mechanism: Power electronics. Power electronics work on both AC and DC, so transformers with their huge requirements for copper and steel are no longer necessary.
We need to accelerate the transition of our grid to DC because DC grids are simpler and cheaper than AC grids.
Grid-scale power electronics are also extremely niche and expensive, perhaps moreso than transformers. HVDC is used where it has a significant advantage, but ease of conversion is not one of those.
> so transformers with their huge requirements for copper and steel are no longer necessary.
smelting some copper and steel and wounding it up is far, far, far, far cheaper than replacing it with power silicon(which might be smaller, but overall needs tons more of energy to produce)
It will be also less reliable. Transformers deal with any overload far better and routinely run for like 50+ years
This problem threatens more than just business as usual. In the past, intense geomagnetic storms have caused big transformers to explode. Without a backup transformer supply, entire regions may face long-term power outages.
The basic problem is easy to grasp, like the mess with charging cords for laptops before it, every large power transformer is a custom design. The fix would be to standardize on a much smaller number of options, and parallel them for the desired loads.
Think of it as analogous to USB-C power, on the megawatt/gigawatt scale. ;-)
People already parallel transformers. That's nothing new but it's usually undesirable because the extra ancillary equipment costs make paralleling more expensive than having a single transformer of equivalent rating if you are building it all at once.
But even fairly small standard specification distribution transformers are custom designs or very short runs. It's not economical to make the same design year after year because the relative prices of copper and core steel vary over time. A design made last year can be uneconomical to make this year because last year copper was relatively cheap so the designer used a lighter core and more copper to achieve the required efficiency. But if this year the copper price has gone up while the core steel price has gone down it would cost more to make the same design while the same specification could be achieved for a lower material cost by making a new design.
The new design is not a new type and for distribution transformers the effort required to design it is of the order of a man hour or two, far less than the difference in material costs.
For very large transformers (megavolt HVDC for instance) the situation is somewhat different and the design can take a very long time. But the opportunities for standardisation are relatively small because the quantity of units in the market is small and the manufacturers and regulators are always chasing ever greater efficiencies.
A far as specifications go there is already quite a lot of standardisation. But standards evolve over time and transformers can last for over half a century so you inevitably end up with a mixture of device types
Also, if one of your paralleled large power transformers fails you can't just buy an off the shelf replacement because no one keeps a stock of items that cost a million dollars each.
Switching to USB-C was trivial because most of the devices involved are essentially consumables with lifetimes measured in handfuls of years ad often much less so the old stuff withers away rapidly. That is not the case with large capital projects such as national electrical networks
And if we didn't have all these data centers would we still be "bottlenecked?"
Or is this a case of blind greedy investment outcompeting civilian life once again? Probably given that Bloomberg feels the need to throw it's voice for the cause.
An article that deeply buries the lede under elementary facts about electrical transmission.
Transformers are made in specialized factories and use specialized components made in even more specialized factories. Expanding production requires not just immediate demand but commitment to future demand because a factory is a very expensive thing. The big thing is that increased demand often involves a demand that won't continue for a long period of time.
You could see the same thing with both masks and vaccines during covid - ramping up ten factories to meet a temporary demand would be very expensive.
They're also heavy. The tragedy of Russia destroying the Ukrainian An-225 was it was one of the only ways to move very big grid scale transformers on short notice.
Maybe few manufacturers of specialized components colluded to not increase much production capacity, even with increased demand, so that prices don't collapse.
It's a well known bottleneck for the entire grid stability, not just the EV future. We even had a way to improve the losses but Congress decided that was too progressive and we're back to just not having enough.
The large transformer shortage has been a problem for years. Large transformer making is a craft, where the winding supports are made of hardwood, like furniture, and wound by hand. Then the windings go into a case that's an oil tank.
The build teams aren't that big - 30-50 people. The main barrier to entry is that it takes people who know how to hand-build big transformers. Utility buyers want a supplier who's going to be around half a century from now, since these things last that long.
Here's a summary of the market, from a transformer maker in China.[1]
Here's an AI-generated fake video of large transformer manufacturing. It's about half wrong.[2] But right enough to be worth watching. I'd like to see the prompts for this.
Virginia Transformer is the US's biggest maker of large transformers.[3] They advertise their "short lead times" of two years. The margins are low, and makers don't want to go idle between orders. This is a problem with much heavy machinery. It could be built faster, but when you catch up, everybody gets laid off and the factory sits idle. There goes your profit margin.
[1] https://energypowertransformer.com/2025-u-s-power-transforme...
[2] https://www.youtube.com/watch?v=ZVVCCG0KkaE
[3] https://www.vatransformer.com/shortest-lead-times/
Which half?
You probably got a lot from this video, because you know which half is wrong. I'd probably get negative knowledge from this video, because I don't.
This may be a new incarnation of the "curse of knowledge," where one over-estimates the value of AI slop if they already know the subject...
That's a good point. The AI slop video is inaccurate, but entertaining.
For comparison, here's the real deal - transformer winding at Virginia Transformer in the US.[1] That video provides a good sense of why these things take so long to make. All those wooden parts. All that slowly and carefully hand wound heavy wire. As they point out, if that wire can move at all, as the magnetic fields pushes and pulls on it, the vibration will, over time, wear out the transformer. It's a very fussy job to get the position and tension right, with wire firmly supported against movement in all directions. That's the difference between a lifetime of a few years and many decades.
It's a boring video.
Here's the whole manufacturing process at ETD in the Czech Republic.[2] This shows roughly the same sequence of steps as the fake video, but it's real. Big industrial bay with lots of transformers and overhead cranes. Sheets of lamination steel. Winding. The moving and shipping of the big transformer. All that is in both the real video and the AI slop. This is the real video from the manufacturer, and it assumes that if you're watching, you know what you're looking at. There's little narration.
It's a confusing video.
Here's a small open frame transformer.[3] If you've done much electrical or electronics work, you've seen one, and may have replaced or installed one. When you see the big ones being built, the process makes sense. Same concept, with a laminated core, windings, insulation, and lead wires. The big ones have the same key parts, just much bigger. But if you don't know a transformer from a transistor, the manufacturer videos are just wallpaper.
And there's the problem. The AI slop version will give the average viewer a general idea of the process. The accurate videos from manufacturers require more background knowledge to comprehend. The target audience is different. The manufacturers don't make those videos for the general public.
[1] https://www.youtube.com/watch?v=Bodj4f3L4RU
[2] https://www.youtube.com/watch?v=G3O979En_kQ
[3] https://www.mscdirect.com/product/details/20594073
As far as I can tell, the video itself is 100% fake. A bad fake. I particularly love the part where the worker levitates a large coil of steel with his hands. The narration sounds OK, so just turn off your monitor.
This. Well said!
You'd think if it's causing this much of a problem, there would be money available.
It's a generic problem with flat demand in heavy industry. Shipbuilding, bridges, nuclear reactors - when the production backlog runs down and the factory goes idle, the factory dies. So do the companies that feed specialized parts into the process.
2 replies →
This would be a great opportunity for the government to get involved.. Tell them to just make two of every order they have now and the government will buy the second one at whatever price the customer is paying. Put the spares in a strategic repository and sell them at “cost” to whoever wants them. Would be a much better use of a few billion dollars than some asinine Star Wars II or another half a trillion into the war maw.
27 replies →
> You'd think if it's causing this much of a problem, there would be money available.
There’s plenty of economic solutions if companies are really that desperate. They can pay a premium to encourage more investment. They can invest themselves, or enter into partnerships, acquire their suppliers or even open their own facilities.
Companies often complain about shortages, but it usually comes with the caveat ‘at the price we’re willing to pay’
I worked in a machine shop of a large shipyard and 20 years ago they had CNC Winders for rebuilding armatures for Navy ships.
They replaced older versions that were NC winders.
So this "hand wound" story is just that.
Windings going into Oil Tank? I think you mean varnish tank... After the rotating assembly is balanced they go into a protective coating tank that is a varnish.
Afaik utility scale Transformers operate in cooling tanks. Maybe that's what they meant about "afterward" "oil" covers many things. I believe the constant cycling and thermal load can make heinous PCBs.
1 reply →
> Large transformer making is a craft
Dumb question: why can’t we mass manufacture smaller transformers and join them up?
You can’t “join up transformers”, a single-phase transformer is an iron core with two sets of copper or aluminum windings around it. As far as I know, the only way to increase the volt-amp rating is more iron and copper/aluminum. Wiring them in series doesn’t increase the volt-amp rating but you can use (3) paralleled single-phase transformers for a three-phase circuit, I occasionally see a set of three medium-voltage to 480V ‘pineapple’ transformers (with visible live parts!) at the service entrance for older buildings.
https://en.wikipedia.org/wiki/Transformer#Construction
4 replies →
Why can’t we mass manufacture aircraft carriers by making a lot of small boats and joining them up?
It’s the same kind of problem.
(And notably, it’s not that it’s actually completely impossible to do it that way - just impractical compared to the alternative. You could actually make something that kind of sorta worked for an aircraft carrier by joining tens of thousands of small pontoons and support ships. Operationally, it would just suck compared to the alternative.)
1 reply →
How would you join transformers?
> Here's an AI-generated fake video of large transformer manufacturing. It's about half wrong.[2] But right enough to be worth watching. I'd like to see the prompts for this.
I'd like some sort of shared blocklist support for YouTube and Instagram. I'm sick and tired of content thieves and AI slop farms.
> So how did we get to a point where one component can hold trillion-dollar industries hostage? Turns out, a quirk of history made the entire world’s electricity systems reliant on transformers.
> At the end of the 19th century, when electricity was just starting to become a commercial source of energy, two businessmen fought to control its future in what came to be known as “the war of the currents.” Thomas Edison promoted the use of direct current (DC) and George Westinghouse, inventor and industrialist, was convinced that alternating current (AC) would prove more practical.
> In a clash of personality, finance and some genuine technical advantages, Westinghouse won out and the world has been mostly stuck with using AC as a means of generating and transmitting electricity. Transformers are necessary to make the AC system work.
This entire section is a glaring load of nonsense and needs to be removed. We had to start with AC for a variety of technical reasons, the main one being that boosting DC voltage pre-switching technology was impossible. DC cant pass through a transformer unless it is converted to some form of AC, usually in the form of PWM square waves these days. Before the invention of the mercury arc rectifier (And later valve) in 1902 you had boost DC using mechanical methods: generators. The problem there is physical, they did not have the ability to insulate the generator windings at high voltage potentials. They also had problems with DC voltages over 2000 volts on commutators [1] citing excessive arcing. Commutators are also a limiting factor in machine size as beyond several MW they dissipate too much power. So with all this the highest practical voltage for a DC grid using early electrical machinery is around 2 kV. Now imagine all that mechanical complexity on the distribution end. Meanwhile, early AC transmission was already in the tens of kilovolts: 11/22/33 kV (multiples of the early Edison 110 volt standard.)
As for the whole war of currents, I feel it is vastly overstated and was more a public spectacle than serious scientific dispute. It was already known from early on that AC was the future thanks to its ability to easily be transformed to higher voltages for transmission and back again with no moving parts. The "war" was likely Edison marketing to sell off the remaining inventory less desirable DC machinery.
1. https://en.wikipedia.org/wiki/Commutator_(electric)
Yes this is the most glaring issue. There also two disconnects later in the article: at the end it laments how china has been increasing transformer manufacturing but the US government has done nothing. Then in the next sentence its mentions trumps tariffs have increased transformer costs, I. E. Government action to increase domestic production. It also glosses over the new DOE rule on how transformers are made…so maybe there is a larger story there relevant to the lack of supply.
Tariffs don't help onshore manufacturing when they apply to the materials that the manufacturing needs and might evaporate before the manufacturing capability is actually created. Tariffs needs to be applied carefully and consistently to actually encourage this.
1 reply →
We had targeted policies under Biden to increase US production of grid components. This entailed invoking the DPA and setting aside millions for manufacturing improvements. Trump paused all that and created blanket tariffs that don’t seem like they’re designed to onshore US manufacturing of these very specific components but do increase all the material costs. This is not an easy thing to fix with dumb tariffs, and it’s really easy to make everything worse.
1 reply →
Also it is quite a leap to the conclusion that HVDC eqipment will not have long lead times as well especially since there are quite a few less companies making it.
> practical voltage for a DC grid using early electrical machinery is around 2 kV.
What is a current (pun!) practical limit?
If a 100MW PV farm and a data center are separated by 1km (20 Olympic pools) - is there a way to avoid AC?
I know there are future solutions [1]
[1] https://techcrunch.com/2025/04/07/former-tesla-exec-drew-bag...
The early limit was because high voltage DC required producing it at the generator, whereas you could produce high voltage AC by generating at a lower voltage and then stepping it up with a transformer for long distance transmission.
The rules are changing because of switchmode voltage conversion, using transistors to switch the voltage at a high frequency, where the magnetics (transformers, inductors) can be much smaller and more efficient, then converting back to DC. This is how virtually all smaller power supplies have been made for years, the only question (which I don't know) being how far along we are at reaching the voltage levels of long distance transmission in this way.
I'd think that hustling us towards DC with electronic voltage conversion would be a reasonable strategic goal for dealing with the transformer problem, worthy of support by a government.
5 replies →
That link talks about 5MW 35kv AC / 800v DC converters.. completely different thing, they try to sell a single-source PV invertor-to-35KV AC solution first, then 35KV to 800V DC second, to have a sorta complete solution of PV-to-datacenter. And it's only 5MW. And only 35KV AC. For moving 100MW even over a few km you would need 110KV at least. I think. An overhead wire can handle about 600A of current, that's the physical limit and the reason for kilovolts there.
Consider also that there is nothing existing in transmission and switching gear certified for HVDC it being rare one-off projects so far, while AC is ubiquitious, more-or-less mass-produced and many people are trained in its maintenance.
> If a 100MW PV farm and a data center are separated by 1km (20 Olympic pools) - is there a way to avoid AC?
Not in any economical sort of way. A rectifier and two transformers is cheaper than directly switching HVDC. If you step up the voltage to 115kV, a 100MW three-phase AC circuit is only 500 amps.
HVDC transmission over 100kV lines are common now. https://www.emeranl.com/maritime-link/overview
Yup. The only thing missing from the writeup is a eulogy for the death of the rotary converter.
I mean if you wanna leapfrog china just throw more money into switched mode inverters and rectifiers. You don't have to use transformers. As long as you have a black cheque, there are plenty of options.
*blank
I think the article has things backwards. It's the shortage of stable demand that is holding back the building of transformers. A transformer factory that can make reliable, efficient, large transformers takes a long time to create because a lot of it relies on institutional memory. But it can be destroyed much more quickly by adverse market conditions and impatient investors.
Remember that the product has a typical lifetime measured in decades, there are huge numbers of large power transformers that have been in near continuous operation for over half a century. When one of those fails it is often more economical to repair it than replace it with a new one but that depends on there being institutions that understand what was done fifty years ago. All this requires the opposite of modern move fast and break things investing.
We had a decent domestic transformer manufacturing ecosystem for decades, and all the major manufacturers shut down their NA plants and moved them to South America to make a little bit more money. The problem is, as it often is, the perverse incentives of the religion of quarterly earning reports.
Capitalism is a fire - if you tend it well and regulate it, it serves a useful function. Let it burn out of control and it will consume everything.
We should learn something from the Cuban heavy transformer manufacturing enterprise. I mean I get that critical theory is that easy, but the step everyone always seem to miss is a credible explanation how a centrally managed command economy is better. There is also basically no free market mechanisms functioning in power delivery in the United States. You can’t stick a solar panel in your backyard and sell the output.
3 replies →
Possibly the easiest way to bring any metropolitan area or region into the Stone Age for unknowable amounts of time is simply to destroy large, bespoke transmission (rather than distribution) transformers. Crazy people shooting out the cooling systems have done this several times.
Meaningful grid security means these items need rapid, standardized, domestic production capacity and cold spares distributed offsite and ready to be deployed should anything happen to ones in use. These are critical items that must not be neglected to reactive actions disaster recovery.
https://en.wikipedia.org/wiki/Metcalf_sniper_attack
https://en.wikipedia.org/wiki/Moore_County_substation_attack
https://en.wikipedia.org/wiki/Electrical_grid_security_in_th...
I live near Palo Alto and a transformer blew at the closest PG&E substation and it took six days to ship in a replacement and they cut electricity to the whole area for that time, I want to say last year or the year prior. So it’s not like they had to wait years as described in the article for new builds but I wonder where they got it from and how many do they have.
It might be easier for DC transmission components to be standardized. Sure, anything with complex controls has a lot more opportunity to fail to interoperate, but DC gear can often be configured for different voltage ratios and can much more directly control how much current flows where.
Maybe the grid needs a multi-source agreement for equipment like the network industry has for optics.
Yet another good reason for at-home solar and storage.
Does very little to offset things like no power for hospitals, refrigerated food distribution, manufacturing of almost anything.
14 replies →
I've been wondering for awhile about the economics of the AC vs DC grid thing. Historically, AC made a lot more sense because transformers are simple and relatively straightforward to make. But now we have amazing capabilities to handle enormous amounts of power with modern IGBTs and similar power-switching transistors. (A modern high-end EV motor controller, for instance, might be able to handle a megawatt of power. Not continuously, but still.) Is a DC-DC converter now more economically viable than an equivalent transformer? The former is more techincally complicated, but the latter is bulky and requires large quantities of expensive input materials like copper.
A big problem with solid state electronics is fault current handling. The grid would become extremely brittle if it was purely a DC conversion setup. Semiconductors don't do too well outside their happy zone. All it takes is some wind and tree to fire up a very large arc welder. If you can't momentarily handle 10x+ the rated system capacity, you are gonna have a really bad time. Ordinary transformers in oil bath can take a hammering for many cycles. A semiconductor wouldn't make it through one.
Still roughly 2x the cost and about 10x lower MTBF.
That seems like it's within striking distance of competitive, no? You get some major advantages in size and production automation. Perhaps it's ok for it to die sooner if you can get it built now and then replace it later.
3 replies →
I can think of thousands of components that can hold trillion dollar industries hostage.
I challenge you to name one that cannot and that also makes it into high school curricula or How Things Work.
https://mst3k.fandom.com/wiki/A_Case_of_Spring_Fever_(short)
https://m.youtube.com/watch?v=vzKfAFsbRSk
If you are not ready to lock yourself in a bunker after reading the article and watching that short, I strongly suggest you consider the inclined plane.
You’d better do it now. Very few locks work in the absence of transformers, springs and inclined planes.
why not link the original? https://www.youtube.com/watch?v=4ttYlcrA7ys
This is so great!
transformers are infrastructure, 100% duty cycle with a significant overload capacity that can be 3 times name plate, right there with dams and bridges, and if one developes a fault, realy bad things happen and you get a crater. There very nature makes them imensely heavy and very compact, all of the equipment used to form the parts is gargantuan, and materials to build them come in units that must be moved by house sized forklifts, consider changing a tire on such a forklift. Remember that the largest transformers travel on the heaviest rail cars, specials, these things are way heavier than anything else per ft³. Which gets us to cold,warm, or hot idle, or decomisioning, which are your choices when a huge factory has no work, hot idle means limited production, warm means some of the guys hang out and tinker with stuff, cold means, locked up,no employees but security, as decomisioning something like this has strategic considerations, or should.
I don't understand: is it illegal to buy transformers from China? It looks like they are building them like crazy, and are available for sale:
E.g., https://lindahongli.en.made-in-china.com/product/SAapQolWVUY...
That's in the ballpark of the Heathrow transformer that blew, I think.
I understand they will be not cheap, with tariffs and all, but nothing the Magnificent 7 or Heathrow could not afford.
It seems to me that (as the article points out) that production facilities are pretty old and production COULD be much more automated, and products improved if there was a will.
However, "Now those firms are seeing a rise in demand for transformers alongside the buildout of data centers for AI, but remain unsure if the trend will continue, says Gonzalez Isla. “Transformer companies aren’t going to open new plants only to shut it down after 10 years of business,” she says."
And THAT seems to be the crucial difference here between the transformer industry and, say, NVIDIA.
Standards/regulations that the transformers almost certainly don't meet, produced in factories that don't follow any standards or regulations, and then add in the cost and delay in shipping something like that.
Maybe.
https://evernewtransformer.com/pt/how-to-purchase-power-tran...
If I were to desperately need a power transformer, I'd consider going down this route in parallel to waiting for years for a "blessed" one.
China has the most sophisticated grid in the world, and is spending $100B a year on expanding and upgrading. They have a uniquely high share generated by renewables. It runs 800kV and will go higher after the upgrade. The first Small Modular Reactor will come online this year. If you think that’s all just being built in random factories without standards you’re very much mistaken.
3 replies →
And yet somehow the biggest electric grid in the world is running on them just fine.
"Transformers are necessary to make the AC system work."
This isn't quite wrong but the motivation is backwards: AC is necessary to make transformers work.
1. All grids need to move energy at high voltage and low current to minimize losses.
2. This requires a mechanism to step voltages up and down for transmission.
3. In 1890 the only such mechanism was the transformer.
4. Transformers only work on AC, not DC.
Hence our legacy grid is AC.
Nowadays we have an additional mechanism: Power electronics. Power electronics work on both AC and DC, so transformers with their huge requirements for copper and steel are no longer necessary.
We need to accelerate the transition of our grid to DC because DC grids are simpler and cheaper than AC grids.
Grid-scale power electronics are also extremely niche and expensive, perhaps moreso than transformers. HVDC is used where it has a significant advantage, but ease of conversion is not one of those.
> so transformers with their huge requirements for copper and steel are no longer necessary.
smelting some copper and steel and wounding it up is far, far, far, far cheaper than replacing it with power silicon(which might be smaller, but overall needs tons more of energy to produce)
It will be also less reliable. Transformers deal with any overload far better and routinely run for like 50+ years
This problem threatens more than just business as usual. In the past, intense geomagnetic storms have caused big transformers to explode. Without a backup transformer supply, entire regions may face long-term power outages.
https://pmc.ncbi.nlm.nih.gov/articles/PMC12303307/
The basic problem is easy to grasp, like the mess with charging cords for laptops before it, every large power transformer is a custom design. The fix would be to standardize on a much smaller number of options, and parallel them for the desired loads.
Think of it as analogous to USB-C power, on the megawatt/gigawatt scale. ;-)
People already parallel transformers. That's nothing new but it's usually undesirable because the extra ancillary equipment costs make paralleling more expensive than having a single transformer of equivalent rating if you are building it all at once.
But even fairly small standard specification distribution transformers are custom designs or very short runs. It's not economical to make the same design year after year because the relative prices of copper and core steel vary over time. A design made last year can be uneconomical to make this year because last year copper was relatively cheap so the designer used a lighter core and more copper to achieve the required efficiency. But if this year the copper price has gone up while the core steel price has gone down it would cost more to make the same design while the same specification could be achieved for a lower material cost by making a new design.
The new design is not a new type and for distribution transformers the effort required to design it is of the order of a man hour or two, far less than the difference in material costs.
For very large transformers (megavolt HVDC for instance) the situation is somewhat different and the design can take a very long time. But the opportunities for standardisation are relatively small because the quantity of units in the market is small and the manufacturers and regulators are always chasing ever greater efficiencies.
A far as specifications go there is already quite a lot of standardisation. But standards evolve over time and transformers can last for over half a century so you inevitably end up with a mixture of device types
Also, if one of your paralleled large power transformers fails you can't just buy an off the shelf replacement because no one keeps a stock of items that cost a million dollars each.
Switching to USB-C was trivial because most of the devices involved are essentially consumables with lifetimes measured in handfuls of years ad often much less so the old stuff withers away rapidly. That is not the case with large capital projects such as national electrical networks
On a national scale, keeping a stock of transformers is peanuts. For this to be viable, they only need to be interchangeable, not identical.
1 reply →
And if we didn't have all these data centers would we still be "bottlenecked?"
Or is this a case of blind greedy investment outcompeting civilian life once again? Probably given that Bloomberg feels the need to throw it's voice for the cause.
One of the big complaints you’d hear a lot from some about EVs early on was the tax on the electric grid.
But whenever it’s AI data centers being discussed, they’re not talked about as a tax, but that they don’t have enough power.
As if the former is a detriment to humanity and the latter is a benefit that deserves more resources, when the opposite is extremely true.
“Hello! I like money!” - Mr. Krabs
Or perhaps there’s even some Malthusianism at play?
The AI companies don't have the fossil fuel companies paying for propaganda against them.
An article that deeply buries the lede under elementary facts about electrical transmission.
Transformers are made in specialized factories and use specialized components made in even more specialized factories. Expanding production requires not just immediate demand but commitment to future demand because a factory is a very expensive thing. The big thing is that increased demand often involves a demand that won't continue for a long period of time.
You could see the same thing with both masks and vaccines during covid - ramping up ten factories to meet a temporary demand would be very expensive.
They're also heavy. The tragedy of Russia destroying the Ukrainian An-225 was it was one of the only ways to move very big grid scale transformers on short notice.
This is a problem in strategic reserve territory.
Maybe few manufacturers of specialized components colluded to not increase much production capacity, even with increased demand, so that prices don't collapse.
Also https://archive.ph/yn2It
It's a well known bottleneck for the entire grid stability, not just the EV future. We even had a way to improve the losses but Congress decided that was too progressive and we're back to just not having enough.
I dug into root causing this and copper prices using Claude to gather data and visualize it. I learned a bunch. The write up with tldr at the top: https://usevawn.com/blog/copper-electrical-equipment-prices/