For decades, the physical equipment underpinning the electric grid has remained largely unchanged: passive, “dumb” devices installed as far back as the 1970s that lack much real-time control. But today, in the face of skyrocketing energy demand, a new class of technologies has emerged.
In this episode, Drew Baglino, the founder and CEO of Heron Power, returns to the show to discuss his company’s new generation of solid-state transformers, or SSTs. After a 17-year career at Tesla — where he led energy and powertrain development — Drew is now focused on replacing the grid’s aging infrastructure with these advanced power electronics.
Shayle and Drew take a deep dive into the history of the power transistor, and then explore how the SST has the potential to transform the grid into a highly optimized and intelligent machine. They cover topics like:
- The evolution of power electronics
- Why we still haven’t fixed the transformer shortage
- How Heron Power’s SSTs remove legacy transformers and switches to create a substantial uplift for project developers
- The potential to remove 70% of traditional electrical equipment at data centers by distributing power directly to the rack
- Why Drew thinks SSTs offer a “pathway toward affordability”
Resources
- Catalyst: Drew Baglino on Tesla’s master plan
- Latitude Media: Inside Heron Power’s plan to transform the grid
- Catalyst: Understanding the electric transformer shortage
- Open Circuit: The grid resilience dilemma
- Latitude Media: These Autogrid alums want to change how data centers use power
Credits: Hosted by Shayle Kann. Produced and edited by Max Savage Levenson. Original music and engineering by Sean Marquand. Stephen Lacey is our executive editor.
Catalyst is brought to you by Uplight. Uplight activates energy customers and their connected devices to generate, shift, and save energy—improving grid resilience and energy affordability while accelerating decarbonization. Learn how Uplight is helping utilities unlock flexible load at scale at uplight.com.
Catalyst is brought to you by Antenna Group, the public relations and strategic marketing agency of choice for climate, energy, and infrastructure leaders. If you’re a startup, investor, or global corporation that’s looking to tell your climate story, demonstrate your impact, or accelerate your growth, Antenna Group’s team of industry insiders is ready to help. Learn more at antennagroup.com.
Catalyst is brought to you by EnergyHub. EnergyHub helps utilities build next-generation virtual power plants that unlock reliable flexibility at every level of the grid. See how EnergyHub helps unlock the power of flexibility at scale, and deliver more value through cross-DER dispatch with their leading Edge DERMS platform, by visiting energyhub.com.
Transcript
Shayle Kann: Coming up: at long last, an ode to power electronics.
I’m Shayle Kann. I lead the early stage venture strategy at Energy Impact Partners. Welcome. All right, so my friend Drew Baglino came on this podcast about two years ago, right after he’d left Tesla after a 17-year stint, culminating in him leading energy and powertrain and a whole bunch of other stuff there.
At that point, he was taking some time off and figuring out what was gonna come next for him. It turns out it was power electronics. Drew started a company called Heron Power, and Heron is introducing a new type of device to the grid that combines solid state power electronics with software and controls to dramatically simplify a whole class of grid infrastructure while simultaneously imbuing it with a host of new capabilities. For all the time on this pod that we spend talking about what’s going on in electricity, I think we actually haven’t spent enough of it talking about the actual equipment that underpins the market. We know that there are long lead times for things like transformers and switchgear, but is there an opportunity to leverage the unprecedented growth in the market right now — that we’ve talked about many times — to catapult a new class of technology onto the grid at scale? I think so. For disclosure, I’m an investor in Heron and I have been since their first external round, and actually just this week, Heron announced a $140 million Series B led by Andreessen Horowitz, where we at EIP also doubled down.
Anyway, here’s Drew.
Drew. Welcome back.
Drew Baglino: Thanks, Shayle. Happy to be back.
Shayle Kann: All right. Let’s talk about power electronics. I want you to start by explaining basically what power electronics are, but maybe through a history lesson. Tell me the history of power electronics.
Drew Baglino: Yeah, I’ll try my best here. I think people have heard many times about the history of the transistor, right? And Moore’s Law and how transistors went from vacuum tubes to three nanometer devices inside of GPUs. I think most people are familiar with that, but at the same time, and using some similar technologies, an equivalent thing happened with power transistors or power semiconductors over basically the same timeframe — from like the seventies through to today. But what was improving was not the size, although the size did improve, and it wasn’t just the size alone, but actually some other very important things for power semiconductors, like the voltage that you can block with the power transistor, or the power MOSFET, or the power IGBT. And also it was the current capability — or really the current density — that you can get through the power transistor. And then one of the more important things, more recently, is actually the thermal conductivity that you can achieve through the device. If you have better thermal conductivity, it’s easier to keep the device cool, which means it’s easier to go to higher current density. And then maybe one of the most important things that has improved over the four or five decades is the switching speed of the device — how quickly can it change states?
And you might be asking: well, why does that matter? To think about why it matters, it’s useful to consider the world of electricity more broadly. How does electricity work? Until batteries existed, you couldn’t really store it. It’s generated in one place, and it’s connected through a contiguous circuit — a continuous conductor — to where it’s used at the end use. Any branches or Y’s in the circuits are all simultaneously affecting each other, unless you have devices in the middle to decouple the flow of electricity. If you don’t have any devices to decouple the flow of electricity, anything that is connected to the circuit affects anything else that is connected to the circuit — and instantaneously. It’s amazing actually how this happens. And that’s why when people say the electricity grid is like the world’s largest manmade machine, they’re not wrong, because all of the devices, the motors, everything that’s plugged into every wall is in some way affecting everything else that is plugged into the wall.
The only thing that can change that is if you can control the flow of electricity. And that is what power electronics, as they have improved over the past four or five decades, actually allow you to start to do — dynamically, and with modern devices made out of silicon carbide and gallium nitride — millions of times per second, stop and start the flow of electricity through a power transistor and ultimately control the power flow through circuits in a device or on the grid. That’s a very zoomed-out view, and I can certainly go into more details.
Shayle Kann: Well, I think a key point to make here is that you talked about the electricity grid and what power electronics can do on the grid, but that’s actually not mostly what’s on the grid today. And second of all, it’s actually not mostly where power electronics are used today. So I want to spend a minute on where power electronics are used today — one of the places where you have a bunch of personal experience — and then get back to what that enables for the grid, which is what you’re building at Heron. Am I wrong?
Drew Baglino: No, you’re right.
Shayle Kann: We have power electronics on the grid, but it’s not common.
Drew Baglino: Yeah, the first applications of power electronics in like the late seventies and early eighties were built on relatively slow switching, relatively large format thyristors and IGBTs. The first place they went to be applied was towards variable frequency drives on large industrial motors.
This is a great example of — before power devices existed in this way — those motors were kind of just always spinning, always ready to go at full power. Even if the pump in your factory or the fan on some large air handling system didn’t need to run at full speed, you had to have large mechanical relays to switch it on and off, but you could only do that a couple thousand times before the switch would fail. And it wasn’t a fast-moving switch — you’d go over and hit a breaker and turn it on, and it would be on for the whole day.
Those first applications — variable frequency drives — instead matched the need of the load, whatever the water flow rate you wanted or whatever you wanted to do with your electric motor in your manufacturing process, to the electrical load. The electrical load would match the mechanical load and all of a sudden you had a lot of efficiency in industry because of variable frequency drives.
The next real application was large AC to DC to AC switching stations on the grid. These were air-insulated, like the size of central exchanges — if you remember what telecom buildings used to look like. They would allow you to decouple one islanded grid from another islanded grid using a DC link between them. These are uncommon infrastructure, but actually useful when you look at the US with five separate major balancing authorities in the electricity grid. If there are DC links between them — they’re not very high power capability, but they are there — they allow the frequency to be different in different places, and yet you can actually tie power flow between them.
Shayle Kann: Right. So it’s like when people talk about ERCOT as an island grid — it is technically an island grid, but it’s not like there isn’t a physical connection to, I guess, probably three of the other —
Drew Baglino: Yeah, there are these DC links that allow some power flow between the different regional authorities. And there are other DC links — like between Europe and the UK, or between the North Island and South Island in New Zealand. They’re pretty common. The way they were built in the eighties used early power electronics devices that switched really slowly — maybe a thousand times a second — but that allowed you to do this AC to DC to AC kind of conversion.
Then power silicon started to become better — these are silicon MOSFETs, like 100-volt, 200-volt silicon MOSFETs in the early eighties. You started to see these in switching power supplies on PCs, VCRs, TVs, cordless phones, answering machines — all of the consumer electronics of the eighties and early nineties had some small switching power supply, or they had silicon diodes in them with a traditional transformer. So you’d have a 60 hertz transformer in a wall wart on your wall that would go from 120 volts down to like 10 or 8, and then that would go through a bridge rectifier — a power device — to make a DC voltage that would go into the electronics device.
That was edge-based power electronics for consumer, low-power applications. Then silicon IGBTs got better and were able to do like 600 volts or maybe even 1200 volts. In the early nineties you started to see solar inverters and early drive inverters for electric vehicles, and maybe variable speed fans for HVAC systems in homes — all these interesting edge applications where you needed to go either AC to AC at different frequency for variable speed motors, or DC to AC with solar, or AC to DC for a battery, or in an electric vehicle, from the DC battery to the AC motor.
These were awesome applications of devices that existed at the time, that could switch tens of thousands of times per second, and could handle 600 to 1200 volts — maybe 100 amps in a single device. Then as you get into the 2000s and 2010s, some researchers in the US started working with some new wider band gap semiconductor materials — silicon carbide and gallium nitride — that had some intrinsically awesome characteristics. Silicon carbide can switch super fast and has really good blocking voltage capability. While I was at Tesla, we started using silicon carbide to make drive inverters in cars because the incremental cost of the more expensive transistor was more than outweighed by the savings in battery, because the drive inverter could be so much more efficient using silicon carbide. So while you might spend $100 more on silicon carbide devices in the car, you’d save $400 to $500 in battery.
Shayle Kann: Is it true that electric vehicles were what really drove the supply chain scale-up for silicon carbide? What is the supply chain like for silicon carbide and how has it matured over the past decade now?
Drew Baglino: Yeah, in 2010 the supply chain for silicon carbide was tiny. Silicon carbide was used in LEDs and nothing else really. But some folks at Wolfspeed and Infineon and a few other device manufacturers were like, this is gonna be an amazing power semiconductor platform. And they started to develop a whole bunch of different devices — first in the 600-volt class to support EVs, and then later at higher voltages to support grid applications.
The first way we incorporated it into Teslas was with the Model 3 onboard charger. We wanted to make the onboard charger more affordable. The best way to make power electronics systems that involve isolation more affordable is to go up in frequency, because to get isolation you basically need to use a transformer of some type, and transformers become smaller as you go up in frequency. It’s a linear relationship between frequency and size, based on how much energy you can store in an inductor and how quickly you’re charging and discharging that inductor — if you charge and discharge it faster, you’re moving more energy per unit time and you can make the inductor smaller. We really wanted to make the onboard charger smaller. So we used these early silicon carbide devices to make the onboard charger. We increased its power density by like a factor of two and dramatically reduced its cost. At the same time, that onboard charger also did the DC to DC conversion between the battery bus and the low voltage net in the vehicle, so it was a great integration play.
Silicon carbide then went into the drive inverter to make it about 1% more efficient — which sounds like not a lot. But when you think about it: you size the battery to give you, say, 300 miles of range. That 1% is worth three miles worth of battery, and 1% of battery is a lot.
Shayle Kann: All right. So you walked through a good history of power electronics, ending with your own personal experience with silicon carbide specifically as a class of power electronics within Tesla vehicles. Let’s contrast that to what’s on the grid today. Like, what do we use today at those branching Y’s on the grid?
Drew Baglino: We use — prior to power electronics really becoming a thing in the seventies and eighties — the only way you could switch electricity or the flow of electricity was with mechanical switches. Think of the breakers in your breaker panel, or maybe you’ve looked into your neighborhood utility switch yard and seen these huge armatures that spring open to disconnect one feeder or reconnect another. These are large, bulky, slow — slow meaning they actuate in hundreds of milliseconds — and can actuate maybe once every couple of minutes, and are really not meant to actuate more than a couple thousand times in their total lifetime. That’s how electricity is controlled at the grid scale. There’s really not a lot of real-time, millisecond control. And this contrasts sharply with the latest generation of battery inverters or solar inverters — or the way you charge an EV — where the power electronics are actively controlling voltage and current hundreds of thousands of times per second, using really small magnetic devices.
Grid designers and electrical engineers working on power systems are really limited in the tools they can use. They have these slow switches, which they use to isolate a fault or reroute through a different line because one line is overloaded — doing that rerouting slowly, on a one-second or once-an-hour timeframe. They also don’t have any dynamic control over voltage, frequency, or power factor using power electronics at thousands of cycles per second. They just have static voltage transformers, AC to AC transformers — those gray boxes you see on your street corner or on your telephone pole. Those are passive, fixed-ratio, passive voltage dividers or voltage multipliers. There’s no control over how power flows through there. It just passively moves, following the path of least resistance.
And that is still the state of the art. What I described was true in 1970 and it’s still kind of true today in the 2020s.
Shayle Kann: In fact, many of the transformers on the grid today were installed in the 1970s. They’re very old on average.
Drew Baglino: Yeah. Over 70% of distribution transformers are over 30 years old — something like that. Some crazy statistics. There’s some stuff on the edges starting to join, some additional tools in the toolkit in the last five to ten years — STATCOMs are an example, where you have these switched capacitors that you can use to do some power factor control, and there’s some power electronics in those. They’re not used that often, but they do exist.
But what you can do with power electronics is much more than that. What we’ve seen happen with silicon carbide — silicon carbide 10 years ago was 600 volts or 1.2 kilovolts. Nowadays there are silicon carbide devices that are 2.3 kV capable or 4.6 kV capable. When you look at distribution voltages in the US — 7 kV, 12 kV, 20 kV, 21 kV, 35 kV — you don’t need too many of those devices in series to interact with the grid at those voltages. With the progression of silicon carbide, it’s now possible to make solid state transformers that can be much more capable than just a passive voltage divider.
Shayle Kann: Yeah, I want to come back to what solid state transformers can do in various applications. But before we get off the traditional transformer topic — I’m curious your perspective on that market. In the traditional oil-filled transformer world, we’ve had a supply chain that’s been gummed up for years. It dates back to when all supply chains started to get gummed up during COVID, and then one by one most supply chains cooled off and lead times for most stuff went back to normal. It did not happen with transformers. The lead times now for traditional transformers — I think both distribution and high voltage — are basically as long as they’ve ever been. I’ve had a lot of people express mystification about this, because they’re kind of dumb things that we’ve been producing for a hundred years. You’d think we could solve that problem quicker. What’s your perspective on why, absent new technology, haven’t we just solved the transformer shortage?
Drew Baglino: Yeah, there are so many factors at play. I’m not gonna try to get them in order — I’m just gonna start rattling them off. First is straight-up demand. We now have growth again, and it’s broad-based growth. There’s growth of loads interconnecting at transmission, like large data centers. There’s growth of large generation — partially because old assets are being retired, and partially because we just need more generation. So there’s a bunch of generation transformers and large transmission load interconnect transformers. And then we have broad-based distribution load growth from electric vehicles, home electrification — some of that policy driven, some just pure demand driven.
We have broad-based increases in demand. In fact, power transformers — generator transformers — demand is up over double since 2019. For generation step-up transformers, it’s up over 250%. Distribution transformers are up over 100%. And you can’t say the demand increase is just load growth, because it’s not — some of it is replacing aging infrastructure. A lot of these core transformers on the grid were built in the seventies, and at some point they need to be replaced. So some of it is aging infrastructure, and some of it is load growth.
Then there’s regulatory uncertainty. The DOE started saying things like, we’re gonna change the basic materials in transformers — or at least take some public comment about potentially doing that — in the name of making transformers more efficient. For some background: transformers are generally 99, 99.2, 99.3% efficient, depending on how they’re loaded or sized. So there’s not a lot of room to make them more efficient, but they’re everywhere. One interesting thing about that efficiency rating — this is for traditional transformers — is that it’s measured at rated load. Actually there are losses in transformers that never go away: the magnetizing losses in the steel. That’s one of the reasons why transformers use laminated grain-oriented electrical steel — to reduce that vampire loss or idle loss. And most of the transformers on the grid are not fully loaded, so you end up with a lot of that idle loss adding up all over the place. That DOE investigation was about reducing that idle loss.
With that regulatory uncertainty, maybe some people didn’t make investments in expanding grain-oriented electrical steel supply, which is one of the most important — and by mass, the biggest — contributors to passive transformers. And maybe some people were thinking solid state transformers were going to replace them. I’ve been thinking that — it’s one of the reasons why I started Heron Power. But you have people in this industry wondering whether the grain-oriented electrical steel is gonna be designed out by policy, wondering whether we’re just in a bubble of replacing a bunch of stuff built in the seventies and whether this electricity demand growth isn’t gonna be sustained. So they’re not investing as quickly as they otherwise could.
Shayle Kann: Yeah, all those things are true. The only thing I would add, having spoken to a bunch of old school legacy transformer manufacturers — they have gone through boom-bust cycles in their business over their lifetimes and they’re reticent to get out over their skis. They’re expanding, like everybody is expanding capacity —
Drew Baglino: There are a lot of announcements.
Shayle Kann: Yeah, but they’re measured about it. They’re not expanding by 5x. They’re like building a new factory and expanding by 2x, and that takes a couple years, and by the time you catch up, the data center demand forecast has gone up by another 2x anyway. And so we’re still behind. I think that’s part of it too — just reticence to overinvest amongst the incumbents. People can fault them for it, but it’s actually a reasonable position. It’s a tragedy-of-the-commons type problem: any given one of them would rather be in an undersupplied market because then they have pricing power and margin power, rather than an oversupplied one. So they might as well expand to whatever they feel highly confident they’re gonna be able to sell.
Drew Baglino: Yeah. There’s another regulatory uncertainty item I didn’t mention — tariffs. With the rapidly changing set of tariff rules and regulations both from the US and from other countries, sometimes it’s hard to know where to build a factory. These factories are relatively large investments, and it’s not just the investment that’s at risk — if you pick the wrong location, you could be on the other side of a tariff that you hadn’t predicted. So people are waiting for a lot of these things to shake out when making expansion decisions.
Shayle Kann: All right. Which creates, in some part, the market opportunity for you to come in and introduce new technology. So back to what you are doing — which is solid state transformers. But I think that kind of undersells it in some ways, because what you’re building contains a solid state transformer, but it actually replaces more than that in terms of what would otherwise have to get built if it didn’t get deployed.
I just want to talk about what this class of power electronics — these solid state transformers — can enable, by different categories. As you said, they’re used all over the place. So let’s talk about the markets you’re focused on. Starting with: if I’m going to connect a new solar project or a new battery to the grid, what is the list of things I normally need to go from generator to grid? And then in contrast, what does it look like if I install a Heron Link, which is your product?
Drew Baglino: Yeah, so I’m building a 100-megawatt solar facility. What’s on my single-line diagram? You start with trackers in the field and some combiner boxes collecting DC at around 1500 volts DC. That 1500 volts DC is brought into, most of the time, central inverters. These central inverter skids have a DC to AC inverter modular to like the one-megawatt level — so maybe four one-megawatt DC to AC inverters. The input voltage is 1500 volts; the output voltage is 690 volts AC. Then on the other side of the 690 volts AC, you have some protection devices — maybe a main breaker, some fusing — and then you connect to the low side of a step-up transformer. A medium voltage transformer, usually oil-filled. Sometimes it’s a dry-type transformer, and on the other side of that transformer you’ve got 34 kV AC, most typically. There’s also some fusing and potentially switchgear there on the skid.
Then you connect that 34 kV in a daisy-chain configuration to a bunch of these inverter skids — maybe five or six — and eventually you get to a medium voltage feeder breaker, about 600 amps worth of 34,000-volt inverters, usually something around 30 to 40 megawatts at that breaker. Then on the other side of that breaker, you have a generation step-up transformer typically rated for the full 100 megawatts. On the other side of that, you’ll have hundreds of kilovolts — 200 kV, 300 kV, 600 kV, depending on the grid.
That’s the typical single-line diagram for solar. Batteries look very similar. Inside the skid, you’ve got companies like SMA, EPC, Power Electronics, Huawei, SunGrow — they make the power electronics part, the DC to AC part. Some of them might make the transformer, most of them don’t.
Shayle Kann: They co-pack sometimes, right? They’ll put a transformer in a box with an inverter, but —
Drew Baglino: Yeah, they’ll put the transformer on the skid so it’s easy to land, but they usually don’t make the transformer. Transformers are generally made these days in China, India, and Mexico. Very few are actually made in the US.
That total system — you’ll have a 99% efficient transformer and maybe a 98% efficient inverter, so about 97% efficient total conversion, maybe 98.5% if you’re lucky. So roughly 97.5% efficient total conversion system.
When we do this with a solid state transformer, we basically move the 60 hertz transformer to a 100 kilohertz transformer, which makes it much smaller — 50 to 100 times more power dense. Now we have power electronics control on both sides of that 100 kilohertz transformer. And we don’t have a modularity of a megawatt — we have a modularity sized to that small isolation transformer, somewhere around 100 to 200 kilowatts.
The interesting thing about that level of modularity is it gives you robustness to fault, because if you have a fault, you only lose around 100 kilowatts, not a megawatt — or in the case of the transformer on that skid, if it failed, you’d lose four megawatts and you’d need a crane to replace it, possibly waiting weeks to months to get that replacement transformer.
What we remove from the single-line diagram is: we remove the legacy transformer, we remove that 690-volt breaker or fuses and protection. We also remove some power-factor-correcting capacitors at the central plant, because we no longer have the inductance of that medium-voltage transformer. And we get simpler protection on that medium-voltage feeder because we don’t have a transformer that could have a hard fault and catch fire — we have a power electronics front end that, if it faults, will be like 1.2 per unit, just slightly overloaded current. So the protection gets a lot simpler as well. That’s the kind of thing we’re doing for solar, batteries, and also data centers — although the data center story is a little different.
Shayle Kann: Yeah. We’re gonna get to that one in a second. For solar and batteries — I think there’s been a long history in the broader electricity world, and to some extent in the solar battery inverter world, where people introduce new technology and make a pitch that “this is better, it enables some control you didn’t have otherwise.” But better isn’t always what wins in the electricity market. The important thing is: what’s the net outcome that actually matters?
In the case of solar and batteries, it seems to me — and I’m curious what you think — the rank-order killer apps, because this is one of those technologies with numerous benefits, but which ones really matter? It does enable greater control, but to me it seems the biggest ones for solar and batteries might actually be reliability — like a step-function change in reliability, which people don’t appreciate, given how much failure there is of utility-scale solar in particular because of inverters — and maybe transformers to a lesser extent. But reliability, space savings, CapEx — when you’re talking to customers, what do they care the most about?
Drew Baglino: Reliability is a big one. Solar inverters are the largest source of underperformance on utility-scale solar plants. It’s not the modules — you’d think it would be the modules because there are so many of them out in the field and you worry about hail and whatever else. But actually it’s the inverters. Central inverter availability is, on average in the industry, 97.5 to 98%, which basically means two to two-and-a-half percent of the time when they should be producing power, they’re not. That’s straight bottom line on your project.
It’s not just the inverters either — it’s also the transformers. From some statistics we’ve gathered, transformers are not really designed to run at their rated power as long as they do in these desert power plants, where they’re sitting in the sun and running at nameplate rated power for eight or nine hours a day. They’re failing about 1 to 1.4% per year on average. So that transformer — which is pretty hard to replace — needs to be replaced, and if you have 100 transformers on your utility-scale solar facility, you’re replacing a transformer or more a year. And the last thing is they have no monitoring. There’s no real intelligence built into these transformers. I’ve been talking to large owner-operators of renewable power plants and they have to send people out to measure oil health and look at all the bushings and do all of these things to make sure they don’t have thermal events in the field. So they’re a big pain point that people want to get rid of.
Reliability is one. You mentioned another: the solution is about 1% absolute more efficient, which drives production value. For battery installations, it’s round-trip efficiency improvement — so it doesn’t just count once, it counts twice. We’re also taking this opportunity to simplify O&M. We don’t have any of the transformer O&M — you don’t need to check the oil or replace the oil. We also don’t have that switchgear. Altogether, we see a 5 to 6% NPV uplift for our customers building with this type of inverter versus an alternative.
Shayle Kann: All right. Let’s talk about the large load or data center use case, where there’s a similar set of benefits from switching to solid state transformers, but actually one big difference — at least as I’ve seen it — on the “delete a bunch of stuff” side, because it seems there’s a lot more stuff to delete in the data center use case.
Drew Baglino: Yeah. Data centers still distribute power today the way they did when they first came into vogue in the nineties — all the racks are connected at AC, usually like 240 line-to-neutral AC, and like 450 line-to-line. That means you’re starting with hundreds of kilovolts outside the data center. You’re doing sub-transmission voltage — 13 or 34 kV — to the different data hall areas. Then you have a three-megawatt medium-voltage transformer going from that medium voltage down to about 400 volts AC. That 400 volts AC is brought into the data hall through a gray space area with UPS systems and protection and power distribution, and then through bus bars overhead above the racks.
In a world where data center racks are 10 kilowatts, that’s maybe fine. But as they become 100 kilowatts or a megawatt, it starts to look like EV charging or grid batteries or solar — and it needs to change. Rather than using AC as a distribution means, you start looking at power electronics to go directly to DC at higher voltage as well. Now the rack, rather than having a backplane voltage of 48 volts — which is just a legacy thing from telecom switching stations of the eighties and nineties — now the backplane voltage of the racks will be 800 volts or even higher. Then you can use an SST-based solution to go from medium voltage 34 kV all the way to 800 volts with no gray space rooms with UPS systems and power distribution panels — no additional transformers at all.
You can incorporate just the amount of energy storage you need on that 800-volt side to handle GPU ripple or whatever other power ripple you have, and also allow for 30 seconds or so of hold-up time to support facility transitions — to generators or from one medium voltage connection to another. You can remove 70% of the stuff in the electrical diagram and a similar amount of footprint. And you might ask: does that really matter? The GPUs are where all the money is. That’s true, but where a lot of the time and labor shortage is, is in the certified electricians doing a ton of AC electrical work — and you’re removing all this copper demand because you’re not distributing power at low voltage anymore. You’re bringing high voltage as close as possible to the rack. So it’s a major headache alleviator — or painkiller, as you like to say, Shayle.
Shayle Kann: It’s a painkiller for sure. And the other thing is space, right? You’re deleting a bunch of stuff, which frees up a bunch of space, and space is at a premium in data centers.
Drew Baglino: Yeah, you get to bring the stuff that needs to be low-latency and close together as close together as possible, because you’ve removed all this power distribution equipment that would otherwise be occupying white space.
Shayle Kann: Okay, so we talked about solar and batteries, and we talked about data centers. Let’s go back to the grid then, just to wrap up. Over time — and obviously this will take a long time — if we go throughout the transmission-distribution system and replace, one by one, all of these traditional oil-filled transformers on the grid today with solid state transformers, what does that enable from a grid management perspective?
Drew Baglino: Utilities and grid operators right now are facing a lot of pressure. They’ve got aging infrastructure, growing demand, and they’re in the market for new solutions. Luckily, SSTs can provide a ton of value propositions beyond just voltage transformation.
An SST can have a cost similar to a traditional oil-filled transformer, but at the same time provide functions that would otherwise be provided by popcorn components around the transformer — functions like overcurrent protection and fault isolation, what an automatic tap changer does for voltage correction, what three-phase balancers do to enable higher utilization on the different phases in the distribution grid. They can provide the spinning inertia type functionality that synchronous condensers do for frequency regulation. And they can also take the place of cap banks for power factor correction.
So with the choice to go SST — the next time they need to place a distribution substation or replace an aging 50-year-old 34 kV to 208 transformer — they’re at the same time getting all of those other value-added functions kind of for free. And what those other value-added functions do is enable more utilization of the existing poles and wires. Utilization is the key to affordability.
If you look at the rate cases for public utilities at PUCs around the country, they take their total costs of new CapEx and maintaining existing CapEx, and divide that by kilowatt hours served. The best way to serve more kilowatt hours is to increase the utilization of the existing poles and wires. And to do that, you need intelligent infrastructure that can dynamically respond to the conditions of every circuit and maximize the utilization of every circuit.
So not only will SSTs ultimately cost less per unit of voltage conversion, but they’ll also add all of this additional value-added functionality that allows you to get more out of every wire — existing and new — that utilities build. And that is the pathway towards affordability. That is what the 21st century grid will look like.
Shayle Kann: When you were describing how the grid works before, I was thinking about a network of tributaries — a river system, right? At every spot where two rivers converge, there’s a Y. It’s possibly the opposite direction of what I’m imagining from a river system perspective. But if we’re trying to find the right metaphor — at every one of those connection points, we’ve always had to build a dam that allows us to control the water flow. But we used to build it with sticks and rocks, and now we have concrete and whatever the Hoover Dam is built out of. We can control it to a much higher degree.
Drew Baglino: Well, I think a better analogy — if we’re gonna use a water analogy, because I’ve thought about it — would be this: the way the grid works today is like if you have 100 units of water flowing through the upstream side of the river, and if you had control in the past, it would be that 10 units go one way and 90 units go the other. And you couldn’t really change much about that — it was always gonna be that way. So if it were 200 units coming down the river, it would be 180 and 20. But with power electronics, you can have whatever you want on the other side of that dam.
And another example is locks. Locks are used to adjust levels — imagine locks as taking a long time to move a boat up and down in potential. That’s what you’re doing: changing the potential of the boat, literally the gravity, how high above sea level the boat is relative to other parts of the river. That’s kind of what power semiconductor devices are. The most recent generation of devices can move thousands of volts in nanoseconds — and volts are potential, that’s the analog. And that’s compared to mechanical switches in the past, which were moving in milliseconds or tens of milliseconds or even seconds to do the same thing.
Shayle Kann: The locks metaphor really clicks for me specifically. I grew up — you know this — I grew up in Madison, Wisconsin, and I literally grew up across the street from a locks. There’s a river between two lakes in Madison — shout out to anybody who lived in the Tenney-Lapham neighborhood of Madison, Wisconsin, who knows the locks. The locks take forever. They really do. And I actually brought my 3-year-old back to Madison last year, and it’s a big activity — you can go watch the locks and waste a whole bunch of time with a 3-year-old. That’s great.
Drew Baglino: Yeah. There are a lot of hydrology analogies to electrical circuits. Do you know water hammer?
Shayle Kann: I’ve heard water hammer. Yeah.
Drew Baglino: Water hammer is basically like an amperage transient. If you go and turn off your water faucet from full flow to off, you get oscillations in the water column, and you need something to damp that out. If you’re a good plumber, you add that. If you don’t, that oscillation could last forever. The same thing exists in electrical circuits. And you can harness it for good — that’s what resonant converters do. They use that oscillatory behavior to have more efficient soft switching when changing from one voltage to another, or one frequency to another. But it can also be bad, and you can get oscillations that end up with grids going unstable — like what happened in Spain.
Shayle Kann: Yep.
Drew Baglino: Water hammer on the grid.
Shayle Kann: Water hammer on the grid. There it is. We figured it out. All right, Drew, this was awesome. Thank you so much for your time.
Drew Baglino: Absolutely. Thanks. Always a pleasure.
Shayle Kann: Drew Baglino is the founder and CEO of Heron Power. This show is a production of Latitude Media. You can head over to latitudemedia.com for links to today’s topics. Latitude is supported by Prelude Ventures. This episode was produced by Max Savage Levenson. Mixing and theme song by Sean Marquand. Stephen Lacey is our executive editor. I’m Shayle Kann, and this is Catalyst.
