The Art of Quality Sep 18, 2026 1h 11m 49m saved
With Landon Mossburg, founder and CEO of Peak Energy, who spent four years in Sweden bringing up Northvolt's first battery pilot line and before that worked at Tesla from the start of the Model S
About 90% of the failures on grid-scale batteries in the field come from the cooling system, Landon Mossburg said, and roughly 80% of the fires start there too.
Every grid battery built before his company's used water cooling, which consumes about 5% of the energy stored to run it. Peak Energy's product has none.
"So every other battery, before us, on the grid, has to be actively cooled — literally uses water cooling."
Mossburg joined Tesla in 2013 at the start of the Model S, moved into manufacturing, then spent four years in Sweden bringing up Northvolt's first pilot line before leading its North American expansion. Peak Energy is three years old, connected its first system to the grid a little over a year ago, and holds $1.1 billion of deposit-backed customer reservations.
The full interview is covered here so you can skip it. 71 minutes of audio, 22 minutes of reading.
Here are the 18 insights that matter.
Key Takeaways
Roughly 90% of grid-battery field failures and 80% of the fires come from the cooling system, which Peak Energy removed
Water cooling also burns about 5% of the energy stored just to run, and it is the reason batteries are not built into solar panels
A battery with no site costs around it would be five times cheaper than the best projects available today
Batteries and solar are on learning rates he puts as high as 20% a year; gas-fired power is on about 1%
Cheap gas has gone from under $50 a megawatt-hour to about $80, with behind-the-meter deals above $100
Solar plus storage is now about 30% cheaper than gas on his arithmetic, having been roughly level before the data-center boom
Fusion and small reactor companies are aiming at a target batteries and solar have already reached
xAI has installed 4 GWh of batteries at three data centers in eighteen months, because AI load swings damage gas turbines
One server rack is approaching a megawatt, enough for 800 to 1,000 homes
Peak Energy holds $1.1B of deposit-backed reservations against $10M of revenue this year and $100M expected next
A single wrong machine oil on a cutting knife stopped a battery factory for a month and took a hundred people to find
1. The Kardashev Framing
The host opened by tracing energy from fire through fossil fuels to electrification and then storage. Mossburg's reply reached for a scale of civilizations measured by how much of a star's output they can capture.
Energy use is the proxy he prefers for capability
to the extent you can harness and control energy, it's sort of a good proxy for how sophisticated and capable you are as a civilization
Landon Mossburg
Matter and energy are functionally related, he said, close to different states of the same stuff. His observation about the general public is that energy only shows up as an annoyance, in the form of a phone to charge, a tank to fill or a utility bill, while it underwrites everything. The first way any organism gets energy is by breaking chemical bonds in matter.
2. Two Kinds of Battery
Mossburg split batteries into two classes before discussing either.
The distinction most people grew up with and stopped noticing
Primary battery just means it's not reversible
Landon Mossburg
A primary cell stores energy once and is thrown away. Rechargeable cells are secondary batteries, and the process is reversible. He then extended the same description to fuel.
Combustion is a battery being discharged
If you think about fossil fuels, like oil, coal, or sugar — anything that can burn — all that's doing is storing a tremendous amount of energy in a chemical bond, and we break that bond, you get the energy out as heat.
Landon Mossburg
On the host's guess that a dam is the oldest form of storage, he agreed and explained pumped hydro: two reservoirs at different heights, water pumped up to charge and released through turbines to discharge.
The economics of pumped hydro have been overtaken
Up until very recently it was the cheapest form of energy storage on the grid — had a huge amount of capex, but really long life and very low cost to operate once you build it. Now batteries are cheaper by far.
Landon Mossburg
3. How Grid Batteries Got Cheap
Peak Energy's first product is a battery energy storage system, the grid-scale box the industry abbreviates to BESS. Mossburg dated the first attempts on the grid to about fifteen or sixteen years ago, with lead-acid banks before that — small, expensive to build and operate, and unpleasant to work around.
Lithium-ion raised the energy density of rechargeable cells, and stayed in consumer electronics until cars arrived.
The first electric-vehicle packs were consumer cells
We took basically laptop batteries and put thousands of them in car packs.
Landon Mossburg
The volume a Model S required doubled industry capacity quickly, then doubled it again, and the learning rate brought the price down. Before that, putting lithium-ion on the grid had been considered and dismissed.
His description of the old cost problem
it would be like using diamonds as body armor — just not practical, because you need so much
Landon Mossburg
Falling prices made small grid trials viable, then real products. Tesla entered early with Megapack, and Mossburg made a claim about where that business now sits in Tesla's accounts.
Where he says Tesla's profit came from
pretty much all of Tesla's margin, profit, last year came from Megapack — they made almost no profit on cars
Landon Mossburg
He flagged that this is not widely understood outside investors who follow the company closely, and said Megapack is very similar to what Peak Energy does.
4. Batteries as Market Makers
Most grid storage is not paired with generation at all, Mossburg said. A developer finds a node on the network with congestion, where transmission capacity is short of demand at some point in the day and price volatility is therefore high, and sites a project there.
The core operating strategy
There's a few other ways to operate a battery, but the main way is you just buy energy when it's cheap, and you sell it when it's expensive.
Landon Mossburg
The host's translation was that the operator is a market maker in financial terms, storing when cheap and releasing when expensive, which Mossburg accepted.
5. Why Solar and Storage Split
The question Mossburg said the industry should be asking is why hybrid solar-and-battery sites are rare. His answer starts with electricity and ends with construction finance.
Solar panels produce direct current. Batteries charge and discharge in direct current. Alternating current is what the grid transmits, and direct current cannot travel far without heavy losses or heavy spending, so inverters sit throughout a solar field converting close to the panels. Every conversion costs an inverter to buy and maintain and loses a few percent of the energy.
The efficient design connects the battery straight to the panels on the direct-current side. Almost nobody builds it. Instead developers build an alternating-current solar field and a separate alternating-current storage project, and pay the site-level costs twice.
Why paying twice is the rational choice
So you'd rather pay that twice and have much more certainty that if one of these things is delayed, the other can still make money.
Landon Mossburg
Integrating both on site means neither earns anything until all of it is finished and tested. A month of delay on the battery is a month of construction financing on the whole project, which he said wipes out the savings from fewer inverters and better round-trip efficiency.
6. The Cooling System
There is a deeper reason the two cannot be combined in a factory, and it is the constraint Peak Energy was founded to remove. Grid batteries have to last at least twenty years outdoors, and each of the company's smallest units stores enough to power hundreds of homes for four hours, which generates a great deal of heat.
The design every competitor shares
So every other battery, before us, on the grid, has to be actively cooled — literally uses water cooling.
Landon Mossburg
Running that system consumes about 5% of the energy stored in the battery. It is also the part that fails.
The failure concentration
It's the stuff that breaks — about 90% of the failures in the field are caused by components related to the cooling system.
Landon Mossburg
The fire record points the same way.
And the fires start there too
80% of the battery fires you've read about, if you've read about Moss Landing or any of those, are generally initiated by the cooling system failing, which is intuitive, because it shorts usually
Landon Mossburg
High voltage plus a water leak produces a short, and the short produces the fire. The system is also expensive and loud, because large fans have to exhaust the heat. And it is what makes factory-integrated solar and storage impossible.
The arithmetic that rules out the combined product
putting thousands of air conditioners through your solar field, or tens of thousands, is a great way to go bankrupt
Landon Mossburg
A solar panel needs no air conditioner. A battery bolted to the back of every fifth panel would need one.
7. A 5x Cheaper Battery
Mossburg described Peak Energy's philosophy as removing every constraint that adds cost, complexity, safety risk, financing uncertainty or deployment time and is not inherent to the useful component itself.
What the useful component alone would cost
that theoretical cheap battery would be five times cheaper than what we currently pay, the best projects
Landon Mossburg
Everything stacked on top of the cells is where the money goes. The host offered a friend's line that all value creation is compression; Mossburg preferred the inverse framing, which is to take complexity out.
The company's stated objective
And so that's the mission of Peak — basically drive the cost of energy down to its fundamental component cost limit, both cost in terms of money and in time.
Landon Mossburg
That defines what the company does not do.
The division of labor he chose
Which means we don't focus on trying to make a better solar panel, or trying to make a better battery cell — we try to make better energy infrastructure out of these components
Landon Mossburg
Cell and panel manufacturing are far more capital-intensive, he said, and there are good companies already driving those costs down. He drew the parallel to his old employer: Tesla did not invent batteries or electric motors, and worked on them only when it had to push the industry.
8. Wright's Law in Batteries
Mossburg's reason for thinking the outcome is settled is the learning rate on the three technologies he named — batteries, solar and, to a degree, wind.
The definition he gave for a general audience
every time you double the amount of production, the rate of learning is how much the price falls by
Landon Mossburg
The host identified it as Wright's law and made the investor's case for why it matters: power-law growth combined with power-law cost improvement compounds into something unusual.
The rate he puts on batteries and solar
If you look at batteries, solar and wind, they're — batteries and solar are enjoying really high learning rates, I think they might be as high as 20% or something like that.
Landon Mossburg
He gave the price series for a prismatic lithium iron phosphate cell, the legacy technology Peak Energy competes with: about $50 per kilowatt-hour now, roughly $100 five years ago, about $180 five years before that, and about $300 five years before that. Panels are now cheap enough that a developer routinely overbuilds a solar field by 20% to 30%, so the rest of the infrastructure is fully used even in weak light.
His framing of the difference in rates was that oil and gas improve more slowly than a Treasury bill and that solar and batteries look like the best fund performance ever recorded.
9. Gas Lost Its Price Lead
Mossburg pointed at the published cost-of-energy comparisons and said the ranking has changed. Before the data-center boom, gas was the cheapest thing in the United States, with solar plus batteries about 10% behind on many projects and ahead in some parts of the world.
The learning rate on the incumbent
And you look at the learning rate on gas-fired power, it's like 1%, if that, maybe five
Landon Mossburg
That leaves an inflection: a developer familiar with both technologies who has the space will pick solar and batteries, because it will be cheaper. Then supply caught up with gas.
The manufacturing bottleneck he named
I think there are only two locations in the world right now that can forge turbine blades
Landon Mossburg
Cheap gas-fired power in the United States was below $50 a megawatt-hour, he said, and is now around $80, with some behind-the-meter deals above $100 — substantially above batteries plus solar. On his own arithmetic, taking today's solar cost and adding only the cell plus a 20% pack adder and no site-level costs, the combination has gone from roughly level with gas to about 30% cheaper.
He named three reasons the switch is not complete: physical space, because panels need a great deal of it; policy that does not track the economics, with wind as his example; and geopolitics, since almost all solar manufacturing and the majority of battery manufacturing happens in China, which he called a justifiable worry for every other country.
10. Fusion Aims Too Low
The host raised the number of publicly traded fusion and small modular reactor companies and called their valuations a form of bubble at times. Mossburg's objection was about the target rather than the technology.
Every optimistic projection he has read aims at the price of cheap gas, with the most aggressive claiming five years and most research suggesting ten at the earliest, or never.
The problem with aiming at gas
And the problematic thing about that is that they're aiming at a target that today batteries and solar are approaching.
Landon Mossburg
Which makes the timeline the flaw
So if they get there in five years, they're too slow.
Landon Mossburg
He said he watched the same mistake at Tesla. A year after the Model S got traction, most other manufacturers launched a competitor designed to match the Model S on price and specification — and from inside the company he knew Tesla would be two or three revisions ahead by the time those cars shipped.
His broader point was about the pace energy investors are used to. Batteries and solar are driven by consumer manufacturing and behave more like semiconductors than like construction; gas does not, because it is construction, and factories scale better than building sites.
11. Why Nobody Noticed
Asked why this is not more widely discussed, Mossburg gave a plain answer.
His explanation for the silence
I just think people don't like to talk about — it's energy, it's not a sexy topic, it's something that people don't want to spend time on.
Landon Mossburg
Inside the industry the price trend is well known. What slows the response is silos: developers who know oil and gas do that work, developers who know renewables do that. Add change management and the fact that the battery half of the pairing only became cheap recently. Solar without storage produces a steep evening demand ramp and a different grid problem.
The host said one house in thirty in his Florida neighborhood has solar and that this had always puzzled him.
The residential comparison
Well, residential solar is wildly more expensive than utility solar — wildly, wildly.
Landon Mossburg
The panels cost slightly more at that scale, but the gap is customer acquisition cost and the risk of spending on a deal that never closes, plus the per-roof construction and its delays. The route to fast improvement, he said, is large scale with minimum barriers, and as much integration moved into the factory as possible, so that field work is closer to setting the product down.
12. Astronaut to Batteries
Mossburg said he was not interested in batteries until about ten years ago. His first role at Tesla was connectivity, managing the relationships with the companies supplying data and software to the car, including AT&T and Google. He then wrote software, which brought him close to the factories.
Why he went there at all
But actually, I went to Tesla in the beginning not because I was interested in EVs — I'm not even really a car guy, either.
Landon Mossburg
He joined in 2013 hoping to meet Elon Musk and move to SpaceX, because he wanted to be an astronaut. It did not happen, and he told the host he tells his children he is fairly sure he will still get a shot and certain they will.
The route into the subject
manufacturing was my gateway drug to get into batteries
Landon Mossburg
Factory work in his last year at Tesla is what he enjoyed, so he joined Northvolt to work on battery factories, spent four years in Sweden bringing up the first pilot line, then led the North American expansion. Europe was behind the United States on grid-scale storage, so the American market is where he first saw the gap: storage on the grid cost five to ten times the cost of the cells doing the work. On a car, by contrast, the pack around the cells adds roughly 10% — $10,000 of cells in a pack that costs perhaps a thousand dollars to build around them. Hardware is now around 20% of a grid project, he said, and much of the rest is financing, construction and operating cost, because the systems consume auxiliary power and break often.
13. What Batteries Do for Grids
Without storage, electricity has to be consumed the instant it is generated. The host said the grid working at all still feels like magic to him: switching on a light requires a proportional injection of energy somewhere, immediately.
The mechanism that has always handled it
And the only reason that works is because we grew the grid by using inertial generation — we spin rotors to generate energy, with thermal, like coal and stuff, that's all they're doing.
Landon Mossburg
Flipping the switch raises resistance somewhere, pulls on a spinning rotor and slows it by an infinitesimal amount, and physics does the balancing with no intelligence involved anywhere. As the number of spinning rotors falls, batteries take over the job, and they are more controllable and faster to respond. Mossburg was careful about the trade: replacing the rotor means adding engineered complexity to solve something physics currently solves for free. He also said browning out is now avoided daily in places because storage is present.
14. A Megawatt per Rack
The change Mossburg said is happening regardless of his company comes from data centers. Tesla and xAI got there first, because Musk did not want to wait for grid interconnection and bought gas turbines to site next to three xAI data centers in Mississippi.
Running AI workloads exposed the same inertia problem. A training run has a power profile that swings hard, because training has to stay roughly synchronous, and energy density per rack is climbing fast.
The scale in a single rack
one megawatt of continuous power is about 800 to 1,000 average homes
Landon Mossburg
He said a product has already been announced with about a megawatt per server rack, and that the next generation needs a separate cabinet beside the rack purely for power supplies. Those swings are the danger.
What xAI worked out before the rest of the market
So they realized very early on — a year ago, or two years ago — they have to put batteries there.
Landon Mossburg
Gas turbines respond badly to inertial demand, he said, and the swings tear them up.
The scale of what they built
They've installed 4 gigawatt-hours of batteries at three data centers in the last year and a half.
Landon Mossburg
Unbuffered swings destroy downstream equipment and brown out the grid, which is why, on his account, grid operators now require a mitigation as a condition of connection: "you can't connect to the grid unless you provide a solution to protect the grid from those spikes." Other solutions exist and he said all of them are worse and more expensive than batteries. He described single-site deals on the order of 50 gigawatt-hours, most of them not public.
15. The Case for Abundance
The host said the rate of change in aggregate is hard to hold in one's head, and quoted Al Bartlett: "the greatest shortcoming of humanity is our failure to understand the exponential function." He asked what the positives are.
Mossburg placed himself on the conservative side of AI optimism: large productivity gains, a move into physical systems inside a working career rather than in thirty years, no intelligence explosion next year. In that world the binding cost is raw materials and energy.
Why he thinks the pieces are already in hand
We don't need that — we have the technology today, that is the hopeful thing.
Landon Mossburg
The alternative he was describing is a future where the only escape from a higher-emissions energy build-out is a material-science breakthrough making fusion cheap, or small reactors somehow reaching economies of scale nobody can yet describe.
The nature of the remaining work
That's a systems engineering problem, it's not a material science or physics research problem — it's an engineering problem.
Landon Mossburg
That converts a hard-to-quantify research risk into scaling work, which he was clear is very hard but is work. His reason for staying with it, having said there are easier ways to make money, is what energy is attached to.
The macro link
When energy prices increase, inflation goes up.
Landon Mossburg
Countries without reliable access to their own energy tend to become unstable, go to war, or lose wars. He cited a history of the oil industry, which he did not name, for the argument that oil is much of why Japan entered and lost the Second World War, and named Churchill's move to take the Royal Navy off coal as one of his best decisions before becoming prime minister.
16. Why He Bets on Solar
Every fuel traces back to the same source, Mossburg said: heavier elements were made in a star, and burning oil is burning stored sunlight that went through several intermediate processes. Photovoltaics skip the intermediaries.
The reactor already in place
Yeah, and we just harness the fusion reactor in the sky — it just works.
Landon Mossburg
It should keep working for a few billion years, which he said is hard to beat. He was complimentary about fusion as a technology and doubtful about it for terrestrial power.
His long-horizon call
But I would not bet against solar. I think solar is going to win.
Landon Mossburg
He attached a warning for investors: a physics-correct decision about energy may not pay in capital markets, because the politics are important right now and highly volatile. Asked about mirrors in orbit, which the host named Reflect Orbital, he said there are risks worth thinking about and that it is worth trying. He also made the case for putting AI compute at Lagrange points, where constant sunlight removes the need for a battery at all.
Why the company defines itself broadly
we're not a battery company, we're not even really an energy storage company, we think of ourselves as an energy company
Landon Mossburg
One of Peak Energy's stated values is to face the brutal truth, he said, with no sacred cows on product or direction. The mission is the cheapest dispatchable electron, and if orbital mirrors turn out to be the route, the company will help that happen.
17. $1.1B of Reservations
Asked what help he wants, Mossburg said hiring first: anyone in the energy industry, any good engineer, anyone motivated to work on the problem. Then he gave the numbers.
The order book and the ramp
We're going to do $10 million in revenue by the end of our fiscal '26, in a few months here, and then $100 million next year — we are scaling, we're moving.
Landon Mossburg
Behind that sits committed demand: "we have $1.1 billion in customer-backed, deposit-backed reservations." The company is three years old, connected its first system to the grid a little over a year ago, and expects 4 gigawatt-hours of production capacity next year.
He described the technical proof as done and the commercial proof as the current job. People close to the field who have tried to remove active cooling find it hard to believe Peak Energy solved it, he said, and the next year is about crossing from pilot scale to full scale — financing, bankability and insurance, in parallel with scaling manufacturing. His closing ask was political as much as commercial: tell your representatives you care that energy is cheap, because countries with more efficient energy infrastructure win.
The cost of being 20% behind
I mean, if your energy costs you 20% more than your rivals' energy, it may not kill you — it's not going to kill you next year, you may not notice it, but you're going to definitely notice it on a 10 to 15 to 20 year horizon.
Landon Mossburg
The host's version was that if energy converts fairly directly into intelligence, then competitive advantage runs downstream of who manages energy well.
His amendment to that
Yeah, who can create it, who can store it, who can deploy it.
Landon Mossburg
18. The Wrong Machine Oil
Asked for his most memorable experience of quality, Mossburg went to the Swedish factory. Making a modern battery cell is, on his ranking, second only to making a leading-edge semiconductor in difficulty.
The comparison he drew
probably the only thing that's harder to make is a semiconductor, like a cutting-edge one
Landon Mossburg
The process moves from continuous, fluid, kiln-based steps into a high-speed assembly process with tolerances around ten microns. There are more than twenty steps, and a defect introduced at step two may only appear at the end, where the signal tells you something is wrong without telling you what.
During the pilot ramp about thirty people brought sleeping bags and slept in conference rooms in the factory for four weeks, then moved to hotels for another six months. Two months in, the line was running and then started producing heavy scrap with no identifiable cause. Nothing had changed and nothing they checked explained it.
The answer was a lubricant. A mechanical tool that cuts sheets has to be serviced every hundred thousand cuts or so, and the knife oiled. The oil the equipment maker specified was import-restricted and unavailable in Europe, so a maintenance technician bought a high-quality machine oil locally, as a routine part of his job, and told nobody. Root-causing it took roughly a hundred people about a month, with no battery production in that time.
The rule the episode ends on
So if you're doing mass manufacturing like that, one of the things you got to do is, if you change anything, you got to write it down.
Landon Mossburg
The first rule is not to change anything, he said, and the second is to write down what changed.
Bonus Insights
Culture is part of why Asia is better at tight tolerances
For very high-tolerance, high-speed manufacturing, Mossburg said, it helps to have a workplace culture in which people follow instructions closely and ask permission rather than take initiative. He thinks that is part of why Japan, China and to some extent Korea do this better than Western countries, and was explicit that the same trait costs them elsewhere.
The cost of change is what caps what gets built
The write-it-down rule is good discipline for a running line and bad for developing new products. Mass manufacturing is the only place the learning-rate curve shows up, Mossburg said, and it works precisely because the process gets into a groove and repeats. If AI lowered the cost of changing a process, he said, the range of things a society can make and the quantity it can make would change substantially.
A nuclear plant for one data center
On the scale of what these sites draw, Mossburg said a gigawatt of power is why hyperscalers are buying nuclear plants: the entire output of a plant can be consumed by one large data center. The host's own comparison point was his car, at roughly 70 kilowatt-hours, against 4 gigawatt-hours at the xAI sites.
He has been doing it almost constantly for three years
Mossburg said he has worked on this nearly around the clock for three years and still learns something new almost every day, generally from talking to somebody operating in the field. The host said the same thing is what keeps him sane about the frustrating parts of the job.
Mossburg's bottom line is that the cheapest energy is already solar paired with batteries, that the gap is engineering rather than physics, and that the single largest piece of avoidable cost in grid storage is the cooling system his company removed.
Products, Companies & Tools Mentioned
Peak Energy (Three years old, $1.1B of deposit-backed reservations, $10M of revenue expected this fiscal year and $100M next, 4 GWh of planned capacity, on a passively cooled grid battery)
Tesla (Where he worked from the start of the Model S; he says almost all of its profit last year came from Megapack rather than cars, and that Megapack is close to what Peak Energy builds)
Northvolt (Where he spent four years in Sweden bringing up the first pilot line, then led the North American expansion)
xAI (Bought gas turbines rather than wait for interconnection, then installed 4 GWh of batteries at three Mississippi data centers when AI load swings started damaging the turbines)
Nvidia (Its next generation needs a separate cabinet beside the rack purely for power supplies, as racks approach a megawatt)
Reflect Orbital (The host's example of orbital mirrors aimed at solar farms; Mossburg said there are risks but it is worth trying)
Lazard and the International Energy Agency (The published cost-of-energy comparisons he uses to argue solar plus storage now beats gas)
Moss Landing (The battery fire he names as the type initiated by a cooling-system failure)
If this was worth your time, send it to someone closer to the industry than you are.
Get the latest market chatter as it happens:
Liquid cooling causing most field failures makes sense once you picture thousands of cells plumbing water through them around the clock. Take the plumbing out and the leak risk is just gone. And that learning rate gap is wild, storage shaving costs 20% a year while gas barely moves 1%.