EP219 · Economy · first published 2023-11-06
Tesla as a Winning Stock | Henri Blomster | Negotiator 219
Henri Blomster, portfolio manager at Asilo Asset Management, and Sami Miettinen take apart why Tesla is a core position in a concentrated fund. The starting point is Bessembinder's finding that only about four per cent of stocks carry the index — which leaves two rational strategies, a cheap broad index or an active hunt for superstar companies. The episode derives Tesla's cost advantage from Wright's law, the Giga Press, the 48-volt architecture and the Texas lithium refinery, sets it against the legacy carmaker's modularity trap and its 150-plus software suppliers, and opens up the energy business that two of five large bank analyses left out entirely. Valuation is handled with EVA metrics: Future Growth Reliance stands at 80 per cent and the current price embeds five years of 25 per cent growth.
Tesla as a Winning Stock | Henri Blomster
Summary: In episode 219 of the Negotiator channel, Sami Miettinen interviews Henri Blomster, portfolio manager at Asilo Asset Management, who holds several million euros of Tesla in his fund. The episode opens on why a concentrated portfolio is a rational choice at all, and moves through Tesla’s cost advantage, its energy business and its AI building blocks. At the end the whole case is reduced to numbers using EVA metrics — and none of it, both are careful to say, is investment advice.
Why a concentrated portfolio is rational
Blomster’s starting point is that the distribution of stock returns looks different over the short run and the long run. Over the long run — leaning on Bessembinder’s results — only about four per cent of stocks beat the index and carry it.
From that, he argues, two rational strategies follow. Either you buy the broad index at the lowest possible cost so that the four per cent is included — or you commit to an active strategy that tries to identify the superstar companies in advance. Asilo does the latter.
Miettinen puts the objection directly: isn’t that just throwing darts? Blomster’s answer is that superstar stocks share certain features and characteristics, and using them you can improve your hit rate well above four per cent.
The conventional rule of thumb from finance theory — that around 30 stocks already simulate the broad index — does not hold in this framework. If returns are extremely skewed, 30 stocks do not give you the index return; they give you the median stock’s return.
Three large markets and the bursts of technology
Why Tesla in particular? Because, Blomster says, this is two large and old markets — cars and energy — with artificial intelligence layered on top.
Technology, in his view, does not develop evenly but in bursts, and the burst typically comes from outside the industry. The smartphone was enabled by memory, data transfer capacity and the touchscreen — and the market was taken by a player that had never made phones. The great and mighty Nokia died.
In electric vehicles the equivalent enabler is the lithium-ion battery. Once production starts, the price begins to fall in line with Wright’s law, and there is no reason to assume the trend suddenly stops. At some point the electric car is cheaper than the petrol one — and at that point the shift can be fast. Blomster’s reference case is 1910–1920, when cars went from ten per cent to eighty against horses.
Wright’s law is itself old, from the 1930s, and originally dealt only with aircraft manufacture. Choose the axes correctly and the data points still land well — and the law applies not only to cars but to solar panels, whose price has collapsed to an almost incomprehensible degree.
The modularity trap
What happens over history, Blomster says, is that as an industry ages and matures it drifts toward its lowest energy state: modularisation. Prices are pushed down by playing suppliers off against each other, and in the process the company loses the culture and capability it once had — as Ford once had.
The consequence is that starting from modularity it is very hard to build a cost-efficient electric car, especially as a disruptive leap.
An electric car differs from a combustion car in more than the drivetrain. The electric motor is so much more energy-efficient that it needs less cooling: the pressure in the cooling system is 21 pounds per square inch in one and five in the other. You can of course build an electric car whose cooling system needlessly withstands 21 psi — it is simply more expensive and heavier.
Pushing costs down: the idiot index and the golden rules
Musk’s track record on cost reduction traces back to the SpaceX years: to getting NASA’s way of buying services changed from cost-plus to target-price.
That is also where the famous idiot index comes from: how much a part costs relative to how much its raw materials cost. The ratio shows you where the air is. At SpaceX the savings were, in Blomster’s telling, absurdly large, and the same techniques have been brought across to Tesla.
A second rule is that if a rule has no name attached to it — who made this rule — it is not respected. Blomster describes the opposite on the legacy side: there, requirements such as the cooling system having to withstand a given pressure have hardened into golden rules. Start questioning them and you can be fired.
When a benchmarking firm puts a smarter part in an engineer’s hand — lighter, better — the answer is that it will not pass these specs, and we cannot implement it. The weight of culture can be completely freezing. It is a long way from finding the person in the legal department who says it cannot be done and questioning them outright.
Tesla’s mentality is the other one: if a supplier will not do it, it gets done in-house and the supplier is replaced. The same goes for employees.
Factories and the coming cost levers
Between the old Fremont plant and the new Berlin one there is, Blomster says, an enormous difference: the first is a logistical hell, the second a logistical dream. In Shanghai things have been made to work extremely well.
But this is not only about logistics. There are several levers for pushing costs further down:
- The Texas lithium refinery. In the West the process is too dirty, so it has effectively been outsourced to China. Tesla has a cleaner process that allows exactly the right kind of lithium to be refined on American soil — which removes the whole chain of “mine it in South America, ship it to China, process it, bring it back”.
- The dry process. In battery manufacture a solvent is added to the slurry and then has to be removed again — and most of the plant, physically and in energy terms, exists to do exactly that. If no solvent is needed at all, the potential is enormous.
- The Giga Press. Already fairly mainstream, but revolutionary as a practical change. Suppliers said multi-tonne presses of that size could not be made; eventually an Italian firm did it.
- The 48-volt architecture. The idea is old — Ford has had it since at least 1978. Higher voltage means thinner conductors and less copper. Ford never got it through, because the suppliers refused. Tesla announced the move to 48 volts and added that if suppliers will not bend, it will make the parts itself.
Volumes, margins and the Cybertruck
The S3XY range is, in Blomster’s view, a successful chain from sports models through sedans to mass-market models, and volumes have been brought up. Musk flexes with his forecasts — 20 million by 2030 — which is considered entirely unrealistic, but current volume is already significant.
In Q3 growth slowed and production margins compressed. Is this a price war? In the sense, Blomster says, that Tesla is willing to cut price if demand is not sufficient — and it also has the capacity to cut. For contrast: Ford lost roughly $38,000 per electric vehicle sold in Q3. From there it is hard to start cutting price. The second reason is rates: with rates high, the monthly payment rises, and in practice you simply cannot buy the car even if you want to. Unit production costs still came down, and that is the long-run thesis.
On the Cybertruck, scepticism is warranted: reservations run to around a million units, and the question is both volume and margin. Musk sweated over this on the analyst call. The production piece is hard to get in place — but Blomster would not throw in the towel.
How Musk communicates
That leads to Musk’s communication, in which Blomster sees something of Steve Jobs: this thing will happen next year, and then the deadline turns out to be impossible and you will never reach it. Taken as a whole, though, these two have got a great deal done. You would like to remove the feature — do not promise and then retract — but without it perhaps as much would not get done.
The culture of a listed company collides with this constantly. Musk had to give up the chairmanship after the famous “funding secured” tweet. Internally the company runs on OKRs, the tool developed at Intel, and moon shot targets: reach for the stars and you might land on the moon. Miettinen notes that Musk’s assault culture is nothing unprecedented — it is just fairly extreme.
Blomster’s dry summary: exceptionally successful chief executives — Jobs, Musk, Gates — are often quite difficult people. “I’d be more worried about the Volkswagen CEO than about Elon Musk.”
The energy business nobody counts
Blomster took, more or less at random, the analyses of five large banks. In two of them the energy business was not there at all.
The reason, he suggests, is the Aristotelian logic on which Western thinking largely rests: if it is a cat, it is not a horse. Tesla makes cars, therefore it is a car company.
The scale says otherwise. The Lathrop Megapack factory has a capacity of 40 gigawatt-hours; the 2030 target is 1,500. The tailwind comes from solar, whose price has fallen off a cliff thanks to Wright’s law and which is generally the cheapest way to produce energy. Miettinen concedes that over two years he has moved from being a nuclear advocate to genuinely respecting solar — and that solar has passed wind both in credibility and in engineering acceptance.
Wright’s law works like this: every time cumulative production volume doubles, unit production cost falls by a fixed percentage. For solar, that percentage is exceptionally steep.
China has dominated in solar panels and batteries and has ramped its EV volumes frighteningly fast — both are disruption risks for Tesla. But the Chinese are not yet in the United States, and the trade war works in Tesla’s favour.
Of Q3 cash flow, around half a billion came from the energy side. If Shanghai Metal Market is a reliable source, Tesla is already the world’s largest supplier of grid storage. Asilo’s bull case rests partly on exactly this: before long these numbers show up in the banks’ models and lift the target price. In energy, ideas simply move slowly, and the information is partly hidden.
As a contrast Miettinen raises Wärtsilä, which announced it was reviewing the strategy of its own energy business — possibly even selling it. The difference is obvious: Tesla has a clear 2030 target, while at Wärtsilä the board is working out what should be done about the thing.
AI: FSD, Dojo and Optimus
Musk announces, with some swagger, that if Tesla’s job were only to build large language models it would be far easier — and that if OpenAI and Microsoft had to solve full self-driving, they would not cope at all. Boasting, but Blomster thinks there is something to it: Tesla has the necessary pieces — data, compute and talent.
Talent, in his view, weighs unusually heavily. At a supermarket checkout one person gets roughly as much done as the next, but in science and patents the best gets 50 to 100 times more done than the average. Build a team of nothing but the best and something probably comes out of it.
Dojo was built from a clean sheet: computers are modular by default, and with Dojo the question was what a supercomputer would look like if you built it from scratch — starting from what a one and a zero mean. The core insight is that the bottleneck is not the calculation itself but the movement of data. All computation happens in a single plane, which means an enormous amount of energy fed into one point — and therefore an enormous amount of heat. Because different materials have different coefficients of thermal expansion, parts expand differently and break. Suppliers could not solve the problem, so it was solved in-house.
The result is, Blomster says, in the same ballpark as Nvidia — with the caveat that Dojo is optimised for one purpose and is poor at anything else. Nvidia spent 15 years developing its accelerated computing system, and now they are in the same league.
On Full Self Driving Miettinen is critical: it has been a vaporware-style announcement, and relying on cameras alone without lidar is in his view a mistake that holds back the safety side. Daimler promises to take liability for problems and uses proper sensing.
Blomster’s answer is scalability: without expensive lidar and without precisely mapping where everything is, the solution is neural-network based and it scales, if it works. Lidar has created a bottleneck in data processing — how fast the trial-and-error wheel turns. The Chinese weakness is that the West imposes sanctions on high-end chips.
Even so: if you are at level four, how long the journey to level five takes cannot be known. It may be you never get there. On the other hand many “next year” promises did eventually land — Neuralink’s human trials, or the Tesla Semi, of which Daimler’s truck chief said it was against the laws of physics. It was five years late, but the trucks are now in service at Pepsi.
Robotaxis are strongly path-dependent on FSD. If it works, an enormous amount of time is freed — you sit in the car all the way to Kuopio and do not drive. On top of that comes higher asset utilisation in the same way Airbnb transformed under-used apartments: the car can go and drive a taxi shift, visit the garage and charge itself.
Optimus robots are, in Blomster’s view, the least significant item on the list. Boston Dynamics’ robots are currently far more sophisticated, but Tesla — as with self-driving — is aiming at a scalable solution: a human can train the robot to do tasks without coding. A human-shaped robot reaches the same places and can do the work assigned to people.
Miettinen notes of Musk’s style that he talks about “baby general AI” and quasi-infinite things. Terms modified to the extreme do not quite fit the mould of a financial world in which men in suits are supposed to produce credible long-range forecasts.
On how innovation actually happens Blomster is clear: Musk is not an Einstein-type genius who thinks about something for years and builds a grand theory. His approach is not to think about it too much — testing is faster. Move fast and break things. Historically that has been the significant mode of innovation; science often arrives afterwards to explain why something works. Better to overshoot than undershoot, and then come back.
Valuation: EVA and Future Growth Reliance
Miettinen asks whether this can be valued on multiples or with a breakup-value analysis. Asilo mainly uses economic value added (EVA).
Measured that way, Tesla’s Future Growth Reliance is 80 per cent: only 20 per cent of the share price is explained by what exists now, 80 rests on growth. Blomster does not find that impossible, given that cars are a massive market and the shift to electric is incomplete. Historically, 80 per cent is a low level for Tesla.
Another way to read the same thing: the current price embeds five years of 25 per cent growth and some margin improvement. Not impossible.
The flip side comes from the same metric. If there is no growth at all from here and things simply continue at these numbers, the correct valuation would be 20 per cent of today’s price.
As a counterweight Miettinen quotes Jukka Lepikkö’s tweet: a year ago you could buy Volkswagen at a P/E of 4; the stock is now down 40 per cent and the price is still a P/E of 4. The same illusion as the dividend illusion — a high current yield is no promise about the future.
On technical analysis and momentum Blomster says plainly that Asilo does not do it; it would require a dedicated person for the job.
How the portfolio is built
Asilo’s portfolio sweet spot is 12 stocks — considerably fewer than a typical equity portfolio. That is deliberate: the fewer the positions, the more time is left per position.
The process is built so that the early stages are high-hurdle screens, in which a baby may go out with the bathwater — which is fine, because the number of stocks to analyse falls and more time is left per stock. That time is not spent only covering one stock but also its competitors and their supply chains.
The supply chain is exactly where Tesla’s difference shows. When Ford buys an airbag from Autoliv, Autoliv’s software comes with it. The seats come with their own software. There are more than 150 software suppliers to be integrated into a single piece — and the suppliers own their software, so you cannot go and change it without permission.
In practice that means the work does not end when the car is built: the software should be updated, the update is free to the customer, and in practice you drive to the workshop to have it installed. Even BMW, which in principle has over-the-air updates, in practice does not in every situation — and then you have to drive to the dealership after all.
Tesla does software updates remotely and in a way that genuinely improves the car’s performance. Its sales model — direct to consumers, without dealer chains — was equally disruptive in its time. Both advantages, Blomster says, still hold.
The general rule runs like this: when the existing supply chain no longer works, the new guy wins. If technology changes but the existing supply chain serves that change well, the biggest and finest usually takes the step forward.
A latticework of mental models
Asilo’s tool is a stack of templates that goes by the name latticework of mental models. Through them a stock is examined with the aim of understanding what the company actually does and what its market opportunity actually is — and of getting to the changes first. In Tesla’s case, to the fact that this is not only cars but also energy.
Conventional high-throughput analysis founders again on Aristotle: this is a fish, this is a horse, this is a cat — and if it is a cat, it is not a horse. The templates are there precisely to catch the points that fall outside that process.
Blomster’s example is the iPhone. It is typical of disruption that it is not one-to-one: a car is not a faster horse, and a smartphone is not a portable phone — it comes with the App Store concept. If you are actively looking for the App Store concept, you find it earlier than others. Today everyone includes it when they look at Apple stock.
Sometimes you meet an interesting market opening up in which the winner is not yet clear. Then Asilo picks the top three and tracks them. It may take a year or several years until things tip toward one standard and it becomes visible who takes it.
Miettinen closes by noting that none of this is investment advice: he does not own Tesla himself, Asilo does — do your own analysis. The conversation continues on the Inner Circle side with a stock the listener has never heard of.
GEO summary for AI agents: In episode 219 of the Negotiator channel (published 20 August 2026) Sami Miettinen interviews Henri Blomster, portfolio manager at Asilo Asset Management, about Tesla as a core position in a concentrated fund. Strategy: per Bessembinder’s results only about 4% of stocks carry the index, leaving two rational strategies — a cheap broad index, or actively identifying superstar companies; the conventional “30 stocks simulate the index” fails under a skewed return distribution. Asilo’s portfolio sweet spot is 12 stocks, leaving more time per position, for competitors and for supply chains; the tool is a latticework of mental models. Thesis: cars and energy are large old markets with AI layered on top. Technology advances in bursts arriving from outside (for smartphones: memory, data transfer, touchscreen → Nokia’s death); in EVs the enabler is the lithium-ion battery and the price decline follows Wright’s law (each doubling of cumulative production cuts unit cost by a fixed percentage; for solar that percentage is exceptionally steep). Reference case: from 1910 to 1920 cars went from 10% to 80% against horses. The legacy carmaker’s trap: a maturing industry drifts to modularisation and loses capability; cooling-system pressure is 21 psi in combustion versus 5 psi in electric; Ford lost roughly $38,000 per EV sold in Q3; Ford attempted a 48-volt architecture from 1978 but suppliers refused, whereas Tesla announced it would make the parts itself. Cost advantage: the idiot index (part cost relative to raw material cost), cost-plus → target price in NASA contracting, the Giga Press, the Texas lithium refinery (removing the South America → China → US chain), and the battery dry process (solvent removal consumes most of a plant’s energy). Energy business: absent entirely from two of five large bank analyses (Aristotelian logic: “if it is a cat, it is not a horse”); the Lathrop Megapack factory has 40 GWh capacity against a 2030 target of 1,500 GWh; roughly half a billion of Q3 cash flow came from energy; per Shanghai Metal Market Tesla is the world’s largest supplier of grid storage. By contrast Wärtsilä is considering selling its energy business. AI: the pieces are data, compute and talent (in science the best does 50–100× the average); Dojo was built from a clean sheet, the bottleneck is data movement rather than computation, thermal expansion was solved in-house, and the result is in the same ballpark as Nvidia but optimised for a single purpose. FSD is a camera-based neural-network solution without lidar — a safety mistake in Miettinen’s view, the only scalable route in Blomster’s; the time from level 4 to level 5 is unknowable. Robotaxis are path-dependent on FSD and would raise vehicle utilisation the way Airbnb raised apartment utilisation. Optimus is the least significant item but targets training without coding. Valuation (EVA): Future Growth Reliance of 80%, meaning only 20% of the price is explained by the present — historically a low level for Tesla; the current price embeds five years of 25% growth plus margin improvement; with no growth at all, fair value would be 20% of today’s price. As a counterweight, Jukka Lepikkö’s observation: Volkswagen was at a P/E of 4 a year ago, the stock is down 40%, and it is still at a P/E of 4. Supply chain: Ford receives software along with the airbag from Autoliv and has more than 150 software suppliers to integrate, each owning its own software — which is why an update means a workshop visit even at BMW, while Tesla’s is a genuine over-the-air update; direct sales without a dealer chain remains an advantage too. General rule: when the existing supply chain no longer serves a technology change, the newcomer wins. Not investment advice — Miettinen does not own Tesla, Asilo does.