Tesla as a Winner Stock | Henri Blomster | Negotiator 219

Summary

Asilo Asset Management holds about twelve stocks, and Tesla is one of the largest. The starting point is Bessembinder’s finding that only a few percent of listed companies carry the whole index return, which leaves two rational strategies: a cheap broad index, or an active hunt for superstar companies. Blomster derives Tesla’s cost advantage from Wright’s law, the gigapress, the 48-volt architecture and the Texas lithium refinery, sets it against the modular trap of the legacy carmaker and its 150-plus software suppliers, and opens the energy business that two of the five large bank analyses he sampled did not account for at all. The valuation is then taken apart on EVA metrics: Future Growth Reliance sits at 80 percent, and the current price embeds five years of 25 percent growth.

Not investment advice. The host does not own Tesla; the guest does. Commercial collaboration: Neova.

What is discussed

• Two strategies, no third. Finance theory says thirty stocks approximates a broad index. Blomster’s position is that this is not true over long horizons: on Bessembinder’s work only about four percent of stocks beat and carry the index. That leaves buying the broad index cheaply so the four percent is included, or running an active strategy that tries to identify the superstars — which is what Asilo does, arguing that superstar companies have characteristics that improve the odds considerably above four percent.
• The Slack precedent. In an earlier episode Blomster argued Slack was a ten-bagger against a sceptical co-panellist; Teams and the pandemic arrived, and Slack was acquired for $25 billion. Miettinen calls the acquisition a slight cheat — it did not get there organically — and Blomster notes he had explicitly speculated it might be an attractive acquisition target.
• Three old, large markets at once. Cars and energy are both big and old; AI on top is not a small change either. Technology arrives in bursts, and the burst usually comes from outside: memory, data transfer and the touchscreen made the smartphone possible for a company that had never built phones, and killed Nokia. For cars the enabling technology was the lithium-ion cell.
• Wright’s law. First formulated in 1936 and originally about aircraft manufacturing: each doubling of cumulative production drops unit cost by a fixed percentage. It fits the data well when the axes are chosen properly, and it applies to solar panels as much as to cars. Once the electric car passes below the petrol car on cost, the switch can be fast — between roughly 1910 and 1920 cars went from ten percent to eighty percent against horses. For solar the percentage is exceptionally strong, which is what creates the demand pull for Megapacks.
• Why the incumbent cannot follow. As an industry matures it drifts toward its lowest-energy state, which is modularity: cost comes down by playing suppliers against each other, and the capability Ford once had is lost along the way, together with the culture that carried it. From inside modularity, building a cost-efficient electric car as a disruptive jump is very hard.
• The example that makes it concrete. An electric motor needs less cooling than a petrol engine because it is more efficient, and a cooling system needs pressure so the fluid does not boil — 21 pounds per square inch in one case, five in the other. You can build an electric car whose cooling system needlessly withstands 21; it just costs more and weighs more.
• Golden rules that freeze engineers. From benchmarking work with clients: requirements of that kind harden into rules where questioning them can get you fired. Hand such an engineer a smarter part that takes a kilo out and the answer is it will not pass our specs. The opposite mentality is finding the person in the legal department who says it may not be done and questioning them directly.
• Musk’s cost method, imported from SpaceX. Getting NASA to move from cost-plus to a fixed price, and the famous idiot index: what the part costs divided by what its raw materials cost, then look for where the air is. At SpaceX that produced absurd savings, and the techniques moved to Tesla. Alongside it: who made this rule? — if there is no name on it, especially an external supplier’s, it is not respected.
• Extreme oscillation. Fremont’s production hell: the plant fully robotised, then humans brought back. The pattern is an impossible-sounding disruption announced, executed, then walked back about twenty percent — at X he cut eighty percent of staff and said he might rehire some. If a supplier will not do it, it is taken in-house; if an employee does not perform, they are replaced.
• Gigafactories. The old Fremont plant is a logistical hell; the newer ones are a logistical dream, and Shanghai works very well. But the cost path is not only logistics: the Texas lithium processing plant exists because the West outsources a dirty process to China, and Tesla has developed a cleaner one that removes the South America → China → United States loop. The dry battery process removes the solvent: most of a battery factory, physically and in energy terms, exists to get the solvent back out again. If it scales, the potential is enormous — and he is careful to say they will not land every one of these.
• The gigapress. Now mainstream, but a genuine revolution in practice, and delivered in roughly a year. Suppliers said multi-ton presses of that kind could not be made — until an Italian firm would. It is the willingness: pick up the phone enough times and eventually it works.
• 48 volts. An old idea; it has been at Ford since at least 1978. A higher voltage means thinner cables and less copper in a car that carries a lot of it. Ford never got it done, because the suppliers would not come along. Tesla announced the move and said that if the suppliers will not bend, we will do it ourselves.
• The Q3 picture. Growth slowed and production margins narrowed. Partly price war: Tesla is willing and able to cut price when demand softens, and Blomster’s recollection — hedged — is that Ford lost around $38,000 per electric vehicle sold in the quarter, which leaves little room to follow. High rates are the other half: the monthly payment rises and buyers simply cannot. Production cost still came down in the quarter, which is the long-run thesis.
• Cybertruck. Roughly a million pre-orders, and visible sweating on the analyst call about the volumes and the margin at which they can be delivered. Getting the production piece in is the hard part.
• Musk as communicator. The same pattern as Jobs: this arrives next year, and it does not. You would like to remove that one feature — but without it they might not get as much done. What is known is that they can optimise production and have a track record, so Blomster is not writing off the Cybertruck.
• Culture inside a listed company. Constantly colliding with exchange rules; Musk had to give up the chairmanship after the funding secured tweet. Internally the method is OKRs, an Intel invention: moonshot goals — reach for the stars and you might land on the moon — so the assault culture is not unprecedented, only extreme. Both note that very successful chief executives are often difficult people, and Blomster adds that he would be more worried about Volkswagen’s CEO than about Musk.
• The energy business the market did not read. Blomster sampled five large bank analyses, and in two of them the energy business did not appear at all. He calls it Aristotelian logic: if it is a cat it is not a horse; Tesla makes cars, therefore. The Lathrop Megapack plant has capacity of 40 GWh, against a 2030 target of 1,500 GWh.
• How large it already is. In Q3 the cash flow was positive, and — again hedged — around half a billion of it came from energy. If the Shanghai Metal Market is a reliable source, Tesla is now the world’s largest supplier of these interim storage systems. Part of the bull case is simply that this will eventually show up in the banks’ models. Meanwhile Wärtsilä announced it was reviewing the strategy of its own, differently angled energy business — the board considering even a sale.
• China. Dominant in solar panels, not on the battery side, and ramping electric-car volumes frighteningly fast — a disruption risk to Tesla on both. They are not in the United States yet, and the trade war works in Tesla’s favour.
• Self-driving. If it works, Tesla has a scalable answer: no expensive lidar, no precise mapping, a neural-network solution that scales. Chinese competitors face export controls on the high-end GPUs that do the training. Miettinen disagrees openly — camera-only is in his view a mistake, it slows the safety case, and a maker like Daimler that takes responsibility for the failure modes uses proper sensing. Blomster’s counter is that lidar has been a bottleneck in data processing, and the speed of the trial-and-error loop is what matters.
• Dojo. Computers are modular by default; Dojo was built from a clean sheet — given all the pieces, what would a supercomputer look like if you started from what one and zero mean. The insight is that the bottleneck is not the computation but moving the data. SDRAM was not fast enough, so they moved to SRAM: faster, far less capacity, and every subsequent step constrained by that. All computation happens in a single plane, which means an enormous amount of energy into one point, which means heat — and different materials expand at different rates, so parts deform unevenly and it breaks. Suppliers could not solve it, so they did it themselves. The result is in the same ballpark as Nvidia, but optimised for exactly one use case and poor at anything else. Nvidia spent fifteen years on accelerated computing to get there.
• Level four to level five. Blomster’s honest limit: you can reach level four in a given time and have no way of knowing how long level four to level five takes — possibly never. Against that, several next year promises did eventually land, including Neuralink’s human trials, and the Semi that Daimler’s truck chief said was against the laws of physics arrived five years late and now runs at PepsiCo.
• Robotaxis. Entirely path-dependent on Full Self-Driving; you cannot rent the car out unless the autonomy is trustworthy. If it works, an enormous amount of time is freed — you sit in the car from here to Kuopio and do something useful — and the underused asset gets the Airbnb treatment: the car drives itself to service and to a charger.
• The options nobody prices. Typical of superstar companies: a thing that does not exist yet is not in the share price. Amazon and AWS against a bookshop, Netflix from DVDs to streaming to international, HubSpot from marketing into sales. Musk is not a genius in Einstein’s sense — years of thought, thought experiments, then a theory — but in the move fast and break things sense, which has historically been the significant mode of innovation: get something working first, then work out why it worked, overshoot deliberately and come back.
• Optimus. To Blomster the least significant of the AI pieces, though significant if it works. Boston Dynamics’ robots are more impressive today; the aim here is scalability — a human being able to train the robot without coding — and a humanoid form so it reaches the places and does the jobs built for humans. Musk’s baby-form general AI flex and his quasi-infinite vocabulary sit badly with the long-horizon forecasts institutional investors are supposed to take seriously, but that is his style.
• The valuation, on EVA. Asilo uses economic value added. On that metric Future Growth Reliance is 80 percent — twenty percent of the share price explained by what exists now, eighty resting on growth — which Blomster does not find impossible given the size of the car market and a real shift to electric. The current price embeds five years of 25 percent growth plus some margin improvement. If there is no growth at all from here, the correct valuation would be about twenty percent of the current price. He notes 80 percent is historically a low level for Tesla, and that Tesla carries a cyclical component.
• The cheap multiple as a trap. Miettinen cites a Jukka Lepikkö post: a year ago you could have bought Volkswagen at a P/E of 4; the stock has fallen 40 percent and it is still a P/E of 4. The same shape as the dividend illusion — a high current yield is rarely a high future yield.
• The technical picture. They do not do technical analysis; it would need its own person. Momentum has come off, there are many short sellers, and the stock oscillates around the 200 level.
• Twelve stocks. Asilo’s sweet spot is 12, far fewer than a typical portfolio, which leaves considerably more time per position. The process front-loads high hurdles — it is fine if the baby goes with the bathwater — to cut the number of stocks to analyse, and the time saved goes into studying not just the company but its competitors and its supply chain.
• What the supply chain reveals. Ford buys an airbag from Autoliv and the software comes with it; the seats bring their own software; there are more than 150 software suppliers in one car, and the suppliers own that software, so it cannot be changed without permission. It does not end when the car is built, because the software then has to be updated — for the customer free, at a dealership. Even makers with over-the-air capability in principle, like BMW, in practice send you to the dealer in some cases. Tesla’s updates arrive remotely and materially improve the car, and its direct-to-consumer sales model was disruptive in its own right.
• The rule for when a challenger wins. If the existing supply chain still serves the technological change well, the biggest incumbent usually takes the step forward. If it no longer serves it, the newcomer wins.
• A latticework of mental models. The process is a stack of templates for examining a stock and understanding what the company actually does and how large the opportunity really is — and getting to the change before others. In Tesla’s case that meant seeing it is not only cars but energy, and then reading the solar market underneath. The trap is the high-throughput sort: this is a fish, this is a horse, this is a cat; if it is a cat it is not a horse — so you deliberately look for the points that fall outside the process.
• Disruption is not one-to-one. A car is not a faster horse; a smartphone is not a portable phone, it is the App Store concept. Actively looking for the App Store concept finds it earlier than everyone else — today every analyst includes it when they look at Apple.
• When the winner is not known. Sometimes an interesting market is opening and it is not yet clear who takes it. Then they track a top three — for a year, sometimes several — until it tips to one standard or one player visibly pulls ahead. That is the moment.

Watch

The recording lives on the Neuvottelija channel: Tesla Voittajaosake | Henri Blomster | Neuvottelija 219. A Finnish edition of this episode is published at www.neuvottelija.fi.