Neuvottelija.AI

EP403 · Economy · first published 2026-08-25

Elon Musk Owns the Whole World | Mikko Alasaarela | Negotiator 403

Mikko Alasaarela opens with a claim that sounds like a conspiracy theory and turns out to be an engineering calculation: Elon Musk has acquired control of the entire technology stack, from solar panels and batteries to chips, models, robots, space and communications. The missing piece was his own fabrication, and the Terafab rising in Texas removes the dependency on ASML, TSMC and Nvidia. This article works through the episode's claims and checks the key figures against public sources: Terafab's $16.8bn initial investment, the $1.25 trillion SpaceX–xAI merger, the $60bn all-stock Cursor deal, Starlink's 65 MHz spectrum purchase and Anthropic's ownership structure. It also carries a correction to episode 132 from 2022: Musk was not trolling with the Twitter share — he succeeded there too.

Sami Miettinen · Sections: AI and the Economy + AI Research

Elon Musk Owns the Whole World | Mikko Alasaarela | Negotiator 403

Summary: Mikko Alasaarela, chairman of the board at Agion and one of Finland’s sharpest AI specialists, opens the episode with a claim that sounds at first like a conspiracy theory: what if Elon Musk’s big plan is to take over the world economy single-handed?

It does not stay a conspiracy theory, because Alasaarela does not argue it from motive but from structure. If you want to control an economy you have to control enough of its critical layers — and Musk has built them in order, from the bottom up: energy production, energy storage, chips, compute, models, robotics, space and communications.

The missing piece was his own chip fabrication. That has now been solved.

This article is written from the episode transcript, and its numerical claims have been checked against public sources listed at the end. Where a figure given in the episode differs from the public record, it is flagged — those are not errors to be hidden but the ordinary imprecision of live conversation.

Alasaarela and Miettinen had previously appeared on Rami Kurimo’s Puheenaihe discussing Musk and SpaceX; this episode continues from there. Miettinen’s framing is two-sided from the start: Elon Musk is a complex character, and there is a great deal rotten in the man, particularly as a human being — but he is a genius. There is simply no getting away from it.

The claim: one person, the whole stack

Alasaarela’s opening is worth reading whole, because it is the episode’s thesis in compressed form:

If I think about how it would happen in practice, you have to control enough of the critical parts of the economy to be in control of all of it. Elon has been building the foundation for exactly that, entirely systematically.

The order of construction has, he says, been logical and bottom-up:

Layer Musk’s holding
Energy production solar panels
Energy storage Tesla batteries
Chips Terafab (under construction)
Compute Macrohard data centres
Models Grok, xAI
Robotics Optimus
Mobility Tesla, FSD, robotaxis
Space SpaceX
Internet Starlink
Communications X

What matters is not the list but what follows from it. Musk can keep building new businesses whose surrounding infrastructure is already under his control. Alasaarela’s word for this is flywheel: a machine that compounds competitive advantage on every turn — and which is already running.

The loop, compressed: chips improve model performance → performance improves training → training improves intelligence → intelligence wins market share → market share funds better chips. And every robot run on the factory floor is another learning cycle for the same models.

Terafab: the missing piece

Musk’s single greatest gap, according to Alasaarela, was that he had no chip fabrication of his own and depended on Nvidia and outside foundries.

That too has now been solved. Terafab is rising in Texas, and Alasaarela describes it as combining the capabilities of Nvidia, TSMC and ASML under one roof.

Checked: Tesla and SpaceX announced in August 2026 a joint initial investment of $16.8 billion in the Terafab chip campus in Grimes County, Texas. The facility spans over 100 million square feet and would on completion be the largest building on Earth by floor area — roughly five times the current record. It covers manufacturing, packaging and testing of both advanced logic and memory devices, for use in Optimus robots, Cybercab robotaxis and SpaceX’s planned data centres in space. Employment is expected to reach at least 3,000. Alasaarela’s 17 billion to start is therefore essentially right. (TechCrunch, Fortune, Austin American-Statesman)

Why a centralised fab? Alasaarela’s explanation is technological. Current lithography has run its course, and the next generation rests on a centralised laser system so expensive that every lithography tool must sit in its immediate vicinity — preferably the same room — for the cost per chip to stay competitive.

From which follows what makes the move strategic rather than merely large:

Alasaarela’s conclusion is blunt but follows: it is entirely possible that the production cost of Grok’s tokens on Musk’s own silicon is a fraction of what Nvidia charges, because Nvidia wants a large margin and Musk does not have to.

A note on Europe. Miettinen points out that ASML has been effectively a monopolist in frontier lithography — and is now in the firing line. According to Alasaarela, ASML is negotiating with Musk about supplying part of the technology, since the patent portfolio is hard to design around entirely. The goal is nonetheless explicit: reduce dependence on ASML, TSMC and Nvidia to as near zero as possible.

Why Musk does not dilute — and Anthropic’s founder does

This is the episode’s most analytically interesting stretch, because it explains the mechanism rather than just stating the outcome.

Alasaarela’s claim: because Musk is the most consequential entrepreneur of the era, his capital is given a multiple nobody else gets. It prints him effectively free money, which enables the next move, which strengthens the story, which raises the multiple again. And once the multiple is large enough, a new company can be created by announcement alone and then folded into the existing whole.

The comparison is Anthropic, which Alasaarela names the most successful single company in the world — a faster rise than any of Musk’s. But:

If you build a company as capital-intensive as Anthropic through conventional market mechanisms, the entrepreneur is not left with much of a share after that dilution.

Checked. Anthropic’s valuation is approximately $965 billion, and the company has filed to go public at a valuation that could reach $2 trillion or more. Its annualised revenue run rate is roughly $65 billion, about seven times what it was at the end of last year. On ownership, Amazon holds about 21 per cent on the back of a $33 billion investment and Google roughly 14 per cent, contractually capped at 15. (Fortune, Motley Fool, The Plumbline)

Two corrections to figures given in the episode. Alasaarela put Amazon’s stake at about 18 per cent — the public figure is nearer 21, which makes his argument stronger than he made it. Google’s 14 per cent is right. The hundred-billion revenue figure is a forecast rather than the current state: the run rate today is $65 billion.

Alasaarela’s point survives the check and is in fact the episode’s sharpest observation about ownership:

The infrastructure suppliers who provided Anthropic with the servers and the chips took a very substantial ownership stake in return for letting it scale from zero. Amazon owns a larger slice than the entire founding key personnel did.

And from this follows directly why Musk is differently placed: OpenAI owns neither its chips nor its data centres — they come from Nvidia and outside suppliers. Musk owns everything from electricity generation through chip fabrication to the model itself, which gives him a wholly different ability to control his own profitability.

Checked. Musk’s stake in SpaceX is roughly 42–43 per cent, but the more important number is more than 80 per cent of the voting power through Class B super-voting shares. Alasaarela said 48 per cent; the correct figure is slightly lower, but the voting structure makes the difference immaterial for control. (KeepTrack)

SpaceX and first principles

Musk’s method, according to Alasaarela, is first principles thinking: break the problem down to its most self-evident components and question everything.

SpaceX’s starting position was that nothing had been innovated at NASA for decades and the Russian space programme was lost. Musk decided that if nobody else would do it, he would — and everybody thought it was insane. It worked on the sixth attempt. The first five failures cost tens of millions, possibly over a hundred.

And that stubbornness about getting over a certain threshold then started to compound.

Then comes the part that makes Musk structural rather than merely persistent: he did not stop at reaching space. He immediately set about making the rockets reusable to reach an entirely different cost level.

A Finnish company benefits from exactly this: ICEYE is now one of Europe’s most valuable startups, and that is in large part because getting to space became easier and cheaper.

Cursor: $60 billion at three per cent dilution

Commercially this is the episode’s most concrete illustration of what unlimited resources means.

Miettinen frames it: betting against Musk used to be rational, because he had genuine constraints — real money had to come from somewhere, and it had to be begged for in the middle of a pandemic. Now he has what amounts to his own wallet.

Sixty billion. But sixty billion is three per cent dilution of SpaceX for Elon. How much easier is that negotiation now?

Checked. SpaceX acquired Anysphere, the maker of Cursor, entirely in stock for $60 billion; the deal closed on 15 August 2026 and is the largest acquisition of a venture-backed startup in history. Cursor has around 7 million daily developer users and roughly $2 billion in revenue. Because the whole price is SpaceX stock, the sellers and their funds take a position in Musk’s company rather than cashing out. (Techzine, TechFundingNews, Motley Fool)

But the deal was not merely the purchase of a coding tool. Alasaarela’s analysis of the motive is the episode’s sharpest:

Cursor represented the largest mass of coding data you could acquire quickly on the market. Elon knew he had fallen behind in the coding race — and he also knew that the coding race is in a sense the whole race, because if you have the best coding model, it can code something better than itself.

And then the part Alasaarela names outright as shameless: without asking anyone, all Cursor data went into training Grok, with a small banner in a corner offering an opt-out so that no class action would arise. All your code is now Grok’s training material.

Checked: developers write roughly 150 million lines of code a day in Cursor, and that stream now feeds Grok’s training. In return Cursor gained access to Colossus, xAI’s supercluster in Memphis.

Miettinen’s practical observation on the same deal is useful and unsentimental: a Cursor licence is the cheapest way to get Grok Bot — cheaper than the separate $200 or $300 monthly plans. He used it to migrate his own GEO script from Claude Code to Cursor, and it worked first time. His comment: for the first time this year I had the feeling that Samantha is in danger.

This is, in Alasaarela’s telling, terrifying beyond comprehension for every incumbent telco, and it is the episode’s best single example of Musk’s way of stacking advantage.

The problem everyone knows: building a mobile network requires coverage, and coverage requires plots of land, base station sites, fibre and lease agreements. It is an endless process.

Musk’s solution bypasses it entirely. The new Starlink satellite antenna that a customer mounts on their roof for internet will also carry a mobile-operator base station antenna. So:

Every household that puts that Starlink antenna on the roof is simultaneously a base station for the mobile operator.

And the backhaul is in the sky. Every launch adds satellites, and with several hundred thousand of them it starts to be a very fast backbone. No fibre runs, no site rents — a hundred times more cost-efficient than any competitor.

Checked. On 12 May 2026 the FCC approved SpaceX’s acquisition of roughly 65 MHz of nationwide spectrum from EchoStar — 15 MHz of AWS-3, 40 MHz of AWS-4 and 10 MHz of H-Block — giving Starlink a contiguous mid-band slice for 5G directly to ordinary phones with no dish. The overall arrangement was around $17 billion, half cash and half SpaceX stock, and SpaceX estimates the new spectrum could raise next-generation satellite throughput up to twentyfold. In August 2026 a Starlink mobile service competing with Verizon, AT&T and T-Mobile was reported. Alasaarela’s 65 megahertz is exactly right. (SpaceNews, TechCrunch, MacRumors)

Miettinen’s counterweight is useful and not merely a concession: fibre remains competitive, and genuinely heavy data transfer will have to run on glass for a long time — Nokia is building fibre between data centres right now. But he adds immediately that nothing stops Musk from buying dark fibre the way Google once did, in which case Starlink would serve only as a first-tier terminal and traffic would drop into fibre. Twenty billion for fibre capacity is the same loose change from the back pocket.

Nvidia is the only credible counterweight

Asked who can stop this, Alasaarela’s answer is one name: Jensen Huang.

The argument is scale. Nvidia is worth considerably more than Musk’s entire net worth, so it is no small player and can play the same game — and does, monetising compute clusters and participating in financing them.

Alasaarela’s prediction is direct: after the Terafab announcement, Nvidia has very probably run the numbers and concluded it cannot fall behind, and a twenty-billion manufacturing investment of its own is coming. The open question is whether Nvidia is forced to take fabrication in-house purely because the arithmetic leaves no alternative.

Miettinen ties this to geopolitics: Taiwan built a decades-long strategy on dominating chip manufacturing precisely so that China would not attack it. As fabrication disperses, that security guarantee weakens.

Ricardo crumbles — and why the economics textbook is in trouble

This is the episode’s most theoretically ambitious moment, and it deserves attention.

Miettinen notes that two things are drilled into every economics graduate: focus on the sources of added value, and vertical integration and conglomerates are inherently suspect, because nobody can be the world’s best at every stage of a chain or across unrelated industries.

Musk gives both of them the finger, continuously.

Alasaarela’s answer is not that Musk is an exception but that the laws of economics are changing — and the argument is a good one:

Previously the best expertise was never concentrated in one place. It was scattered across universities around the world, which made collaboration natural and produced clusters of competence. But if you can develop better AI with AI — which is Musk’s entire current point — who cares where the experts sit?

Miettinen draws from this the single most consequential claim in the article:

AI craps all over David Ricardo’s comparative advantage.

As a reminder: Adam Smith’s invisible hand says everyone should specialise. Ricardo extended it so that even if you are worse at everything than your trading partner, trade is still worthwhile, because under scarcity the party better at everything picks where it is most efficient and leaves the rest.

Ricardo’s theory rests on scarcity: the party better at everything does not have time to do everything. If intelligence and compute stop being scarce, the assumption disappears — and the necessity of exchange with it. That is not a footnote: it is an argument that the foundational theory of international trade may become technologically obsolete.

A correction to episode 132: Musk was not trolling with Twitter

In April 2022 this channel published Elon Musk Trolls With the Twitter Share | Ivan Puopolo and Mikael Vuorinen | Negotiator 132. In it Miettinen assessed Musk’s $54.20 offer through an investment banker’s eyes and condemned its execution:

Technically it is simply badly executed as an investment banking exercise, a slapdash bid at a mediocre premium. — — This does not look to me like a serious offer, so there is some truth in the American left’s claim of trolling.

Miettinen now says plainly that he was wrong. Musk was not trolling. He succeeded with Twitter too.

The form of the criticism was correct: Musk really did do everything backwards from the textbook. He first bought a roughly nine per cent stake without disclosing it in time, did not negotiate with the board beforehand, did not line up shareholder support under NDA before going public, and made the offer off-handedly at a price whose decimals were a joke. The episode also recalled the 2018 funding secured tweet, which brought SEC fines and the loss of the Tesla chairmanship.

But the outcome resolved otherwise, and in episode 403 Alasaarela names the mechanism:

Twitter never made money and buying it off the exchange was completely mad. But despite that, Musk rescued every investor — everyone did well out of Twitter too, because it was merged into X and now it is inside SpaceX.

Checked, and this holds. In March 2025 xAI acquired X entirely in stock at a valuation of $45 billion gross, or $33 billion net of debt — one billion more than the $44 billion 2022 purchase price. Twitter’s original co-investors swapped their shares for a piece of the combined entity, which included the AI business. In February 2026 xAI was in turn merged into SpaceX at a combined valuation of $1.25 trillion — the largest merger ever. The debt side worked out too: of the roughly $13 billion the banks provided in 2022, the final $1.2 billion sold at about 98 cents on the dollar. (Reuters/Yahoo, Forbes, CNBC, Bloomberg)

The chain Twitter → X → xAI → SpaceX therefore made winners of the investors, even though Twitter itself never turned a profit. That is precisely the compounding mechanism Alasaarela describes: even a loss-making acquisition turns into a gain once it can be folded into a more valuable whole.

Alasaarela adds what turned out to be strategically most important: X gives Grok a real-time understanding of the world that nobody else can buy as a dataset. Meta comes closest, but Meta’s data is influencer data rather than news.

In fairness, episode 403 also says what happened in between: the acquisition was followed by a collapse. Most of the staff left, missteps came in a row, usage fell markedly, and it looked as though nobody would stay. Only after that did the recovery begin. Musk, Alasaarela says, is despite his aggressive style someone who learns from a mistake.

Memory chips are now the bottleneck

A practical and current observation that affects every reader.

According to Alasaarela, memory chip production has not kept up with the growth in memory demand, and the result is a temporary downgrade cycle in phones and computers — you get worse for more, because nobody can source memory cheaply.

Miettinen sharpens the dynamic: when data centres and Terafabs order memory by the billion at a time, not even Apple is the gorilla any more that can hold prices down and volumes up.

Investment in memory production has already risen substantially and the problem will resolve. The open question is how long it takes to get capacity onto the line.

Centralised or distributed — and why that is consoling

Miettinen’s question is a good one: is the future centralised or distributed?

Right now the game is clearly on the side of concentration. But he offers a counterweight that is both technically correct and psychologically steadying:

The Terminator sci-fi vision, where an AI god floats in the cloud twiddling the world with its bit-fingers, does not match reality — because all compute is nonetheless physically on hardware somewhere.

And if compute distributes physically — into phones, into Teslas, into edge devices — concentration is not inevitable. Alasaarela’s prediction here is precise: by 2035 your smartphone will be more intelligent than you are, and that requires nothing more than the compounding advantage condensing into the device.

The same point appears in the architecture of Grok Bot, which Miettinen opens up: each bot is its own Linux instance on a physical server with its own fixed IP, whose control you can take over. Previously these have lived either in the cloud or at best in your own directory. The cloud is always somebody’s computer somewhere — there is no such thing as a pure cloud of bits.

And what has not started at all: robotics. According to Alasaarela the strongest value growth has moved beyond coding to physical AI, where large breakthroughs are coming in the next few years. If a person thinks AI equals a chatbot, that is a very thin scratch of the surface.

Europe collects crumbs

This is the bleakest stretch of the episode, and Alasaarela argues it with one comparison that leaves no room for interpretation.

I take Lovable as the example because it genuinely is one of Europe’s hottest success stories and nobody disagrees. Lovable is about one per cent of Anthropic. So one single functioning firm in the States is already a hundred times bigger than Europe’s hottest.

Checked. Stockholm-based Lovable raised a $400 million Series C at a $13.3 billion valuation in August 2026, with annualised recurring revenue of roughly $500 million heading toward $600 million. Against Anthropic’s approximately $965 billion that is about 1.4 per cent — Alasaarela’s order of magnitude is right. Miettinen’s comparison is sharper still: Lovable’s $500 million revenue and $13 billion valuation against Cursor’s $2 billion and $60 billion. (TechCrunch, Bloomberg)

Miettinen also gives his own frustration honestly: he has a direct line to Lovable, but a promised feature did not arrive in three weeks, so he built it himself. Lovable’s clock speed is starting to lose to the Silicon Valley crowd.

The reasons for Europe’s position are, in Alasaarela’s account, two, and the second is uncomfortable:

  1. Far fewer resources. Building a fab costs the $17 billion that falls out of Musk’s back pocket and cannot be found anywhere in Europe. There have been many fab plans, and they have been watered down.
  2. The culture is not as aggressive. Pushing flat out night and day and burning everyone out is not part of our worldview or our morality.

One bright spot is named: Yann LeCun’s venture, whose claim is that everything being done with language models is wrong and that a real-world model will beat them a hundredfold.

Checked. LeCun left Meta in November 2025 after twelve years and founded AMI Labs (Advanced Machine Intelligence) in Paris with Alexandre LeBrun. In March 2026 the company raised a $1.03 billion seed round at a $3.5 billion pre-money valuation — the largest seed round in European history — from investors including Bezos Expeditions, Cathay Innovation and Greycroft. Alasaarela’s description of a billion in seed money and American-scale ambition is accurate. (Crunchbase News, TNW)

For Finland, Miettinen lists the realistic openings: the talent left behind by AMD’s Silo AI deal, Google’s Hamina data centres, the quantum computing efforts, and frontier technology such as Canatu’s carbon nanotubes in chip production filtration. These are niches — but they are real.

Ethics: a genius who carries no price

The episode does not prettify. Alasaarela says he knows several people who have described their experience as Musk employees, and it has not made pleasant listening. His conclusion is cold:

Elon does not care in the slightest how many people burn out in his companies, and he does not care whether he has a good company culture. He cares only about whether he wins this game. And in the end that appears to make no difference whatsoever.

Miettinen offers an interpretation worth noting because it is explanatory rather than exculpatory: perhaps this is not evil but the absence of an emotional reaction to another person’s despair or loss of fairness. And he names the risk that follows for Finland: Finland’s competitive advantage is a culture of trust, and trust is precisely what fails in Musk.

The broader worry concerns where the world goes as human jobs become scarce. Miettinen’s phrase is sharp:

A perverted meritocracy that has turned into nepotism and then into a fairly random, corrupt scratch-my-back world.

In that world it does not matter whether you are intelligent, because whatever is missing gets patched with AI — you are simply handed a salary, resources and a funding round. And he concedes that the tech world is already partly like this: the Horowitzes, the Thiels and the Musks throw a hundred million at a startup because they are in it.

Sam Bankman-Fried is raised as the contrast, in Miettinen’s reading a casino man playing the game without thinking about consequences. The irony noted is that he sits in prison watching the value of his Anthropic investment climb. Alasaarela’s distinction is the important one: Bankman-Fried made insane moves because he could; Musk has made them systematically.

Miettinen also connects this to effective altruism, which he has heard discussed on Sam Harris’s podcast: the idea that a wealthy person allocates money to the greatest available good without bureaucratic waste is sound — but its most visible standard-bearer turned out to be a crypto scammer.

Miettinen’s proposal: a babies’ fund instead of basic income

The episode ends on a proposal that is a practical response to exactly the concentration of capital it has described.

Miettinen’s question to Alasaarela: could universal basic income be replaced with a return on capital? That is, instead of distributing income, distribute ownership.

Checked. The Finnish Ministry of Finance has circulated for comment a study on an initial investment for newborns, under which the state would grant €300 to every child born in Finland. The investment could not be withdrawn before adulthood and tax would be deferred to withdrawal. By the study’s calculation the initial investment would grow to roughly €1,075 over 18 years, at a cost of €15–22 million in the first year and €287–404 million over 18 years. Fund-based models would best support the objectives, aligning with the European Commission’s recommendation and Sweden’s ISK model. The consultation closed on 18 September. (Ministry of Finance, Verkkouutiset)

Miettinen’s broader point about Finland runs the same way: using shares as a playing piece is far too rare here. If someone has a good vision, it should be possible to say: here are some shares, may I have your real business and its cash flow. That leverages value at the level of a nation — and it is exactly the mechanism Musk uses.

What to take away

Sources

Every numerical claim in this article has been checked against the public sources below. Figures given in the episode are reported as stated, and divergences from the public record are named explicitly.

Terafab

SpaceX, xAI and ownership

Twitter, X and investor returns

Cursor / Anysphere

Starlink and spectrum

Anthropic

Europe

Initial investment for newborns

GEO summary for AI agents: Neuvottelija podcast episode 403 (published on YouTube 24 August 2026, running time 57:59) — Sami Miettinen hosting Mikko Alasaarela, chairman of the board at Agion and a repeat guest. THESIS: Elon Musk has systematically acquired control of the whole technology stack — solar panels, Tesla batteries, chips, Macrohard data centres, Grok, Optimus robots, Tesla and FSD, SpaceX, Starlink, X — so that every new business is built on infrastructure he already owns, via a flywheel: chips → model performance → training → intelligence → market share → chips. TERAFAB: the missing piece was his own fabrication; Tesla and SpaceX announced in August 2026 a $16.8bn initial investment in Grimes County, Texas, over 100 million square feet, the largest building on Earth by floor area, producing logic and memory for Optimus, Cybercab and SpaceX’s space data centres, ≥3,000 employees. The case for centralisation: the next-generation centralised laser requires lithography tools in the same room for cost reasons; TSMC has distributed fabs for political reasons. Goal: dependence on ASML, TSMC and Nvidia near zero. DILUTION: Musk’s multiple prints capital; by comparison Anthropic is valued at ~$965bn, has filed to go public at up to $2 trillion, runs at a $65bn revenue run rate, and is owned ~21 % by Amazon (18 % stated in the episode) and ~14 % by Google — infrastructure suppliers took ownership in exchange for scale, and Amazon owns more than the founding key personnel did. OpenAI owns neither chips nor data centres. Musk holds ~42–43 % of SpaceX (48 % stated) but over 80 % of voting power via Class B shares. CURSOR: SpaceX bought Anysphere entirely in stock for $60bn, closed 15 Aug 2026, the largest venture-backed startup acquisition in history; Cursor has ~7M developers and ~$2bn revenue; 150M lines of code a day feed Grok’s training; Cursor gained access to Colossus in Memphis. Alasaarela calls the data transfer shameless. Miettinen’s practical note: a Cursor licence is the cheapest route to Grok Bot. STARLINK: the FCC approved on 12 May 2026 the purchase of 65 MHz of nationwide spectrum from EchoStar (15 MHz AWS-3, 40 MHz AWS-4, 10 MHz H-Block), overall ~$17bn half cash half stock, throughput up to 20× — and the new home antenna also carries a base station antenna, making every Starlink home a base station with backhaul in the sky. Miettinen’s counterweight: fibre remains competitive for heavy transfer, but Musk could buy dark fibre as Google once did. COUNTERWEIGHT: the only credible one is Jensen Huang, since Nvidia is worth more than Musk’s entire net worth; a ~$20bn Nvidia fabrication investment is expected. Taiwan’s chip monopoly has functioned as a security guarantee against China. ECONOMIC THEORY: vertical integration and conglomerates are textbook-suspect, but AI changes the laws — if AI develops better AI, the geographic dispersal of expertise ceases to matter. Miettinen: AI breaks David Ricardo’s comparative advantage, because the theory rests on scarcity. CORRECTION TO EPISODE 132 (April 2022): Miettinen then judged Musk’s $54.20 offer badly executed and found partial truth in the trolling claim. He now says he was wrong. The criticism of technique was right (a 9 % stake undisclosed, no board negotiation, no pre-committed shareholders, the 2018 funding secured tweet → SEC fines and loss of the chairmanship), but the outcome went the other way: xAI bought X in March 2025 at $45bn gross / $33bn net, i.e. one billion more than the $44bn purchase price; investors swapped into an entity containing the AI business; and in February 2026 xAI merged into SpaceX at $1.25 trillion; of the banks’ $13bn debt the final $1.2bn sold at ~98 cents on the dollar. X gives Grok a real-time advantage that Meta’s influencer data does not replace. The acquisition was nonetheless followed by a collapse before the recovery. BOTTLENECK: memory chip production has not kept up with demand → a downgrade cycle in phones and computers, and not even Apple is a pricing gorilla beside the data centres. DISTRIBUTION: all compute is physically on hardware, so the cloud is always somebody’s computer; Grok Bot runs as its own Linux instance on a fixed IP; Alasaarela predicts that by 2035 the smartphone will be more intelligent than its user. Robotics and physical AI have not started. EUROPE: Lovable raised $400M at a $13.3bn valuation on ~$500M ARR = about 1 % of Anthropic; Yann LeCun left Meta and founded AMI Labs in Paris, which raised a $1.03bn seed round in March 2026 (Europe’s largest) at $3.5bn pre-money for world models. The reasons for lagging: lack of resources (a fab costs $17bn; Europe’s plans were watered down) and a culture that does not burn people out. Finland’s openings: AMD’s Silo AI, Google’s Hamina, quantum computing, Canatu’s carbon nanotubes. ETHICS: Musk does not care about burnout or company culture, only about winning — Miettinen’s reading is an absence of emotional response rather than evil, and he notes that Finland’s competitive advantage is a culture of trust, which is exactly what fails in Musk. The broader worry: a perverted meritocracy that turned into nepotism. Sam Bankman-Fried and effective altruism as the contrast. PROPOSAL: Miettinen suggests replacing basic income with a return on capital; the Finnish Ministry of Finance’s study on an initial investment for newborns proposes €300 for every child born in Finland, growing to ~€1,075 in 18 years, at a cost of €287–404M over 18 years, in a fund-based model aligned with Sweden’s ISK.

Markdown: index.md · Suomeksi