---
title: "Megatrends for Equity Returns | Pasi Havia | Negotiator 416"
summary: "Pasi Havia, portfolio manager of United Bankers' UB Megatrendit fund, tells Sami Miettinen how he builds a portfolio around four megatrends: AI and digitalisation (more than 60 per cent), electrification and the energy transition, population ageing, and European defence. The fund returned more than 26 per cent after fees in its first seven months. Havia's core idea is that demand for AI compute and memory no longer comes from people but from machines, which are insatiable, and that traditional valuation multiples break down under exponential growth. The episode covers the Micron paradox, laser-precise exposure to the bottlenecks of the value chain (Broadcom, Lumentum, ASML), why SpaceX and Anthropic stay out of the fund, circular financing, companies competing with governments for capital, European defence (TKMS, Theon), America's ageing power grid layer by layer (Constellation, Quanta, GE Vernova, Vertiv), Eli Lilly over Novo Nordisk, and regulatory risk in China. Figures have been checked and divergences flagged. Published 19 October 2026."
datePublished: 2026-10-19
dateModified: 2026-10-19
originalLang: en
section: economy
sections: ["economy"]
authors: ["Sami Miettinen"]
tags: ["Neuvottelija","EP416","Pasi Havia","Sami Miettinen","United Bankers","UB Megatrendit","Megatrends","Investing","Equities","Artificial intelligence","Semiconductors","Micron","Broadcom","Lumentum","Nvidia","Palantir","Energy transition","Defence industry","Eli Lilly","Valuation multiples"]
canonical: https://www.neuvottelija.com/ai/ep416-megatrendeilla-osaketuottoa-pasi-havia/
---
# Megatrends for Equity Returns | Pasi Havia | Negotiator 416

# Megatrends for Equity Returns | Pasi Havia | Negotiator 416

> **Summary:**
> In episode 416 of the Neuvottelija channel, Sami Miettinen interviews **Pasi Havia**, portfolio manager of United Bankers' **UB Megatrendit** equity fund. Havia builds the portfolio on four megatrends: AI and digitalisation (more than 60 per cent of the fund), electrification and the energy transition, population ageing, and European defence. At the heart of the episode is a question every investor in the AI boom has to face: how do you value companies whose demand no longer comes from people but from machines, and what happens to traditional multiples when growth is exponential? Havia's answer is practical. He takes "laser-precise exposure" to the bottlenecks of AI infrastructure, invests only in profitable companies and expects to move on once the cycle shows signs of cooling. Sami, for his part, is preparing for his own AI investing challenge and tests his thesis that machines will replace human labour. Published 19 October 2026.
>
> **A note on reading this.** The episode is a conversation between two investors, not investment advice. Sami says so himself: *"no investment recommendations are given here"*, he is not an investment adviser and receives no compensation from United Bankers. Havia explains his own picks and reasoning; Sami gives his own views and forecasts. The figures are the speakers' own, and some are from memory. They have been checked against public sources where possible, and divergences are flagged in the text rather than silently corrected. The subtitles carry no speaker labels, so speakers were inferred from the flow of the conversation; quotations are translated from Finnish for sense. **Disclosure:** Sami says in the episode that he owns Nokia and Wärtsilä and invests mainly through index funds and private equity funds.

---

## Who is talking

**[Pasi Havia](https://www.neuvottelija.com/podcast/guests/pasi-havia/)** is portfolio manager of United Bankers' UB Megatrendit fund. The fund started on 6 March 2026; Havia joined United Bankers in mid-January. He has about 12 years of professional portfolio management behind him: the first fund he began managing was set up in 2014, and he already used megatrends in his previous fund. At the end of the episode it emerges that he lives in Pärnu, Estonia.

**Sami Miettinen** hosts the Neuvottelija channel. He is blunt about himself as an investor: index funds and private equity, but *"as a stock picker I'm a bit crap"*. That makes the episode a training session for him too. He has signed up for the next calendar year to a group that invests real money with the help of AI, and specifically in individual stocks rather than indices.

## 1. A megatrend portfolio is a weighted view, not a permanent allocation

Havia's fund has four megatrends, and their weights show where he sees returns right now:

| Megatrend | Weight | Havia's reasoning |
|---|---|---|
| AI and digitalisation | over 60% | The biggest return potential right now, and it has driven the fund's returns |
| Electrification and energy transition | second largest | Overlaps with AI, because part of the energy demand comes from data centres, but there are other drivers too |
| Population ageing | about 10% | A more predictable and "flatter" trajectory |
| Defence and security | about 10% | The chosen line is specifically European defence |

The AI overweight is deliberate: *"I'm overweighting it right now and taking a view there."* At the same time Havia stresses that none of the weights is permanent. Megatrends can be swapped, a fifth can be added or the list cut to three. If something happens with AI, he can shrink its weight and grow, say, ageing, *"which doesn't correlate so strongly with tech companies"*.

This is an essential feature of megatrend investing that is often missed. A theme is not an index. It is the portfolio manager's view of where the world is going, and the weights measure the strength of that view. In Havia's portfolio, three of the four themes currently play supporting roles.

**Returns.** According to Havia, the first seven months returned *"over 26 per cent after all costs"*. It was not an easy start. When the fund launched in March, *"Trump started rattling around Hormuz"*, and the portfolio first went down before climbing back. According to the YouTube description, the gains were driven above all by AI-infrastructure names Micron, Broadcom, Lumentum and AMD.

## 2. Miettinen's "nasty megatrend number one": human labour is replaced

During the episode Sami lays out his provisional thesis for next year's investing challenge:

> *"Human labour will be disrupted by AI, first through our friends the frontier models and their deployment campaigns and other partnerships, so that people become androidised on the intellectual side, and then a bit later through robotics and real-world AI."*

Havia is on the same lines but refines the order. The first phase is already visible; physical AI, the replacement of human physical work, is still coming. In his view it ties in with deglobalisation. When a humanoid or a robot does the work at practically the same price at home as in a low-cost country, the cheap-labour advantage disappears and production can move closer. Havia thinks this trend started during Covid, when supply chains jammed, shipping stalled and chips ran short, and he does not see it going away.

Sami pushes the idea further. The old "plain vanilla" megatrends were urbanisation, globalisation and computerisation. Now the Strait of Hormuz is closed, shipping is in trouble and the United States no longer wants to sustain an interdependent world economy, so the globalisation trend could turn into localisation. Havia's answer is that this has already been happening for years: *"I don't see why that would change now."*

The oddest moment concerns Sami's thesis itself. When Havia explains the demand for memory chips, he uses it as his argument: *"I think we come here to Miettinen's nasty megatrend number one, that human work gets replaced, so the machines need such a staggering amount of silicon."* Sami's social pessimism is, for Havia, an investment thesis: if machines do people's jobs, they also do the consuming instead of people.

## 3. Demand no longer comes from people but from machines

The episode's most important analytical idea is Havia's description of who actually consumes AI compute:

> *"Before, the usage was for people, but now it's for the machines, and they are insatiable. Nothing ever seems to be enough."*

He distinguishes three waves. First, tokens were consumed by a person typing a question to a chatbot. Today *"the AI agents use by far the largest share of the token volume"*. The third wave comes when physical AI really gets going. Then part of the compute moves out of data centres into the devices themselves, because perceiving the real world cannot tolerate delay:

> *"You have to process locally when you need to perceive the real world. The latency is too long if you throw it over the internet to a data centre."*

For an investor that has a concrete consequence. Edge computing needs memory and processor chips outside data centres, so demand is not confined to hyperscalers' server halls. Sami puts the same point through Nvidia: Jensen Huang has grown his company from one trillion to the next, but *"the demand doesn't come from people. It comes from those gigantic server farms."*

Havia is also direct about what this means for his own timing. At this stage of the cycle he prefers the companies needed to build today's infrastructure, *"because the demand is so ferocious"*. When the cycle matures and signs of cooling appear, he will probably move on to different kinds of companies. A megatrend does not mean sitting in the same stocks within a theme forever.

## 4. The memory boom and the Micron paradox

Memory chips are the longest single topic of the episode and Havia's *"pet stock"*.

Sami starts with his own miss. In January he bought Apple's most expensive laptop because, in his view, the RAM boom was already obvious and had reached the shop shelves. Recently he ordered the most expensive Mac possible, *"eleven grand plus VAT"*. Yet he did not buy Samsung or SK Hynix shares, because in his head South Korea belonged *"a bit in the China category"*: the chaebols can play their own game.

Havia has no Korean memory cluster. His memory exposure is America's **Micron**, and the choice of country is deliberate. If MAGA politics starts protecting its own market, the political risk is smaller in an American company.

In Micron, Havia sees a market misunderstanding:

> *"Micron has risen about 200 per cent in the last six months and about 500 per cent over the year. The multiples are still exactly the same as six months ago and a year ago."*

The share price has risen with earnings, but the valuation multiple has not expanded. The market prices Micron as if it were permanently at the peak of its cycle: *"right now we're at the peak of the cycle, that's why the multiples are so low. Six months ago we were at the peak of the cycle, so the multiple had to be this low."* Havia does not understand the logic, because the company says the opposite. With its results on the last day of September, Micron said 2027 will be tighter for memory than 2026. When an analyst asked when supply and demand would balance, the company said it had no visibility on that.

His second argument is the contract structure. Micron is signing long agreements with customers **out to 2031** and locking in prices now. According to Havia, such agreements cover about **35 per cent** of the capacity the company can offer through 2030, with a target of **50 per cent**. *"In practice that leaves half the business that could somehow be exposed."* So he struggles to see why the stock is priced as if new capacity were about to crush prices just around the corner.

*Clarification.* Public reports say Micron has 26 strategic customer agreements, two extended to 2031, and expects long-term contracts to exceed 35 per cent of **revenue** by 2030. The figure is a share of revenue rather than capacity, but the order of magnitude matches Havia's. In the results release, management said 2027 and 2028 would be much tighter for memory than 2026, which supports his picture.

Against the history of memory cycles, this is the episode's most interesting claim. Sami recalls that in earlier cycles *"there was always some guy who built overcapacity"*. Havia's counter is that the nature of demand has changed. When the consumer is a machine rather than a person, demand does not saturate in the same way, and hyperscalers' investment forecasts have been raised time after time.

Micron's focus, Havia says, is DRAM, and within that **HBM**, the high-bandwidth memory used in AI accelerators. The company even exited consumer memory sales because it wants to serve data centres at *"fatter margins"*; it still makes NAND storage. Its latest quarterly report had, for the first time, a slide on physical AI, which the company believes could become a significant new business area by the end of the decade.

Sami adds the consumer's view. Data centres buy memory wholesale and at a premium, and not even a giant the size of Apple can compete with them. In his view that is why iPhone prices have had to rise. His advice is half in jest: *"I'd stock up a bit of my own RAM and memory at home while prices are rising."*

## 5. The value chain: ASML, Broadcom, Lumentum and other bottlenecks

Havia's way of choosing companies within the AI theme is the most usable part of the episode. He does not buy the theme; he buys its narrowest points.

**Europe.** From Europe the fund holds **ASML** and **BE Semiconductor**. Havia's problem is that it is very hard to find European companies at the centre of AI: *"Europe has played itself offside."* Sami points to ASML's dominant position and recalls his conversation with [Mikko Alasaarela](https://www.neuvottelija.com/ai/ep403-elon-musk-omistaa-koko-maailman-mikko-alasaarela/) about Elon Musk's Terafab project: if someone develops an entirely new way to make chips, that too would be a disruption. Havia expects even Musk to use other companies at first, until he has built his own teams.

**Custom chips.** Many large companies now design their own chips. Amazon has Trainium and Google has TPUs, but neither makes them entirely in-house. The design work is done by **Broadcom**, which is in Havia's portfolio.

**Optics.** Havia's latest pick is optics company **Lumentum**, and the reasoning shows his method:

> *"With these company exposures I like to take really laser-precise exposure."*

Nokia is involved in the fibre links between data centres, but its business contains *"so much of all kinds of other stuff"*. **Coherent** would have been a better comparison, but its lasers go to many other industries. Havia wanted exposure specifically to the fact that data centres are being built aggressively and capex is surging, so he chose Lumentum rather than Nokia or Coherent. Pure exposure matters more to him than a familiar name.

## 6. Why the fund holds neither SpaceX nor Anthropic

Sami asks about the giant listings to come. Many employees of SpaceX, Anthropic or OpenAI will make a lot of money when these companies eventually go public. What screen would they have to pass to get into Havia's fund?

The answer is a simple rule. Every company in the portfolio is profitable, and *"Anthropic, for example, is deeply loss-making"*. Havia wants a company to show that it can run a productive business. He admits the cost of the rule: if he made the right early pick, returns would be much better. But his experience is that even coming in later, *"there's still runway left"* if the company sits well in the supply chain, and the risk of picking wrong is smaller.

**Musk's conglomerate.** Sami recalls being taught at business school to focus on core operations and to sit at a narrow point in the value chain. Musk does the opposite: he integrates everything he can vertically and then places random bets on top. Havia's reading is that Musk has packaged his companies in layers: a moonshot business, one that already makes money, and a third in between. That makes valuing the whole hard, and he has not invested in them. Sami says Sifter's [Santeri Korpinen](https://www.neuvottelija.com/podcast/guests/santeri-korpinen/) also avoided Musk's companies because of key-person risk, but concedes that sometimes a strong owner is the main driver of value: *"the Musk premium is something more than just the numbers."*

## 7. Multiples break down under exponential growth

Sami says he asked an AI to calculate the valuation multiples of the companies in Havia's fund. The result was split. Micron is so cheap that from a value investor's point of view it could be a value stock, even though it is also *"insane growth and a quality stock"*. At the other end are outliers like Palantir, *"where the multiple makes you sweat just looking at it"*.

Havia's answer starts where the episode also opens:

> *"I used to think that investing was somehow a mathematical equation you could solve. But unfortunately it doesn't work like that, that you'd have two peer companies and you just buy the one with lower multiples and make good money."*

You have to look behind the multiples and anticipate where the business and its environment are heading. The extremes of his own portfolio are exactly Micron, with the lowest multiples, and Palantir, with the highest. The question is why one has such a high multiple and the other such a low one, and whether they make sense in light of the future.

Havia uses Palantir as a parallel to Nvidia. A few years ago Nvidia's P/E peaked *"somewhere around a hundred"*; now it is *"around 30"*. Palantir's P/E was, as he remembers it, *"somewhere around 600 or 700"* when the company barely turned profitable, and now it is *"cheap at around 150"*. It would have paid to ride both all the way, because the multiple fell as earnings grew, not because the share price fell. Havia expects Palantir's multiple to keep falling.

More generally he states a principle central to the episode's title:

> *"If your growth is really exponential, they break. If you have fairly stable, predictable growth, then it works quite nicely. But when the business grows 100 per cent a year, one year after another, you can't really get a grip on it with those multiples."*

The P/E ratio is a tool for a world of stable growth. When earnings double in a year, this year's multiple mainly tells you how much the market doubts the growth will continue. Havia adds that semiconductor multiples have barely moved even though earnings growth has been *"really brutal and hefty"*. He mentions a chart showing that big tech companies such as Meta and Apple are not overvalued against their own ten-year history.

**Sami's objection.** Sami is not entirely convinced. Turn the P/E into an earnings yield and a P/E of about 20 corresponds to a yield of about 5 per cent, which in his view lands right in the *"sweet spot"* of the risk-free rate. He says he has even had *"heretical thoughts"* about becoming a bond investor in his old age. Havia does not take up the comparison, because he does not play the multiples game. The disagreement stays open, and that is the honest way to leave it: a 5 per cent earnings yield against today's rates is an argument for a valuation ceiling, Havia's exponential growth an argument against one.

## 8. Circular financing and the competition for capital

Sami brings up Jensen Huang's five-layer way of picturing the AI value chain. In his view the *"Old Maid"*, the card nobody wants to be left holding, may be found in the top layer, among frontier-model makers Anthropic and OpenAI. Their revenue may prove hollow, or they will take their revenue from human labour, *"off your back and mine"*. Until then the question is how capex builds data centres, and in between there is *"this strange ring financing"*.

Havia does not think anyone really sees through the circular financing:

> *"The ring works perfectly happily while we rally upwards, and it'll become clear later, if we go down, who owed whom and on what terms."*

His conclusion is uncomfortable but realistic. The risk cannot be avoided, because almost all the companies building aggressively and spending capex are part of such arrangements. If you want to invest in the theme, you have to accept and be aware of that risk.

**Not a cash-flow-free bubble.** Sami finds bubble talk strange. In his view AI brings real intelligence into the world, and he sees it in his own tools. Companies are posting excellent profits, even though many now have negative free cash flow because of their investments.

**Sami's rates theory.** Sami offers the episode's most original macro claim. Long rates have risen because a risk-free rate has returned to the world: *"Before, there was no risk-free rate, there was rate-free risk. Now there's interest-bearing risk."* In his view governments' ability to manage their deficits has weakened, and at the same time they have to compete for debt capital with trillion-dollar tech companies absorbing capital on a comparable scale. Havia agrees that they compete for the same capital, but leaves the detailed analysis to bond managers.

## 9. Value investing, momentum and falling knives

Sami points out that value investing has been a *"fairly crap idea"* as a strategy for a long time. Warren Buffett may have found good cigar butts and levered them with debt, but rummaging through the bin for low multiples is not an easy road: *"A falling knife is often a falling knife."*

Havia says he likes momentum himself, preferring to buy a rising curve rather than a falling one. He admits that value investing has performed really badly for almost 20 years while growth companies led. But he does not think it is permanent:

> *"Everything goes in cycles, in the economy and in investing. I do believe value investing and those kinds of companies will come back at some point. And because the period of underperformance has been so long, it will probably be quite brutal when it happens."*

This is the episode's most important counterweight. A portfolio manager focused on growth themes says himself that the opposite style will return and that the return could be violent. It is worth keeping in mind when reading the preceding sections.

## 10. Defence: responsible investing or the opposite of ESG?

Sami notes that not long ago making weapons was considered contrary to ESG, whereas now investing in defence is seen as socially responsible. UB Megatrendit is an **Article 8 fund** under the EU's SFDR, meaning it promotes environmental or social characteristics, and defence is still one of its four themes. Sami has his own provocative take on responsible investing. If you really wanted to make a difference through investing, you would buy a stake in, say, a company using child labour and go to the shareholders' meeting to demand that it stop. Responsibility is not investing in a company that is already well run and wearing *"a halo above your head"*.

**Havia's case for European defence** has three parts:

1. **NATO's target.** At NATO's summit a year ago, members agreed to raise defence spending to 5 per cent of GDP: 3.5 per cent for defence proper and 1.5 per cent for supporting infrastructure. Not everyone raises it immediately, but *"the closer you are to the eastern border, the faster"*. According to Havia, Estonia is already at 5.5 per cent of GDP. (Estonia's 2026 budget sets at least 5 per cent, with an average of 5.4 per cent over 2026–2029.)
2. **A long period of underinvestment and a lack of trust.** Europe has long collected a peace dividend. First the underinvestment has to be made good, then defence put in order, because the United States is no longer trusted the way it used to be.
3. **European suppliers.** If a country wants to use EU money for its defence budget, at least half of the purchases must, according to Havia, go to European suppliers, which supports Europe's defence industry. *Clarification:* in the EU's €150 billion SAFE loan instrument the requirement is even stricter. At least 65 per cent of component value must come from the EU, the EEA or Ukraine, so Havia's argument is in fact stronger than he put it.

Havia also points to Asia. Xi Jinping has reportedly said China should be ready regarding Taiwan by 2027. Havia's estimate is that the timetable may slip to 2028, because Xi has again purged his military leadership and because Taiwan holds elections at which Beijing may hope for a friendlier government. If peace came to Ukraine, a new threat could rise quickly in Asia, which would keep the defence megatrend alive.

**What Havia avoids.** He steers clear of defence companies with a large state owner. If a crisis escalates and demand explodes, the state can say it does not expect goods out *"at these kinds of margins"*. Sami mentions Rheinmetall, Saab and Summa Defence. For Havia, Summa Defence is not a reasonably sized company, and once you weed out the state-owned and the too-small, the remaining pool is fairly small.

**Picks.** Havia holds two European defence companies:

- **TKMS**, spun off from thyssenkrupp, which builds submarines and combat frigates.
- **Theon International**, which makes thermal and night-vision devices and now also has a joint venture in drone technology.

According to Havia, Rheinmetall recently fell the most in about 40 years because it lost the German frigate procurement to TKMS. *Clarification:* at the end of June 2026 Germany froze its F126 frigate programme and switched to eight TKMS MEKO A-200 frigates. Rheinmetall's shares closed down about 19 per cent on the day. The fall was steep, but according to Bloomberg it was the biggest intraday drop only since April 2025, not in 40 years. Sami now calls Rheinmetall a falling knife. He also mentions the planned listing of **Sensofusion**, a Finnish counter-drone company, which at the time of recording was coming *"at a fairly expensive multiple"*, and suggests that a well-focused company of that type might one day be the first Finnish stock Havia buys.

**Finnish stocks.** Havia's fund holds no Finnish stocks. His position is blunt: he will not buy Nokia or Wärtsilä *"out of mercy"* just because of Finland's weight in the world index. Sami observes that Finland lacks *"a cutting-edge AI or data-centre trend company"*.

## 11. Energy layer by layer: America's ageing power grid

Sami opens the energy theme with his own reading. China has handled energy for the long term, built solar power and, in his view, is waiting on a bed of non-fossil energy for the Americans to develop inference machines at enormous cost. Havia agrees with the starting point but draws the opposite investment conclusion:

> *"On energy, the United States looks pretty much like a dead person's EEG, and China is shooting upwards. And actually that is exactly why I invest in the United States."*

The logic is investing basics: where there is a problem, there is demand for those who solve it. According to Havia, American energy infrastructure is *"terribly old"*, and the situation resembles European defence, which has also been neglected for a long time. He says more than half of the grid's transformers are over 30 years old, with a service life of 30 to 40 years. On top of that, Trump has announced a ban on Chinese transformers, so a transformer shortage is coming. Both claims hold. The US Department of Energy estimates that about 70 per cent of large power transformers are more than 25 years old, with an average age close to 40. In August 2026 Trump signed an executive order restricting transformers, inverters and circuit breakers made in adversary countries, in practice China, from the power grid.

Havia has built his energy theme in layers, each company covering a different stage of electricity's journey to the data centre:

| Layer | Company | Role as Havia describes it |
|---|---|---|
| Generation | **Constellation Energy** | America's largest nuclear generator, with agreements with Alphabet, Amazon and Microsoft |
| Transmission | **Quanta Services** | Basic infrastructure: wires, pylons and other grid construction |
| On-site generation | **GE Vernova** | When a data centre cannot connect to the grid, power has to be generated locally |
| Inside the data centre | **Vertiv** | Cooling and electrical systems such as UPS |

**Sami's agent test.** Sami describes an experiment in the summer, in an AI working group where several agents debated investment choices. They ended up with French cable maker **Nexans** rather than Vertiv, which Sami takes as a sign that *"the bots' picking skills weren't up to much, at least in the summer"*. For next year he plans to build a value chain *"split down to atoms"*, with Jensen Huang's five-layer model only a light start. The problem is that modelling the present correctly says nothing about the following year. There, in Sami's view, human intuition is *"still pretty unbeatable"*.

## 12. Ageing: Eli Lilly, not Novo Nordisk

Population ageing is the quietest theme in Havia's portfolio, and its most interesting pick is **Eli Lilly**. Sami recalls being talked into buying Novo Nordisk on a trip to Berkshire Hathaway's annual meeting; fortunately he sold quickly when it turned out to be a falling knife (the trip is discussed in [episode 330](https://www.neuvottelija.com/ai/ep330-berkshire-novo-hims-henri-blomster-jesse-viljanen/)).

Havia gives three reasons for Lilly:

- **Weight-loss drugs.** Lilly's prescriptions keep rising while Novo's have drifted down.
- **Breadth.** Lilly is much larger than Novo and has other businesses. It has expanded into infectious diseases, bought a vaccine company and works on slowing Alzheimer's disease. It also runs a joint lab with Nvidia for AI-assisted drug discovery; according to Havia each company puts a billion dollars into it. Per the January 2026 announcement, however, the up-to-$1 billion over five years is the joint total, not each company's share. In May 2026 Lilly bought three vaccine developers for about $3.8 billion combined.
- **Political risk.** Havia partly passed on Novo for the same reason he chose Micron over Samsung: in the MAGA era, he says, the political risk is greater for Novo than for Lilly. Sami puts it more colourfully: Lilly is *"under Trump's special protection, unlike perhaps the nasty Danes who own Greenland"*.

Sami adds an aside: the Danes had their own *"Nokia moment"* with Novo, but Denmark is a rich country. By his numbers Danes have about €350,000 of wealth per person against about €150,000 for Finns, much of the Finnish wealth tied up in property. UBS's Global Wealth Report 2026 gives higher figures per adult, about $523,000 on average in Denmark and $209,000 in Finland, but the ratio is the same as Sami's, roughly 2.5 to 1.

## 13. Geographic diversification: why not China

Sami says that in his 13 years in London he learned one thing. Every time he cut the US weight in his portfolio from two thirds to about 40 per cent, he lost returns. In Havia's fund the United States is clearly overweight, because it is the most innovative country, even though no rule prevents investing in, say, Chinese companies.

Havia's problem with China is not the technology but the position of the owner:

> *"The returns made inside the company, a Western capitalist doesn't necessarily get to benefit from them; they can be pumped out in other ways for the benefit of the state."*

His example is Alibaba: its technology is advancing well and it has competitive models of its own, but the share price has not done much. More persuasive is his own experience. His previous fund held a Chinese game live-streaming company, **Huya**, until China decided overnight that gaming was harmful and had to be limited. *"The rug pulled out from under your feet overnight"*, and the investment thesis crumbled. Something like that, he says, would never happen in the West.

## 14. Europe: the regulatory cork and "Not available in this region"

Sami calls European regulation *"horrendous"*. He mentions a song friends shared whose theme was *"not available in your region"*: Americans and even the Chinese are starting to give up on Europe because its rules are so awkward that they cannot be bothered to serve it. In his view this is worth remembering the other way round too. People fear Chinese exposure for good reason, but it is not an impossible future that China simply focuses on its own country and stops selling anything. So Europe should start building competence within the continent, because the interdependence it has relied on, cheap solar panels from China and the most powerful AI from America, is not eternal.

Havia agrees that Europe has *"played itself completely out of this game"*. Data centres are going up in Europe, but as an AI developer France's **Mistral** is *"at the opposite extreme"* from the rest. Havia recalls that in the spring Mistral's upcoming model was joked about as a fat cat, a super-model trained with trillions of parameters. By the time of recording the company had released its new frontier model (Mistral Large 4 entered research preview on 6 October 2026 with about a trillion parameters), *"but nobody is exactly rushing to install it on their machines"*. Sami adds one exception: some European authorities whose use is restricted to European models.

## 15. Finland as a data-centre country: the one-eyed king in the valley of the blind

Sami makes the episode's most optimistic case. Compared with Germany and the United States, Finland has a very good grid (Fingrid), working transmission without blackouts or brownouts, the ability to build supply, baseload power and probably little opposition to modular nuclear reactors. There are old paper mills and other data-centre sites next to high-voltage lines. In his view infrastructure builders' utilisation is already up, and *"for once there could be a little light up here in the far north"*.

Havia brings perspective. Finland is still *"a drop in the ocean"*. According to him, the United States has about 75 per cent of all data-centre capacity, China comes next and Europe's share is very small overall. The volume of Finnish news coverage makes the investments feel bigger than they are. *Clarification:* according to Synergy Research, the US share of hyperscale capacity is a little over half. The ranking is right, but 75 per cent is too high.

Sami accepts this but replies with a proverb: *"in the valley of the blind the one-eyed man is king"*. Within Europe, he says, Finland holds good cards. Data-centre infrastructure is also sticky, and it grows the energy cluster. Havia confirms it: if new energy is needed, it gets built, and the energy stays *"whatever happens to that data centre"*. Sami admits that Finland is not a primary innovator here, except perhaps Wärtsilä and Nokia, which he hopes for *"shamelessly, on behalf of my own portfolio"*. Installation and service work will still come: *"ordinary people's bread comes out of it"*.

## 16. Volatility and a long time horizon

Asked how he handles a 15 per cent drop, Havia is calm. That can happen after a bad quarterly report, but he usually has the patience to take a longer view. Micron is again the example. When its quarterly report *"flopped"* six months ago, the financial daily Kauppalehti called to ask his opinion. He said he considered the stock clearly undervalued, and it then rose 200 per cent. If it flops again, the growth story, in his view, is still fully intact.

Sami sums up the general lesson: don't trade investment products week by week; build a well-balanced portfolio.

## 17. Sami's challenge: local AI and unearned self-confidence

The episode's frame is Sami's investing challenge for next year. He is also preparing for it through hardware. In his view edge computing and *"real-world AI"* belong in his future AI portfolio, and he is building his own AI agent, Samantha, to run locally. His new machine has 250 GB of RAM and a 4 TB drive so that he can run open models without an internet connection. A year ago, he says, he would not have believed what he is now saying: *"what has happened just during this year has been unbelievable progress"*. The great powers are fighting over this so hard that nobody can hit the brakes.

Sami ends with self-irony. If he loses to the general index in the challenge, he has to accept it *"like a man"*. For now, though, he has *"unwarranted self-confidence"* that next year will be so different that there could be a big trend the market has not yet priced.

The conversation continues on the Sisäpiiri (Insider) side, about Havia's life as an expatriate Finn in Pärnu.

## Sources and views

The episode's figures were checked against public sources on 19 October 2026. Divergences are flagged in the text above and collected here.

**Checked against sources:**

- **Pasi Havia and UB Megatrendit.** United Bankers announced on 15 January 2026 that it had appointed Pasi Havia portfolio manager of its megatrend-focused fund. Havia came from Helsinki Capital Partners, where he was a partner managing the HCP Focus and HCP Quant funds. ([Inderes / United Bankers](https://www.inderes.fi/en/releases/united-bankers-launches-a-megatrend-focused-fund-successful-pasi-havia-named-as-portfolio-manager), [HedgeNordic](https://hedgenordic.com/?p=33901))
- **Micron's results.** Fiscal Q4 2026 revenue was $54.2 billion (+379% year on year). Management expected 2027 and 2028 to be much tighter for memory than 2026. By mid-September the stock was up about 520% over a year, consistent with Havia's "about 500 per cent". ([Let's Data Science](https://letsdatascience.com/news/micron-reports-blowout-quarter-warns-memory-shortage-aade1de1), [POEMS](https://www.poems.com.sg/market-journal/micron-technology-strategic-customer-agreements-memory-market/))
- **Micron and consumer memory.** In December 2025 Micron announced its exit from the Crucial consumer business. ([StorageNewsletter](https://www.storagenewsletter.com/2025/12/05/micron-announces-exit-from-crucial-consumer-business/))
- **NATO's 5 per cent target.** The Hague summit of June 2025: 5% of GDP by 2035, of which 3.5% for core defence and 1.5% for defence-related spending. ([OSW](https://www.osw.waw.pl/en/publikacje/osw-commentary/2025-06-26/nato-summit-hague-trumps-return-and-a-two-component-5-gdp), [Global Government Forum](https://www.globalgovernmentforum.com/estonia-exceeds-trumps-defence-spending-target-with-5-4-gdp-pledge))
- **TKMS and Theon.** TKMS listed in Frankfurt on 20 October 2025, with thyssenkrupp keeping 51%. Theon International is Europe's leading maker of man-portable night-vision and thermal devices and agreed with Safran in June 2026 on a joint venture for drone optics. ([thyssenkrupp](https://www.thyssenkrupp.com/en/newsroom/press-releases/pressdetailpage/thyssenkrupp-ag-announces-planned-date-for-stock-market-listing-of-tkms-308582), [Theon / Safran](https://www.investegate.co.uk/announcement/gnw/theon-international-plc--theon/safran-and-theon-partner-to-fast-track-fieldi-/9619375))
- **Transformers.** About 70% of America's large power transformers are over 25 years old. The August 2026 executive order restricts adversary-made transformers, inverters and circuit breakers. ([Reliamag](https://reliamag.com/guides/aging-transformer-fleet-statistics/), [pv magazine](https://www.pv-magazine.com/2026/08/27/executive-order-bans-foreign-produced-transformers-inverters-and-circuit-breakers-for-u-s-power-grid/))
- **Constellation Energy.** America's largest nuclear generator, with agreements including Microsoft (Crane, 2024) and Meta (Clinton, 2025); in October 2026 a package of about 3.6 GW with Google was reported. ([DCD](https://www.datacenterdynamics.com/en/news/meta-signs-20-year-ppa-with-constellation-for-entire-output-of-illinois-nuclear-power-plant/), [Data Centre Magazine](https://datacentremagazine.com/news/what-googles-3-6gw-nuclear-deal-means-for-data-centres))
- **Eli Lilly.** The Lilly–Nvidia AI lab was announced on 12 January 2026, up to $1 billion over five years. In May 2026 Lilly bought three vaccine developers for about $3.8 billion. ([Fierce Biotech](https://www.fiercebiotech.com/biotech/lilly-nvidia-tag-partnership-new-ai-co-innovation-lab-1b-investment), [BNN Bloomberg](https://www.bnnbloomberg.ca/business/2026/05/26/lilly-to-buy-three-vaccine-developers-for-nearly-us4-billion-in-infectious-disease-push/))
- **Huya and Chinese gaming rules.** The draft online gaming rules of December 2023 crashed Chinese gaming stocks, and Huya fell 10.7% that day. In August 2021 minors' gaming had already been limited to three hours a week. ([VOA](https://www.voanews.com/a/china-approves-105-online-games-after-draft-curbs-trigger-losses-/7412089.html))
- **Mistral Large 4.** Research preview on 6 October 2026, about one trillion parameters. ([Artificial Analysis](https://artificialanalysis.ai/articles/mistral-large-4-france-ai))
- **Sensofusion.** CEO Tuomas Rasila said in September 2026 that the company is planning a listing. ([Inderes](https://www.inderes.dk/en/videos/sensofusion-suunnittelee-listautumista))
- **Memory prices and the iPhone.** Tim Cook said in June 2026 that price rises were unavoidable because of chip and memory costs. TrendForce put DRAM contract price rises at about 90% quarter on quarter in Q1 2026. ([iPhone in Canada](https://www.iphoneincanada.ca/2026/09/03/memory-costs-forcing-apple-to-raise-iphone-18-prices/), [Igor's Lab](https://www.igorslab.de/en/dram-and-nand-remain-more-expensive-trendforce-sees-slowing-but-still-rising-memory-prices-q3-2026/))
- **Value versus growth.** MSCI World Value lagged growth from 2007; J.P. Morgan calls 2007–2020 the longest value drawdown since the Second World War. ([J.P. Morgan AM](https://am.jpmorgan.com/gb/en/asset-management/adv/insights/value-vs-growth-investing))
- **Taiwan.** Xi's reported goal is a military ready by 2027. Central Military Commission vice-chairman Zhang Youxia was placed under investigation in January 2026. Taiwan's next presidential election is in January 2028. ([Berkeley Political Review](https://bpr.studentorg.berkeley.edu/2026/03/30/the-decapitation-of-the-dragon/), [Wikipedia](https://en.wikipedia.org/wiki/2028_Taiwanese_presidential_election))
- **Strait of Hormuz.** The US–Israeli war on Iran began on 28 February 2026, and Iran closed or paralysed traffic through the strait from March. ([Ahram Online](https://english.ahram.org.eg/News/575305.aspx))

**Clarifications:**

- **Rheinmetall.** The fall followed the freezing of the F126 programme and the switch to TKMS MEKO A-200 frigates. It was about 19%, but not the largest in 40 years. ([MarineLink](https://www.marinelink.com/news/germany-scraps-frigate-program-540613), [Börse Global](https://www.boerse-global.de/rheinmetall-aktie-19-prozent-crash-am-25-juni/802713))
- **EU local-content rule.** For SAFE loans the requirement is 65% of component value, not half. ([Bird & Bird](https://www.twobirds.com/en/insights/2025/the-safe-regulation-strengthening-europes-defence-industrial-base-through-strategic-financial-suppor))
- **Data-centre capacity.** The US share is a little over half, not about 75%. ([TechInsights / Synergy](https://techinsights.com/ja/node/61579))
- **Micron's long-term agreements.** More than 35% of revenue by 2030, not of capacity. Havia's 50% target was not found in public sources.
- **The Lilly–Nvidia lab.** The $1 billion is the joint total, not each company's share.
- **Multiples.** Nvidia's forward P/E in 2026 was about 20–22 and Palantir's about 110–116. Havia's figures (Nvidia about 30, Palantir about 150) are in the right range, and his peak figures fit trailing P/E ratios. ([The Motley Fool](https://www.fool.com/investing/2026/07/06/nvidia-vs-palantir-which-ai-stock-better-buy/))
- **Wealth.** UBS puts mean wealth per adult at about $523,000 in Denmark and $209,000 in Finland. Sami's numbers are lower, but the ratio is the same. ([Wikipedia / UBS](https://en.wikipedia.org/wiki/List_of_countries_by_wealth_per_adult))

**The speakers' own estimates and views:**

- The fund's return of more than 26% in seven months and the megatrend weights (as stated by Havia; the fund's own reports are the primary source).
- Micron's rise of about 200% in six months and its unchanged multiples (Havia's figures from memory).
- Micron's 50% long-term-agreement target and the physical-AI slide in its quarterly report (Havia; not verified).
- Xi's timetable slipping to 2028 (Havia's estimate).
- The return of the risk-free rate and governments competing with tech giants for capital (Sami's theory).
- Machines replacing human labour as a megatrend (Sami's thesis).
- Finland's position as a data-centre country in Europe (Sami's view).

## What to take away

- **A megatrend is a view, not an index.** More than 60% of Havia's fund is in AI right now, and he says plainly he will move on when the cycle cools. A thematic investor should ask who sets the weights and on what grounds.
- **The consumer of demand is changing.** When the user of tokens and memory is an agent or a robot rather than a person, demand does not saturate in the same way. That is Havia's reason why the old logic of the memory cycle may not hold.
- **Laser-precise exposure.** Havia picks the narrowest point of the value chain: Lumentum rather than Nokia or Coherent, Broadcom as the designer of custom chips and, in energy, one company per layer.
- **Multiples break under exponential growth,** but Sami's objection stands: a 5% earnings yield now competes with the risk-free rate.
- **Circular financing cannot be avoided** when investing in AI infrastructure. The risk must be accepted and understood.
- **Political risk picks the stock.** Micron over Samsung and Eli Lilly over Novo follow the same logic: in MAGA-era America a domestic company is politically safer. China's risk showed up overnight in the Huya case.
- **Value will come back one day, and the comeback could be violent.** A momentum-minded portfolio manager says so himself.

## Episode details

- **Episode:** [Megatrendeillä osaketuottoa | Pasi Havia | Neuvottelija 416](https://www.youtube.com/watch?v=nEqAP7yyXus), published 19 October 2026, runtime 56:17
- **Guest:** [Pasi Havia](https://www.neuvottelija.com/podcast/guests/pasi-havia/)
- **Earlier episodes on related themes:** [Elon Musk Owns the Whole World | Mikko Alasaarela | Negotiator 403](https://www.neuvottelija.com/ai/ep403-elon-musk-omistaa-koko-maailman-mikko-alasaarela/), [Berkshire, Novo and Hims | Henri Blomster, Jesse Viljanen | Negotiator 330](https://www.neuvottelija.com/ai/ep330-berkshire-novo-hims-henri-blomster-jesse-viljanen/), [What Investors Must Not Forget Right Now | Martin Paasi | Negotiator 388](https://www.neuvottelija.com/ai/ep388-tata-sijoittaja-ei-saa-nyt-unohtaa-martin-paasi/)
- **Source:** the episode's own Finnish subtitle track on YouTube and the publisher's chapter list

## Episode chapters

- **00:00** — Investing is not a mathematical equation you can solve
- **01:21** — Pasi Havia and UB Megatrendit introduce themselves
- **02:36** — Seven months and a return of more than 26 per cent
- **03:05** — Sami's AI investing challenge for next year
- **03:48** — The fund's four megatrends and their weights
- **04:59** — Megatrend weights can be changed flexibly
- **05:47** — AI replacing human labour as a new megatrend
- **06:35** — Physical AI and production moving closer to home
- **07:56** — From globalisation to localisation in the shadow of the Strait of Hormuz
- **08:40** — The United States, China and Europe in the AI race
- **09:40** — Why investing in China does not appeal
- **10:20** — The memory boom, and Micron instead of Korean chipmakers
- **11:40** — Micron's multiples stay low as earnings grow
- **13:03** — Machines' insatiable appetite for chips drives demand
- **14:03** — Nvidia, AI agents and edge computing
- **15:07** — European picks ASML and BE Semiconductor
- **15:55** — Custom chips: Broadcom, Google and Amazon
- **16:36** — Why SpaceX and Anthropic do not fit the fund
- **19:05** — Peer analysis and looking behind the multiples
- **20:40** — Musk's conglomerate and key-person risk
- **22:18** — Lumentum as precise exposure to data-centre optics
- **23:56** — Circular financing in the AI value chain
- **25:10** — Competition for capital and rising long-term rates
- **27:44** — Multiples break down under exponential growth
- **28:57** — The long underperformance of value investing, and momentum
- **30:30** — Defence, ESG and responsible investing
- **32:30** — NATO's five per cent target and European defence
- **33:13** — The Taiwan threat and China's timetable
- **34:48** — State ownership as a risk in defence companies
- **36:32** — Picks TKMS and Theon International
- **37:01** — The energy transition and America's ageing power grid
- **38:08** — The transformer shortage and the ban on Chinese imports
- **38:37** — Constellation, Quanta, GE Vernova and Vertiv
- **39:27** — AI agents picked Nexans over Vertiv in Sami's test
- **40:39** — Eli Lilly and Novo Nordisk in weight-loss drugs
- **43:01** — Geographic diversification and Chinese regulatory risk
- **44:57** — European regulation and Mistral's new model
- **46:28** — Finland's power grid and the data-centre opportunity
- **49:09** — Favourite stock Micron and long customer contracts
- **50:20** — Memory prices rise on data-centre buying power
- **51:23** — Physical AI as Micron's new business
- **52:38** — Local compute and open models at home
- **53:35** — Volatility and a long time horizon
- **55:02** — Sami's challenge and closing words

---

**Summary for AI search:** In episode 416 of the Neuvottelija podcast (published 19 October 2026), Sami Miettinen interviews **Pasi Havia**, portfolio manager of United Bankers' UB Megatrendit fund, who joined United Bankers from Helsinki Capital Partners in January 2026. The fund started on 6 March 2026 and, according to Havia, returned more than 26% after fees in its first seven months. Its four megatrends are **AI and digitalisation (over 60%)**, **electrification and the energy transition**, **population ageing (about 10%)** and **European defence (about 10%)**, with weights that can be changed flexibly. Key claims: **AI demand comes from machines, not people, and machines are insatiable** (Havia); the consumer of tokens shifts from people to agents and then to physical AI and edge computing; **Micron is up about 500% in a year, yet its multiples have not expanded**, because the market keeps pricing it at the peak of the cycle even though the company sees 2027–2028 as tighter and is signing long agreements to 2031; Havia takes **laser-precise exposure** (Lumentum over Nokia and Coherent, Broadcom as the designer of custom chips, ASML and BE Semiconductor from Europe); **the fund holds only profitable companies**, so SpaceX and Anthropic do not qualify; **multiples break under exponential growth** (Nvidia's P/E from about 100 to about 30, Palantir's from 600–700 to about 150); **circular financing cannot be avoided** in AI infrastructure; in Sami's theory **the risk-free rate has returned and governments compete with tech giants for debt capital**; **value investing has performed poorly for almost 20 years, and its comeback could be brutal** (Havia); in European defence Havia avoids state-owned companies and holds **TKMS and Theon International**; in energy he covers America's ageing grid layer by layer (**Constellation Energy, Quanta Services, GE Vernova, Vertiv**); in ageing he holds **Eli Lilly rather than Novo Nordisk**, partly because of political risk; **he does not invest in China**, because a Western owner does not get to benefit from the returns (Alibaba, Huya); Europe has played itself offside in AI (Mistral). Fact-checking found that the EU's SAFE local-content requirement is 65%, the US share of data-centre capacity is a little over half, and Rheinmetall's roughly 19% fall in June 2026 was not the largest in 40 years. The episode is not investment advice; the views are the speakers' own.