Algirdas

Abeona is a company selling Zevaskyn therapy – hope for patients with a rare, devastating skin disease – recessive dystrophic epidermolysis bullosa (RDEB). It is one of the last speculative stocks we still own. Excited about the promise of cell and gene therapy we bought it in 2021 when our investment style was still taking shape. We would not be buying it today, but we are not yet cutting substantial losses as we want to learn this lesson fully – commercialization story is about to deliver its outcome.

RDEB is caused by genetic mutation. As a result, skin lacks "glue" that is holding layers together and tears at the slightest friction. Patients develop chronic, excruciating wounds over their bodies. Many die young from infection. The standard care is Vyjuvek gel, but it requires indefinite weekly dosing. It is better suited for smaller wounds. Large, chronic, non-healing wounds that define the disease remain a source of suffering.

Abeona's Zevaskyn provides durable wound healing for up to eight years after a single application. A second one will be needed in the patient's lifetime. Both will cost 6.2 million dollars in total. It still makes sense economically, because Vyjuvek costs tens of millions cumulatively in a lifetime. We understand that reimbursement authorization in the U.S. is 100% – health insurance is willing to cover.

Abeona has already identified 100 eligible patients. Management think in the U.S. alone there are 750 people that will need Zevaskyn therapy. That is 4.6 billion dollars in potential revenue on the cards, not counting any international expansion. Even at a valuation of 0.66 billion, which is the price we paid for our shares, there is still a possibility for a multibagger return.

Abeona has manufacturing capacity to treat six patients per month and targets to expand it to ten, which for now seems like a hard limit. If the company can serve more than three patients a month – roughly 40 a year – it crosses into profit zone, according to CEO Vishwas Seshadri. We have taken note, and if it is not the case by the start of 2027 – we will exit the position. The clock is ticking, because Zevaskyn was approved April 2025, and nearly a year since, as of the start of 2026, only two commercial patients were treated amid delays. We think the company will have enough cash to run to the middle of 2027 if profitability is not reached.

CEO Seshadri has some six million dollars of wealth kept in Abeona shares, but he is trimming his share, not buying. He and his Chief Commercial Officer Madhav Vasanthavada both have experience in launching complicated therapy, logistically similar to Zevaskyn. They led Breyanzi (liso-cel) at Celgene-BMS.

Zevaskyn takes patients' own skin cells, transduces them ex vivo with functional genes. It then expands them into credit-card-sized sheets. Those sheets are surgically grafted onto wound sites. The corrected cells integrate permanently into the genome and start producing functional collagen needed to hold skin together. Every batch is manufactured for one specific patient. The sheets have an 85-hour shelf life once released. Patients must come to one of only four qualified centers and stay hospitalized for at least five days. The journey from consultation to treatment takes four months. To sum up, margin for error is thin, but the path to profitability is visible at relatively modest volumes.

Any delays and manufacturing failures will sink Abeona. It will prove that autologous cell therapy commercialization is too complex, too slow, and too expensive for a small biotech. A similar case of Bluebird Bio already exists. It had three comparable approved therapies and failed to achieve viable commercial scale. The uncertainty over the future of Abeona is very high. But we have committed only a fraction of our portfolio to the company. We are ready to see how the case plays out soon to learn the lesson of investing in biotech with our skin, so we never repeat the same mistake again.

Asseco Business Solutions (ABS) is the central nervous system for tens of thousands of businesses in Poland. It provides software essential for running the company: accounting, customer relationships, supply chain, payroll, employee records, project management, etc. These services are called Enterprise Resource Planning (ERP).

ABS was one of our first truly independent discoveries as we were searching the Warsaw Stock Exchange manually in 2021. At the time valuations of companies in the US and Europe were very high amid the post-pandemic boom. Meanwhile, ABS was trading for what would be an average company during average times in the US: 15 times earnings per share. It meant a valuation of 1.17 billion zloty (280 million CHF). The company had virtually no debt, consistently rising revenue, high profit margins and plentiful free cash flow with a stable and increasing dividend. We bought it and added more almost every single year since then.

ABS was forged in 2007 by merging five predecessor companies. Wapro targets micro and small businesses plus accounting firms. It has many clients, but they pay relatively little. Softlab addresses large enterprises with complex needs. These clients pay substantially more, but there are considerably fewer of them. Macrologic Merit serves the mid-market with process-oriented software. All the services are set as a ladder so customers can stay with ABS even if they grow beyond their initial size. Employees learn the system. They build processes around it. Cancelling subscription means month-long migration, retraining and rebuilding processes.

Our understanding is that once ERP is installed, subscribers in Poland will pay every year for around 10-15 years. A customer may start with accounting, but later can add payroll, manufacturing logs, analytics, supply-chain optimization, and increasingly these days, AI assistants. Not impossible to leave, but difficult, so we worry less about alternatives that do exist in Poland. The leader is Comarch with a market share of around one fourth. SAP holds a similar share, but it is concentrated in large enterprises. ABS is probably fighting for a distant third position with a market share somewhere around 10%, according to our educated guess. There are many other dynamic players, including promising start-ups.

ABS belongs to the Asseco Group, Poland's largest IT conglomerate. Asseco Group provides to ABS research and development sharing, brand credibility, and infrastructure. These advantages are inflating margins and returns higher than would be possible for a stand-alone company. We are also aware that parent control means management's primary accountability runs upward to Asseco Poland. Top management at ABS is full of managers working long-term. They were promoted internally within the Asseco ecosystem. The CEO is Wojciech Barczentewicz, who has served since 2012. Collectively, leadership holds more than 5% of shares, with Barczentewicz's part standing at 1%. That is around 5 million CHF. It feels confident parking our share alongside his wealth.

We understand that the recent mandatory KSeF e-invoicing rollout in Poland was a successful tailwind for ABS. Customers prefer software that fits local business practices, regulations, Polish language, and workflows. ABS products serve as systems of record for a lot of data. Even if AI is to play an increasing role inside various companies, it will need that data to function properly. Thus, we see an opportunity for ABS to sustain its clients. The company has its own AI offerings branded under the "Genius by Asseco" name. It offers predictive analytics, smart assistants, and document automation. We have noted that over the past several years ABS consistently increased revenue, but the number of employees stayed stable — around 1,100. Salaries make up an important part of costs in the software business. Therefore, we treat it as a potential sign that the spread of AI for now helps ABS earn more rather than threaten it.

However, AI may open new opportunities for established leaders such as Comarch and disruptive newcomers such as Symfonia or enova365. Only 24% of small Polish firms currently use ERP systems versus an EU average of 33%. Thus, there is still territory to conquer, but ABS is hardly spending any money to acquire or upgrade long-term assets such as technology or equipment. Its three offerings for small, medium and large clients require ABS to keep three distinct technology stacks for maintenance. ABS is optimized for harvesting rather than adapting or expanding. So we do not expect it to perform as a growth stock. But we do expect to get a decent return if things stay where they were for the past five years. It would mean around 1.6 times growth in earnings, and virtually all earnings are paid out as dividends — ABS is the largest contributor to our dividend income. The last price we added meant we expect to get back that investment in dividends after taxes alone in 16 years. If the price dips below, we will be ready to add.

Games Workshop is the only publicly listed pure-play miniatures company of its kind globally. It designs, manufactures, distributes and sells fantasy miniatures based on universes it has created and owns. We discovered Games Workshop in 2022 on a list of companies followed by the investment research and consulting firm Edison Group. It was almost love at first sight, and we bought our first shares while celebrating one of our birthdays in Paris. A perfect gift cost us around 16× trailing earnings — the company valued near £2.2 billion. The stock had been beaten down at the time amid inflation worries. We have added to our position every year since, although, of course, never again at a multiple quite so attractive. Our subsequent entry points have usually been around 25 times trailing earnings.

Games Workshop sells licensed miniatures based on The Lord of the Rings — the wonderful fantasy world created by J. R. R. Tolkien that we both love. For business, however, far more important are the worlds created and owned by Games Workshop itself. First comes the space-fiction universe of Warhammer 40,000 — the flagship — together with Warhammer: The Horus Heresy, which explores events set roughly 10,000 years earlier. Then comes the fantasy side of Warhammer, represented today primarily by Warhammer Age of Sigmar and Warhammer: The Old World.

It has taken more than four decades to populate these worlds with fantastical characters, factions, histories and thousands of interconnected narratives, continuously expanded through novels, short stories, audio dramas and other publications. The universes attract hobbyists interested in science fiction and fantasy. They are collectors, painters, model builders, gamers and book lovers. They assemble and personalize their miniatures using Games Workshop's extensive painting system and then build armies to play against each other, from casual games between friends to organized tournaments.

As a result, genuine connections between people are created and local communities are built — in our eyes, a key reason for Games Workshop's existence.

In that sense, play can become a reflection of life itself, carried forward in a freer form. Of course, all of this can be spoiled by commercialism. Games Workshop is, after all, a for-profit company. Our eyes are open to clearly see that there is a fine line between creating stories and experiences that genuinely matter to people and complacently exploiting what has already been created. For now, we trust CEO Kevin Rountree when he says that the company's focus is on the long term and that Games Workshop intends to make the best fantasy miniatures in the world "forever." That long-term mindset will be essential if the company is to resist overextension.

Kevin made his career at Games Workshop, joining as an assistant accountant in 1998 and climbing the ladder before becoming CEO in 2015. He currently owns Games Workshop shares worth roughly £4.5 million. Shareholder value is created, primarily, by not destroying it, according to Kevin. There is no intention of acquiring other companies. Surplus cash is returned to shareholders through dividends after setting aside what is necessary to keep Games Workshop healthy. All of this reassures us when management says that the company "will never run out of things to explore and detail in our truly unique settings."

Interestingly, when we look at Games Workshop's own disclosed principal risks, we find intellectual property protection as the most important, followed by cyber security, global distribution and supply disruption, and loss of key manufacturing and warehousing facilities. Competition from rivals is not mentioned. That makes intuitive sense to us, because Games Workshop largely inhabits a niche of its own.

Whenever we travel to eastern Switzerland, we like to stop for lunch in Bern so that we can visit the Games Workshop store there. These stores, operated directly by the company, act as gateways into "the worlds." They showcase the Warhammer hobby and offer a fantastic customer experience. We understand that selling through independent retailers and online can be economically more efficient, but the stores are key to recruiting new hobbyists and introducing them to the community. Games Workshop now operates almost 600 stores across 24 countries. All but a few are profitable, according to management.

Every year we routinely read about increased pricing. This year's annual report describes the increase as "3% in line with normal levels." The ability to raise prices is important not only for offsetting ordinary inflation and rising costs, but also for absorbing disruptions in global supply chains or tariffs. Management does not treat tariffs as an exceptional event, but "rather part of the uncertainty of operating globally."

The one factor that worries us is Games Workshop's decision to increasingly license Warhammer to other companies, including video game developers and Amazon, which is working on bringing the universe to screens. Management itself admits that the success of these efforts is "broadly out of our control, reliant on the successful development and delivery of projects by licensing partners." We treat this as a serious risk to the integrity of the storytelling.

The Games Workshop hobby remains physical. People build and paint their miniatures and then get together to play with them. Expansion into digital and entertainment creates a very different kind of experience and potentially attracts a very different audience. This could expand the reach of Warhammer, but it also creates the risk of overextended commercialization and gradually surrendering some control over how these worlds are experienced. We are watching how licensing develops and would be prepared to exit if our fears begin to materialize. We don't see that threat for now.

We therefore remain interested in adding more shares of this exceptional business, which has grown its free cash flow by approximately 15% annually over the past decade. We believe something close to that remains feasible over the decade ahead as additional hobbyists join, including in regions currently underrepresented such as Asia or Eastern Europe. At valuations below 25 times trailing earnings, we would be interested in increasing our position.

For many companies in Poland, Mo-Bruk is the only viable option to get rid of their problem — industrial waste. Chemical factories, medical facilities, refining plants, foundries and authorities themselves pay Mo-Bruk to either incinerate waste into ash, convert it into construction materials or fuels, or simply to clean up historical illegal dumps. If Mo-Bruk did not exist, we understand that roughly a third of industrial hazardous waste capacity in Poland would vanish, because there are few alternatives, especially in the southern part of the country where heavy industry is concentrated.

Our first purchase of Mo-Bruk came in November 2021. When checking the Warsaw Stock Exchange we were impressed by its margins — unprecedented in the waste sector. Lower labor costs compared to Western Europe, a low number of permits to process waste, and expensive-to-clean old "ecological bombs" prevalent in Poland are reasons behind the exceptional financials. They have recently somewhat deteriorated, because Mo-Bruk was expanding capacity with a multiyear capital expenditure plan and settling legal issues related to elevated pricing in the past, disputed by local authorities. We have been increasing our position every year when valuation dipped, because beyond the temporary depressed earnings lies an exceptional niche business.

We can count five facilities in Poland dedicated to hazardous waste incineration. Mo-Bruk operates two of those, in Karsy and Jedlicze. Germany has around 30 and France up to 20. These facilities are not easy to build, because of "Not In My Back Yard" human nature. Local communities strongly oppose any plans the moment a project is announced close to their neighborhoods. Permitting takes years. This is why Mo-Bruk's existing plants are a very valuable asset. They were already built and permitted. Getting those same permits approved from scratch today would be hard. If anything changes here — a political push for more permits — we will reconsider our position.

The four-year capex program must be seen in the context described above. It is set to upgrade existing units. Most of it was funded with operating cash flow, but debt was involved, raising its level to somewhat historically high for the company — still conservative, but equal to three and a half years of the company's average free cash flow. The plan was initiated under the previous CEO, Jozef Mokrzycki, who is the founder's son and led the company for many years. The Mokrzycki family controls Mo-Bruk through three foundations; some family members are still involved in management, but the current CEO, who took over in 2022, is Henryk Siodmok. He said 2026 will be the first year for shareholders to see the benefits of the modernization effort. Revenue is targeted to jump around 20%, partly by using installed capacity. We will be monitoring evidence of growth in processed volumes and whether the upgraded units are running close to capacity. If not, we will be ready to reconsider Mo-Bruk's place in our portfolio.

Another element to watch looking forward is how well recent acquisitions — El-Kajo and EcoPoint — are working to expand Mo-Bruk's presence geographically into northern Poland, and activities-wise into oily wastes and shipyard services. Acquisitions are only part of management's push to increase returns for shareholders. Mo-Bruk has consistently paid a substantial dividend, set above 50% of earnings when debt is not elevated. If the dividend growth pattern holds, the valuation in the first part of 2026 would be expected to return the investment through dividends after taxes alone in ten years. That also assumes no growth in earnings per share five years out, if the valuation at 20 times earnings holds. However, we have seen better entry points in the past, when Mo-Bruk traded at under 15 times earnings. We will be patient to add more — the current valuation seems reasonable, but not a bargain.

BioGaia is a niche company which earned its loyal customers – including ourselves – by taking one bacterium very seriously. When our children struggled with infant colic, BioGaia's probiotic drops helped bring relief to some very sleepless nights. Once we discovered that BioGaia was a publicly listed company, we checked its financial statements, balance sheet, and cash flows. We liked what we found, and we initiated a position during 2025 at around 27 times trailing earnings, valuing the entire company at roughly €1 billion.

Infant colic is a frustrating condition whose exact causes remain poorly understood, although the gut microbiome is believed to play a role. BioGaia's best-known product contains the bacterial strain Limosilactobacillus reuteri DSM 17938. Clinical studies suggest it can reduce crying time in many breastfed infants with colic. Exactly why it works is still not fully understood. For exhausted parents like ourselves, however, only one thing mattered at the time: BioGaia.

BioGaia has built its brand over decades by focusing expertise on a single bacterial species: Limosilactobacillus reuteri. It naturally lives in humans, yet may behave very differently depending on their genetic variant – a strain. We understand that growing bacteria consistently of the same strain is much more difficult than engineering pharmaceutical molecules. Much of the process — fermentation, freeze-drying, strain stability — is a trade secret rather than a patent which expires.

Competitors can sell "probiotic drops," but directly copying BioGaia is difficult due to accumulated process know-how. The company has developed strong relationships with pediatricians, who see that BioGaia continuously publishes studies and that its products are inspected repeatedly and supplied to nearly 100 countries. We think pediatricians are conservative, and are not going to recommend new entrants without the same depth of trust easily.

BioGaia deserves our trust because it has recently narrowed its focus. In 2021 management did try to broaden its efforts when it finalized the acquisition of MetaboGen, a company studying the intestinal microbiome. It did not work, and BioGaia has since acknowledged that the investment was impaired. Other acquisitions around that time targeted companies involved in the manufacturing, development, or distribution of existing BioGaia products.

More recently, under the helm of Theresa Agnew since 2023, management has repeatedly told shareholders that it evaluated several acquisition opportunities but could not find ones that met its strategic and financial standards. The company accumulated excess cash, and the decision was taken to supplement the ordinary dividend with extra payouts of up to 100% of earnings when cash flow permits. We like it when management knows how to return cash to shareholders in the absence of attractive expansion opportunities. We like it less when management does not hold a significant portion of its own wealth in BioGaia shares. We tolerate it.

Looking further ahead, a scenario in which BioGaia almost doubles its earnings per share seems plausible to us. The price we paid in 2025 already reflects that expectation. If we get another opportunity to buy below that valuation, we will be interested in adding more. Paying higher prices, however, would require believing the company can grow faster. One area we do not fully understand is management's strategy of taking direct control of distribution in selected markets. We are not yet convinced how this will affect returns on invested capital. Going forward, we want to see whether organic sales grow faster in markets where BioGaia captures a larger share of the commercial relationship. Our initial intuition was that the previous model, relying on distribution partners, was a simpler and better approach. At the same time, we recognize that Theresa Agnew has spent decades commercializing healthcare brands. She may well prove our doubts unfounded. We are happy to give her that opportunity.

Hermès is the company that "sells time" — and forces even the most well-off to expend their time in exchange. Time is the scarcest asset in the world.

Our first entry into owning a piece of this centuries-old brand was during the summer of 2024, when the price of Hermès retreated from its peaks amid a slowdown in the worldwide luxury market. Valuation at the time was approximately €220 billion euros. Now 2026 marks another dip and we are buying again at €160 billion. These entry levels are very high. We would not allow ourselves to commit to any other company at earnings per share as high as 35. But large parts of the world — outside the West — is getting richer. The Middle East and Far East are part of a well-diversified Hermès worldwide presence. We think it is reasonable to expect profits to double in five years, which justifies the valuation. If Hermès is to repeat the growth of past decades, it would take only 19 years to get back the entire investment through dividends after taxes alone. That is a better deal than any real estate we could think of — and Hermès has outlived quite a lot of those.

It crafts items since 1837, initially providing high-quality equestrian equipment to European noblemen. Its leather and fashion items are now about taste, discretion, and having things that are not instantly obtainable. It takes time to craft an item that lasts in quality and desire. Top bags like the Birkin or Kelly require years of waiting. They had founding moments — the iconic photo of the American Princess of Monaco in 1956, or Jane Birkin's plane voyage in 1983 — that cannot be replicated.

Hermès doesn't have a marketing department and doesn't do traditional sponsorships with celebrities. All the energy is dedicated to craftsmanship. CEO Axel Dumas recounts the choice he made during the 2008 financial crisis amid rising gold prices. A veteran craftsman told him no one would initially spot the lower quality of gold in their products.

Axel ordered to continue the Hermès way. Repairs are granted so objects can be passed between generations. An active resale market exists for iconic goods, which evolved naturally based on quality-driven scarcity. Axel is even frustrated when Birkin bags appear on resale too quickly with higher prices, as it harms client relationships.

The Hermès leader is the sixth-generation member of the family. Axel has successfully passed the exam of an aggressive LVMH take-over bid. It united the fifth and sixth generations of Hermès owners to agree to give the family holding company the right of refusal for any stock sale. In exchange for the lock-up, Hermès pays descendants a consistent dividend. The dynasty has put all eggs into the Hermès bag and watches it very closely, with full control and long-term thinking. Only a third of shares are publicly traded, and we spotted family members — for example Dorothée Dumas — buying at the same prices we added ourselves in 2026. We feel confident keeping part of our wealth alongside the Hermès dynasty.

We are also encouraged by the fact that Hermès leadership avoided association with Jeffrey Epstein, who targeted Axel on numerous occasions — through efforts to refurbish his plane in 2013, and through a charity auction in 2016. On both occasions Axel rejected Epstein. These events happened after 2008 conviction but before the public explosion of 2018. It shows sensitivity to reputation.

There are quite a lot of negative reviews about Hermès shops in Geneva from clients complaining they are not pampered there. Whiners threaten to switch to a nearby LVMH store where they allegedly receive a royal welcome. LVMH does have products to compete with Hermès offerings, but structurally speaking both companies stand on different levels of the luxury pyramid. The closest analogs to Hermès are probably Richemont and Ferrari — both lack entry-level items. Hermès offers things such as scarves, the signature accessories of Queen Elizabeth. While still difficult to craft and expensive, these are relatively accessible for first-time buyers. Products that are scarcer, such as bags, require lasting relationships with Hermès staff before they are offered to you.

We think it is the right balance. Revenue engines are concentrated on very expensive leather goods, but supported by a broad universe of objects that serve as entry points and build client loyalty. There needs to be a link between the opportunity to know the brand and unaffordable exclusivity. The reference point — as a warning — is what Swatch does: designing product lines to generate cultural hype and artificial scarcity. We want to see none of that. It breeds a consumerist culture.

We think of Hermès as a champion of durability. New products emerge only when the time is right, in craftsmanship houses that are largely still based in France — 63 out of 79. No adventures are taken with acquisitions of other brands. Focus is crystal clear on its own centuries-long heritage, which is impossible for competitors to recreate. It narrows the distribution of possibilities for Hermès's future to essentially two scenarios: it will either persist the same way for decades, or it will erode into complacency once a new generation of leadership takes over.

Visa is a company that has a coalition of allies willing to fight for it. It has embedded itself into an ecosystem where all the key actors — banks, consumers, regulators, merchants — are part of the network. There is a balance of power within that network that resists any single source of pressure.

Both of us were educated as political scientists. Understanding how coalitions work is what gave us the instinct to buy Visa. It was one of our first picks, back in 2020, when our style was only starting to take shape. Valuation at the time was around $400 billion. It was expensive — we impatiently paid 35 times trailing earnings. We knew Warren Buffett held the stock inside Berkshire's portfolio, and it was described by many as an exceptional business. As we dug deeper into Visa's reason for being, our confidence only grew. We have added every time valuation dipped since, including at the start of 2026.

Throughout most of human history, people traded using physical items — gold, salt, coin. Physical money is of limited practical use nowadays: difficult to store and hard to transport. Our societies need long-distance, intangible transactions, and trusting that they happened without physical proof requires a mediator — a middleman who tracks exactly how much money everyone has, and how it moves with every transaction. Visa is one of the best solutions anyone has built.

The original problem Visa was supposed to solve was credit access in the 1950s. Bank of America (BofA) knew that middle-class consumers had no easy way to get revolving credit for everyday purchases, so it launched a credit card. The launch was not flawless — fraud was rampant — but by 1966 profitability was obvious, and BofA wanted to scale. The trouble was that it was a California-only bank, unable to open branches elsewhere. The only way to grow was to license the system to other banks. When BofA once declined a license to Marine Midland Bank in New York, that outcast convened rival banks and built its own alternative — what we now know as Mastercard. Seeing a serious rival forming, BofA gave up direct control of its own project and converted it into a cooperative of issuer banks in 1976. Visa was born.

By the 1980s it was clear both cooperatives were solving problems well beyond the original credit plan. A merchant can settle a transaction and, at the same time, gain access to hundreds of millions of consumers who tend to buy more. Those paying in cash feel more pain spending than those who simply tap a card. Try telling your customers they cannot pay by card! Cardholders keep their cards for the credit, the global acceptance, the fraud protection, the rewards. Governments get transaction visibility that helps tax compliance. Banks get interchange revenue and interest on credit — most of the transaction fee, in fact as much as 1–2%, goes to the issuing bank. Visa itself earns only around 0.1–0.15%. When regulators or competitors take on Visa, they quickly discover they are really fighting a broad coalition of allies.

If the relationships within coalition described above substantially shift, we will be ready to sell. So far, everyone participating in the network has been sharing in the value and has an interest in its survival. If Visa were gone, cardholders would lose global acceptance and fraud protection, merchants would lose consumers who experience less pain while buying, authorities would lose tax compliance visibility, and JPMorgan, HSBC, Deutsche Bank, and plenty of other banks would lose interchange revenue. That is a very diverse list of parties with reason to keep Visa standing behind their backs. The ecosystem lets Visa absorb a great deal of pressure: when regulators target transaction fees, it's the banks who bear most of the pain, since they take the largest cut; when cardholders spend more than they can afford, it's the banks who carry the risk, since they issue the credit. As a result, Visa's own margins, returns on assets, and free cash flow stay remarkably high.

Since Visa converted from a bank-owned cooperative into a public company in 2008, an explosion of fintech and cryptocurrencies started. Armies of start-ups keep trying to disrupt the financial sector, and some do find their niches — but Visa still stands, growing revenue consistently by slightly more than 10% a year. It remains the dominant bankcard network, commanding roughly half of total card payments outside China. China is the one place the Visa–Mastercard duopoly was successfully sidelined, by UnionPay, later joined by Alipay and WeChat Pay. Massive centralised effort required simultaneously securing banks, merchants, customers, regulators, and reliable technology all at once. It took decades, and it worked inside China only.

Worldwide, Visa's payment network keeps setting the standard for rules, technology, and security — handling authorization, fraud prevention, settlement, tokenization, data analytics, and risk management. The more widely it is used, the stronger the network effect becomes. It substantially narrows the range of outcomes we need to consider when estimating how Visa develops over decades to come. We are confident it can grow its earnings per share by at least 50% within five years. In fact, it did even better in the last five years when earnings per share nearly doubled. Therefore, we trigger a buy every time valuation approaches 25 times trailing earnings. Yes, its revenue depends on global payment volume, which can slow in a recession. Yes, its huge margins attract challengers, but management is aware of that. It keeps strategically acquiring potential disruptors before they become threats — Earthport in 2019, Tink and Currencycloud in 2021. We think Visa will endure for a very long time.

Vertex Pharmaceuticals is one of the greatest rare-disease strongholds ever built in modern biotechnology. It has developed effective modulator therapies that correct the malfunctioning CFTR protein in patients with cystic fibrosis. These therapies prevent the buildup of thick, sticky mucus that would otherwise clog airways in the lungs and block ducts in the digestive system. While Vertex doesn't cure the genetic mutation itself, which causes the condition, it does address the underlying cause of the disease, offering a remarkable improvement for patients.

Impressed by the financial characteristics of the virtual monopoly described above, we bought our first shares in 2021 at around 20 times trailing earnings, valuing the entire company at more than $50 billion. We added more in 2023 and 2024 at somewhat higher earnings multiples, but at prices we still considered attractive dips. Our position will be increased if the market provides another attractive opportunity, or trimmed if the valuation rises beyond what we believe is reasonable.

There are currently five Vertex CFTR modulator therapies: Alyftrek, Trikafta, Symdeko, Orkambi and Kalydeco. Vertex remains the only company to have successfully brought CFTR modulators to market. These therapies work only in people whose mutations allow their CFTR protein to respond to modulation, although Vertex has progressively expanded the eligible population. Following recent label expansions, its modulators can now treat approximately 95% of people living with cystic fibrosis in its core markets.

Vertex has kept a sustained focus on CFTR biology. It has made iterative improvements in its therapeutic approaches, providing increasingly effective treatments. These are broadly reimbursed and have wide geographical access well beyond the U.S. — more than 60 countries across six continents. CFTR modulators are protected by patents for at least another decade, and there is no meaningful cure currently visible on the horizon. A successful genetic or other curative therapy could eventually make CFTR modulators obsolete — and, hopefully for patients, one day it will — but we don't see that day coming soon.

For now, almost all of Vertex's roughly $12 billion in annual revenue comes from its CFTR modulators. However, the first meaningful sources of diversification are becoming visible following recent approvals.

Journavx (suzetrigine) is a non-opioid medicine approved for the treatment of moderate-to-severe acute pain.

Casgevy (exagamglogene autotemcel), co-developed with CRISPR Therapeutics, is a one-time gene-editing therapy for severe sickle cell disease and transfusion-dependent beta thalassemia.

More new medicines could emerge from the pipeline as management continues directing cash flows generated by the CF franchise toward developing new sources of revenue. Among the different programs, we are particularly interested in the following:

Probably the biggest near-term new contributor to the revenue line could be povetacicept, initially targeting IgA nephropathy. Vertex acquired Alpine Immune Sciences, and with it povetacicept, for approximately $4.9 billion in 2024. The FDA has accepted the application for accelerated approval and set November 30, 2026 as its target action date. If successful, we believe povetacicept has the potential to generate several billion dollars in annual revenue, particularly if it succeeds beyond IgA nephropathy. Vertex, however, is far from the only player competing in this field.

Inaxaplin, being developed for APOL1-mediated kidney disease, could eventually become another multibillion-dollar opportunity. Much more visibility should come from the AMPLITUDE trial, with interim data expected in early 2027.

And then, if we allow ourselves to be very optimistic, we would closely watch the development of zimislecel, which could potentially provide a functional cure for type 1 diabetes. We are not assigning much value to it for now, as it remains a long-term project surrounded by substantial clinical, manufacturing and commercial uncertainty.

We think an attractive price for Vertex stands somewhere around $390 per share. At that price, our model requires earnings per share to grow approximately 1.88 times over five years to deliver an average market return, even if Vertex's valuation subsequently falls to around 17 times trailing earnings — close to the lower end of its historical range. We consider this a reasonable projection and broadly consistent with current expectations of analysts.

Interestingly, this is also the valuation at which CEO Reshma Kewalramani bought 10,000 shares on the open market during the August 2025 slump, investing around $3.9 million. She had also bought 10,000 shares in 2021 — the same year we opened our position — when Vertex was going through another period of investor pessimism. We would therefore feel comfortable adding again if an opportunity emerges around $390. We would, however, consider trimming or even exiting if the share price approaches $600 without a corresponding improvement in earnings expectations. At that price, our model would require Vertex's earnings five years from now to exceed current analyst expectations by roughly 50% to generate the return we require.

Nvidia is the most profitable company ever built on a single technological wave. During the Industrial Revolution, Standard Oil dominated oil. In the twentieth century, AT&T dominated telecommunications. Neither achieved what Nvidia has in the age of Artificial Intelligence.

We started buying in May 2022, during the tech selloff. At the time, even after a big drop, the whole company was valued at more than three hundred billion dollars. We knew that Nvidia's chips — the "brain" for electronic devices — powered gaming and increasingly machine learning, as both activities were hungry for computing. But we had no idea what forces ChatGPT would later in November unleash, eventually pushing the price of Nvidia to five trillion.

We knew the company was exceptional: Glassdoor ranked it 4.7 out of 5 - best in its class - and it still holds a turnover rate of 3.7%, which means it is a magnet for talents. The genius of Jensen Huang has created The Nvidia Way. It is a mixture of unprecedent leadership vision combined with self-regulation regime. It allows ideas to emerge and capitalizes on the best of those moving at the speed of light, if one is to speak in Jensenisms.

In a very short time Nvidia has captured — or created from scratch — approximately 80–90% of the AI accelerator market by revenue. These are the components designed to process AI. They work best alongside all the necessary infrastructure created by Nvidia, including CUDA, the programming system that enables developers to run code directly on Nvidia units.

The rest is history. Revenue has grown eight times during the period we held. Profit margin and return on assets jumped respectively from 37% to 60% and from 22% to 58%.

We are under no illusion — Nvidia's exceptional position is under siege. Microsoft, Google, Amazon, Meta and Oracle account for 40–50% of all revenue. The very same companies are already trying to deploy their own alternatives, alongside promising start-ups. China is building its own ecosystem.

Are we worried? We know that a masterclass on disruption by Clayton Christensen The Innovator's Dilemma is Jensen's favorite book! He saw the wave of AI coming well before the ChatGPT moment. He bought Mellanox Technologies in 2019 and Cumulus Networks in 2020 to expand the Nvidia ecosystem. He can do it again when the next wave comes.

Jensen still owns roughly 3.5–3.8% of the company. Billions of his wealth parked alongside ours. He deserved trust through integrity and resilience. Nvidia does not waste energy on tax optimisation strategies, but pays "without even looking at the bill," according to Jensen. He consistently stresses the importance of "suffering" while growing. His charisma was tested rallying political and business leaders for the AI cause.

Jensen knows Nvidia is in "the infinite game" and designed it accordingly. Behind his back stands a university-like learning hub with exceptional talent and culture. This is the true moat of Nvidia. If we start seeing meaningful defections who create their own start-ups, or if turnover rises while Glassdoor rankings dip, or Jensen himself leaves abruptly — we will be ready to exit. We haven't seen those signs yet.

We are, however, closely watching valuation. Nvidia did experience its stock falling by around 80% in the past — twice. We think it is not sustainable when key actors in the AI race throw ten dollars of capital expenditure only to earn one, as is the case now. It reminds us of Railway Mania, when companies were building competing tracks while in fact one logical network was enough. However, the Rubicon has been crossed and there is no return. We still need faster transportation. We still need faster internet. And we will need faster computing for the age of AI. Nvidia will know how to produce it.

Thinking five years ahead, it is not delusional to expect Nvidia to increase its revenue, earnings and free cash flow between two to three times compared to what it is now. These expectations are consistent with the current valuation of five trillion. However, a market capitalisation of six trillion would trigger us to trim the position — because it would require more things to go right for Nvidia while we live in a complex environment.

First, our understanding of the state of world affairs can envision a scenario where China starts conflict over Taiwan or simply blockades it. It would be particularly damaging to Nvidia in the short term, because it is fully reliant on TSMC foundries in Taiwan. Nvidia could not get their chips. Period.

Second, the size of Nvidia will at some point start limiting it. At 42 thousand employees it already exceeds the size of many ancient Greek city-states! It will become difficult to sustain its exceptional culture. At the same time, it will draw criticism over power. Feudalism concentrated power in landowners. Capitalism switched the balance towards owners of means of production. We see possible switch towards "digital land" ownership, techno feudalism, in the words of Yanis Varoufakis.

Finally, there are energy concerns. AI era needs massive amounts of energy, unsustainable now. We do believe the benefits of AI will outweigh the risks. For all of us it will open opportunities never seen before. However, we agree with Pope Leo and J. R. R. Tolkien. The civilization of love will not arise from a single or spectacular gesture, but from the sum of small and steadfast acts of fidelity that serve as a bulwark against dehumanization.

ASML is a company that sells the most precise manufacturing machine ever built. Advanced microchips which run smartphones, computers, and AI would be impossible without ASML, just as it would be impossible to move goods through Suez without a canal.

We first learned about ASML from a close relative at the beginning of our investment journey in 2020. We did not commit, however, because a valuation of around 50 times trailing earnings simply seemed too demanding for us. Since then, the share price has increased roughly sixfold. We finally initiated our position in June 2025 at around 33 times earnings, valuing the whole company at approximately €245 billion. At the time of writing, our investment has doubled. As you can imagine, we are happy that we finally pulled the trigger. What we are not happy about is that we did not pull it much earlier, and with a machine gun rather than a single shot.

Here comes a note of caution to ourselves and to those reading. We think we understand what ASML does, but that is probably an illusion. The science behind its machines is so close to the frontier of human knowledge that much of it is beyond ordinary comprehension. To give just one example, at one stage of manufacturing advanced chips, tiny droplets of tin are struck by a laser that generates temperatures roughly forty times hotter than the surface of the Sun. What follows below is our honest attempt to explain what we think we understand. But to reach that satisfying "click" moment, one probably needs to spend quite some time studying physics.

Microchips are manufactured using lithography. It is the process of transferring incredibly small and complex patterns onto silicon wafers using light. ASML has built a monopoly in the most advanced form of this technology: extreme ultraviolet (EUV) lithography. To produce EUV light, tiny droplets of tin are struck by a powerful laser, creating temperatures roughly forty times hotter than the surface of the Sun. The light produced is so difficult to control that it is absorbed by almost every material. Ordinary lenses are useless. ASML has created an optical labyrinth of ultra-precise multilayer mirrors, polished to perfection, to guide the light exactly where it is needed. The entire machine contains more than 100,000 parts. It is shipped in around 250 containers and takes hundreds of engineers about six months to assemble and calibrate. Each system costs well over €200 million. The next generation, called High-NA EUV, costs almost twice as much, although it still must prove its commercial value. ASML is already working on the generation that will follow, expected around 2032.

Remarkably, the company manufactures only about 15% of the components for the extraordinary machine described above. The remaining parts come from roughly a thousand suppliers, many of them irreplaceable. TRUMPF is the sole supplier of the powerful laser, and ASML forged a strategic partnership with it. Carl Zeiss SMT is the only supplier of the optical system, and ASML owns a 24.9% stake in the business. Cymer — acquired by ASML entirely — is the only company capable of producing the laser-produced plasma light source. In short, ASML has built an ecosystem based on intimate technical cooperation and decades of accumulated knowledge. A competitor would need to recreate hundreds of engineering relationships or build everything from scratch in-house. In our view, it's more like a mission to attack the Star Wars Death Star.

But even the Death Star has its weak points. Perhaps one day this beautiful and extraordinarily sophisticated machine will remain beautiful and sophisticated, yet no longer relevant. A different computing paradigm could emerge, reducing the need for ever more advanced lithography. History reminds us that technologies we once thought indispensable were eventually displaced — from film cameras to magnetic tapes. The more immediate threat we are watching, however, comes from China's determination to build its own advanced lithography industry, as ASML is currently prohibited from shipping its EUV systems there. The effort has already involved industrial espionage and aggressive recruitment of ASML engineers. Perhaps the best-known case is that of Zongchang Yu, former CEO of XTAL Inc. In 2019, XTAL filed for bankruptcy after a US court ordered it to pay ASML $845 million in damages for trade secret theft. Zongchang Yu — under US arrest warrant — moved to China, where his company Dongfang received "Little Giant" status, given to strategically important technology companies that qualify for state support. We follow every report on China's progress. Our current understanding is that China still lacks what EUV lithography needs, but gradual catch-up is on the cards.

A key part of this race will be ASML's culture: willingness to make long-term bets and perseverance with efforts others would abandon. ASML has never been a founder-led story. It was built over four decades by thousands of engineers, suppliers, and partners, absorbing and outlasting dozens of exceptional individual careers, including the one of the current man in charge. Christophe Fouquet has been with ASML since 2008, became an executive in 2018, and CEO in 2024. We love his plan to simplify the organization, which grew from around 26,000 employees in 2020 to more than 43,000 by 2025. Rapid expansion brought additional layers of management, making the company less agile. There are around 4,500 managers overseeing some 16,000 experts, and Fouquet's effort, as communicated by media, is to eliminate 3,000 — laying off roughly half of that number and transitioning the other half to engineering roles. According to Fouquet, engineers currently spend only half of their time on actual engineering, while up to a third of their time goes to meetings. The target is 80% engineering work.

We believe the culture described above is the strongest predictor that ASML will still be an exceptional company decades from now. From time to time the market offers the opportunity to buy a small piece of this giant at a reasonable valuation. Based on our own assessment, we become particularly interested when the valuation falls below roughly 30 times trailing earnings. Around 45 times earnings can still be justified, but much beyond that requires believing the company will grow faster than most analysts currently expect. The semiconductor industry is cyclical, and the next peak may not be far away. We are happy to wait for more favorable price levels to increase our position. At the same time, we are comfortable keeping what we have in case we are wrong. We will be happy to be proven wrong.

Microsoft is powered by two enormous, roughly comparable engines: a high-growth, increasingly capital-hungry cloud infrastructure business, and a high-margin, capital-light subscription software business. Behind them sits a smaller collection of businesses — Windows, devices, gaming, search. It is the oldest company in our portfolio by years since its IPO, and yet it sits at the forefront of the AI infrastructure build-out that is carrying the promise of transforming the world economy.

We have owned Microsoft since the beginning of our investment journey in 2020. Our first entry came at a valuation of around $1.5 trillion and 37 times trailing earnings. We have added several times since then, admittedly at multiples we would probably not accept today, as our investment style has matured. We still think, however, that Microsoft is a juggernaut worth keeping in our portfolio.

True, Windows is not what it used to be. Bing internet search seems destined to remain a distant second to Google's dominance. Gaming and Xbox devices face intense competition from Sony and Nintendo. These businesses are grouped together under "More Personal Computing," and account for around 16% of Microsoft's revenue. This segment is not the reason we own Microsoft — even if our eldest child loves Minecraft, which belongs to the company.

"Productivity and Business Processes," containing brands such as Microsoft 365, LinkedIn and Dynamics, is a different story. These services are essential to many businesses around the world. Once an organization builds its workflows around Microsoft software, switching becomes difficult and costly. The business is sticky. LinkedIn, meanwhile, benefits from powerful network effects: the more professionals who use it, the more valuable it becomes to other professionals, recruiters and businesses, reinforcing its position over competing platforms. Together, Productivity and Business Processes generated around 42% of Microsoft's revenue in 2026 and more than half of its operating profit. These businesses are the reason we acquired our shares.

Now comes the part we neither like nor dislike, because we don't yet fully understand where its economics are heading. We remain comfortable watching and learning. It is the segment called "Intelligent Cloud," with the Azure computing platform at its heart. Azure has made Microsoft one of the world's three dominant hyperscalers, alongside Amazon and Google — meaning the company operates enormous data centers full of servers, networking equipment and increasingly specialized AI hardware, which other companies can tap into on demand. That has become particularly valuable as demand for AI services has exploded. Intelligent Cloud now generates roughly the same revenue as Productivity and Business Processes. But it is growing much faster, and appears increasingly likely to become Microsoft's largest engine. The price of that growth is colossal capital expenditure to build the infrastructure described above.

Microsoft itself acknowledges that AI infrastructure is being built ahead of fully developed revenue streams. We have no idea how these efforts will end. We don't yet understand what the economics of such an AI empire may look like. Can computing services become sticky, or will computing resemble a commodity — a sort of digital oil, as we suspect may be the case? But even for drilling oil, efficient scale still matters. Does getting bigger reduce competition, because scale lets hyperscalers offer computing at a lower cost per unit? If Azure generates extraordinary profits for the next twenty years, today's enormous expenditures could prove necessary. If computing becomes commoditized, pricing could fall and customers may grow indifferent between Azure, AWS and Google Cloud. In that case, Microsoft may have redirected an enormous amount of its beautiful software cash flow into a much lower-return business.

Satya Nadella, Microsoft's CEO since 2014, clearly believes the company needs to make an enormous bet on AI infrastructure. He personally owns Microsoft shares worth roughly $350 million at recent market prices, giving him meaningful exposure to the outcome. Insider ownership across the executive team as a whole, however, remains tiny relative to the company. This is not an owner-operator situation comparable to Nvidia under Jensen Huang.

Capacity built today will or will not monetize over many years to come. If capital expenditures growth starts slowing while Azure growth holds, that's the best-case scenario. Somewhere around three to five years from now, we will need to see evidence that the returns on all this additional invested capital make economic sense. Each additional dollar of capital should eventually produce sufficiently attractive operating profit. If the amount of incremental capital required for each dollar of incremental profit keeps rising materially, that would become an exit signal for us. Once capital expenditures will turn into deprecation line on financial statement, we will need to see if cloud segment operating margin stabilizes and recovers toward its historical levels. A persistently depressed margin despite continued growth would be another warning sign. In the nearer term, if Azure growth were to slow materially – for example toward 25% – while AWS and Google Cloud consistently grew faster, we would treat that as an early indication that Microsoft's competitive position may be weakening.

Microsoft's valuation moved into the low-20s on trailing earnings during the summer of 2026, one of its more attractive valuations in recent decade. Our modeling suggests that somewhere around $370 per share, we don't need heroic assumptions to justify an investment. Under our assumptions, Microsoft would need to grow earnings per share by roughly 1.45 times over five years to generate the return we require – considerably less growth than analysts expect. At around that price, we would be interested in adding. If higher prices persist, we will wait, because AI infrastructure build-out meant free cash flow has made little progress since 2022. This reduces our confidence in making projections substantially more optimistic than the assumptions already described above.