Digital Diversification

Amid the current hype surrounding artificial intelligence (AI), the market is rapidly creating narratives of winners and losers. This is leading to extreme market volatility in technology stocks. Yet the question of which companies can develop solid long-term business models from technology, and which will come under pressure, is far from being conclusively answered. To benefit from these developments without taking on excessive risk, we advocate a diversified approach.

Technology is a broad field. In theory, the Global Industry Classification Standard (GICS) of index provider MSCI clearly divides companies into eleven sectors, and information technology is one of them. In practice, however, it is far less clear-cut: for instance, online retailer Amazon is classified under cyclical consumer goods, even though the company is also one of the leading service providers in the cloud business. Alphabet, which is not just Google’s parent company, and Meta, which does not focus solely on social networks, are also classified as communication services, even though their business model is primarily based on the sale of advertising. Because technology permeates many industries, we prefer to view current developments against the backdrop of the entire digital value chain.

This includes software used by consumers or businesses, and that which underpins so-called financial information firms. Added to this is the hardware installed in and around data centres. All four sectors are currently influenced primarily by developments in AI and dominated by certain market narratives.

Consumer software: can there be just one?

Whether travel portals, streaming services, ride-hailing platforms or social networks: they all function best with as many users as possible due to the network effect. With this type of software from everyday consumer life, our credo has therefore been that the big players are getting bigger. Currently, however, the question arises as to whether the rise of AI chatbots (also known as assistants or agents) such as ChatGPT is changing this.

One argument in favour of this is that, ultimately, all companies are vying for customers’ attention, yet customers only have 24 hours in a day. An increase in usage for one app therefore means a decline for another. Consequently, the narrative has taken hold that ChatGPT and search engines cannot coexist; instead, the superior option will prevail among customers. In the case of Meta and Netflix, however, the market seemed rather unconcerned from our perspective. Yet ChatGPT stands out for its creativity, which could be utilised in the future to expand into areas such as social media and video production.

Which existing companies may come under pressure from chatbots in the future therefore depends heavily on the specifics of their business models and the particular capabilities of the chatbots. If, for example, it is confirmed that chatbots are particularly good at tracking down specific and rare information, this would put pressure on certain providers, even if they have been the market leaders up to now. So rather than following the broad narratives to which “the market” sometimes tends, business models, risks of disruption and AI developments must be closely monitored. It is unwise to categorise companies as AI winners or losers at this stage.

Enterprise software: do they need more or less?

The enterprise software sector has been characterised by a high degree of predictability for many years, as switching providers usually involves significant effort for corporate customers. Now, however, there is also a debate here as to whether AI will lead to disruption and how providers’ pricing power might change. Three aspects are being discussed in particular.

Firstly, enterprise software is becoming increasingly effective due to the addition of AI features, meaning that business customers could reduce their workforce in future and consequently have less need for software licences for their smaller workforce. Countering this is the argument that employees become more productive through AI and that the resulting efficiency gains are shared between the software provider and the customer. We tend to side with this view for some companies.

Secondly, sceptics argue that enterprise software providers offering new AI features will be forced to adopt a different pricing policy. Historically, they have earned too much because they have insisted on bundled offers that simply combine different programmes, and on subscriptions with the longest possible terms. This could now become a disadvantage. Some companies do argue that the addition of AI features would justify price increases and that usage-based components could be incorporated, meaning that heavy users might be asked to pay more in future. However, there is currently no evidence that customers of established software are willing to pay for additional AI features. We also believe that shifting pricing models away from subscription models towards a usage-based payment method is likely to present challenges.

Thirdly, it is speculated that software providers’ businesses could be at risk if tasks are no longer carried out by humans but by AI in the future. However, we still consider this scenario to be premature for most sectors. Start-ups that build agents, for instance, rely on data. And we have observed for some time that the market leaders among software providers are blocking access to this data. In this way, providers manage to buy themselves time to integrate agent functions into their own products. Services that rely heavily on up-to-date data could therefore feel less pressure from AI.

Financial information software: will data providers become redundant?

Credit card providers, stock exchange operators, and “data and analytics” firms such as S&P Global all not only benefit from the network effect, but also monetise data directly and indirectly.

Credit card providers, for instance, process billions of transactions daily. This generates granular data – who bought what, where and when, and via which device – which is monetised, for example, by selling risk tools for the early detection of fraudulent anomalies. Stock exchange operators, too, sell not only trading data but also analytical tools for risk management.

Will AI make such business more difficult in the future? This concern has been haunting the market since at least mid-September 2025. At that time, a quarterly result from US financial data company FactSet triggered a sell-off in financial information stocks. The management there had announced that it would be investing and that margins would consequently decline slightly in the coming year, without offering any prospect of a corresponding increase in revenue. One broker subsequently wrote: “Even if this does not amount to a complete disruption by AI, the costs of doing business are clearly rising.”

This may be true for companies in this sector that hold only a relatively small proportion of their own data. However, in our view, this certainly does not apply to the entire industry. We therefore delve into the details for each company by asking further questions, such as: Is the data difficult to replicate? Is it embedded in a workflow? Is it enriched by company input? Is it protected by regulation? Does it serve as a mark of legitimacy and credibility for customers? Where the answer is yes, we often see potential.

Hardware: who will prevail?

The fourth and final part of our digital value chain – hardware – can in turn be divided into three areas:

1. chips installed in a data centre;
2. the data centre itself; and
3. everything needed to keep the data centre running.

All these areas have experienced a structural boost in recent years due to demand for AI. Nevertheless, we also see risks.

In the chip sector, these were evident to Intel shareholders by mid-September 2025, as the company has lost massive market share. Even in the semiconductor sector, the rising tide does not lift all boats. However, there are very few boats worth watching. Among chip designers, there are essentially only three significant companies: Nvidia, Broadcom and AMD; in design software, there are two: Cadence and Synopsys; among chip manufacturers, there are also two: TSMC and Samsung; and among equipment suppliers, there are five: ASML, Applied Materials, Lam Research, KLA and Tokyo Electron.

This oligopolistic nature also applies to the second sector: cloud infrastructure. The three providers – AWS (Amazon), Azure (Microsoft) and GCP (Alphabet) – currently account for a market share of more than 90 per cent. However, their largest supplier, Nvidia, is keen to reduce their bargaining power and is prioritising the supply of its high-performance chips to new providers (“neoclouds”) such as Oracle, CoreWeave and Nebius. This is intended to put pressure on the three dominant providers.

Oracle’s shareholders, however, have felt the sting of such attempts. Whilst the share price had soared following the announcement of the partnership with Nvidia from September 2025, it subsequently lost around 45 per cent at times – the market is simply weighing up exactly how profitable each planned data centre is.

Things are calmer in the third hardware sector – the “enablers”. Companies such as connector manufacturer Amphenol, Siemens or Schneider Electric ensure that power reaches the data centre reliably and is distributed there. These companies are just as indifferent to which consumer, enterprise or financial information software prevails as they are to the question of which chip is used. Furthermore, the AI trend, together with automation, accounts for only around 50 per cent of revenue. These companies therefore benefit from the current demand for AI data centres without being dependent on it.

For diversified, balanced investments

The digital value chain thus spans many industries, each with its own specific advantages and disadvantages in terms of AI progress. Whilst common market narratives often drive price movements, we do not believe they are always justified. For some enterprise software firms, for example, we consider expectations of disruption to be exaggerated – on the one hand, changes in this sector do not generally happen very quickly, as demonstrated by the migration of corporate data to the cloud, which has been ongoing for many years; on the other hand, AI can certainly also present an opportunity for market leaders.

We therefore focus on carefully analysing companies and investing in a balanced and diversified manner across the value chain. In this way, we aim to capitalise on the opportunities presented by these extraordinary developments without taking on excessive risks.