European Equities & AI: Key Insights from the BNP CEO Conference 2026

We have been in Paris for the annual BNP CEO conference. It was illuminating and we reveal some core takeaways below.

Key takeaways

- There was little sign of the Middle East conflict weighing on sentiment, and we note that in the follow ups all commentary was bucketed as either “positive” or “neutral”. This was before we had the recent peace deal announcement.

- We also checked in with several companies we own.  We found that:

• The push towards deeper Capital Markets integration should help our holding in Deutsche Boerse;
• Margin improvement continues to outpace management’s expectations at Air Liquide and;
• Our healthcare stocks, like Novartis gave updates on a few exciting pipeline assets that the market seems to currently overlook.

- On AI we would note that:

• The European market has plenty of AI plays and investors are feverishly pouring over them;
• Many companies were asked about AI and talked about how they were trying to use it to help their businesses.  We haven’t yet seen evidence that European companies have materially accelerated revenues or margin improvement because of utilizing AI, but they are admittedly in the early stages.
• AI is being monetised to a certain degree by consumers who are suffering from Enterprise IT infrastructure moving slowly;
• European software management teams are trying to develop their own AI agents (autonomous software that can understand user intent) but at this stage appear materially behind more advanced and better funded US startups.

Company updates

It was interesting to hear from the CEO at Deutsche Boerse, Stephan Leithner, that regulation is continuing to help undergird the jewel in their crown, both their clearing business, Eurex, and the custody business Clearstream. 

There are two key reasons why – one is The European Market Infrastructure Regulation (EMIR 3.0), which mandates market participants to maintain active accounts with EU-based central counterparties (CCPs) and should help Eurex, and the other is the latent potential of the drive for Capital Markets Union to create deeper liquidity (and therefore help their custody business Clearstream).

Currently, the market appears to only price their 8% top line target but not anything if these two directives were to truly deepen European Capital Markets.  In the meantime, we don’t believe it likely anyone could remove the clearing or custody business from their incredibly entrenched positions in financial markets, and we like the dividend and buybacks returned to us as shareholders.

The industrial gases industry is well known to our funds as we have previously owned Linde following the Linde/Praxair deal.  After that thesis played out, we believe there is now greater upside in Air Liquide from margin improvement.  Management remains confident that they can hit their 200bp margin expansion target over 2026/27 and may stay close to the 100 bp/yr pace from 2028, ie higher than consensus expectations. Pricing is tracking ahead of early inflation effects, and we were pleased to hear there had been no major supply-chain disruptions reported at the customer level.  

Whilst some may see Air Liquide as exposed to slower growth markets like Europe’s industrial basin, we like their more measured approach to capital deployment, mainly incremental projects rather than the huge one off projects that US competitor Air Products has tried in the past (we are sceptical in general of large capex all at once and the returns it can generate).  We also appreciate Air Liquide’s growing exposure to semi-conductors (you need a lot of inert gas to produce wafers) and its nascent exposure to space and healthcare, a more defensive and structurally growing market with only a handful of operators and rational pricing. 

If the true value of any business is the discounted value of its future cashflows back to the present day, then the debate on pharma companies is always about the duration of those cashflows because products are always rolling off patent (where cashflows tend to fall precipitously).  It was therefore pleasing to hear Vas Narasimhan at Novartis talk to several big read outs this year and another 5 next years which could be multibillion-dollar indications.  The phrase used, “multiple shots at goal”, is one we appreciate as generalist investors that wouldn’t be able to take a particularly strong view on any one molecule per se. Post the Roche stake sale a few years ago, buybacks were ramped up and we were pleased to see another $10bn tranche announced in July 2025, and a heavy indication from Vas in Paris of more to come.

European market and AI

One interesting observation is the number of people who think that all the AI opportunities are in America and Asia rather than in Europe.  The conference demonstrated that not to be quite true. 

The truth is that the most well attended meetings were companies that either directly or indirectly could benefit from the insatiable demand for AI.   

Those include companies like Infineon, who are producing power semiconductors that are crucial for data centres and managing power voltages, and others like Schneider Electric that are benefitting from the massive power demand.  If you also consider the number of broad industrials that might benefit from automation, power and outright semiconductors like Infineon, ST Micro and of course ASML, then there is a lot of market cap in Europe that could be labelled “AI beneficiary”. 

A cautionary tale for anyone who looks at GICS sectors to try and gauge underlying exposure for stock market structures.

Monetising the spend

The question then is what to do with all this market cap and whether it is investable.  And here we found out something else at this conference – that more peers than we thought are paying to use “consumer” AI, but that this was really “enterprise” spend because their employers wouldn’t give them corporate access.

Given how slow enterprises tend to be to adopt new technologies and the risks around data and protection it is unclear how long this will last for.  For now, though, peers are using AI to augment personal workflows so they can perform research and various other functions outside of their corporate environment.

And so, whilst we were surprised at how many people were paying for AI chatbots, this wasn’t “true” consumer spend, it was niche “enterprise” overflow spend, and the data does confirm this.  The number of monetised users on ChatGPT is only 5% and perhaps as low as 1-2% for Gemini, which suggests the investment community’s use of AI is not representative of wider consumers. The perils of being in an echo chamber!

This has profound implications for the current “demand for compute is through the roof” refrain. When a product is subsidised or provided free of charge, conventional supply-demand dynamics can become distorted because users are insulated from the true cost of consumption.  We believe that we are still partly in this phase as the AI industry develops, and whilst it is true that certain enterprises are paying and some consumers, it is also clear these payors are in a minority when you weigh the putative revenues of Anthropic, OpenAI and Gemini vs. the monumental capex being spent.  There are indications, however, that revenues for these companies are ramping, but the monetisation debate remains uppermost in investors’ minds.

European software

We believe this was also the first BNP CEO conference where there were dual tracks – a public markets track and a private markets track.

Having these two side by side elucidated two key insights – that European software management teams don’t necessarily fully comprehend how serious their competition is to build AI tools to unlock value (agents) and secondly, that many trot out commentary around competitive advantages that may have been true a decade ago but are becoming less true.

In the private markets track, there were 11 companies and it became very clear that these were well funded businesses with billions of dollars of backing and many with a single mission - trying to disrupt the AI industry (mainly by building agents/AI tools).  Talking to their almost messianic dedication to this task, it became clear that with deep pockets and fantastic research teams (Silicon Valley and multimillion-dollar AI coding contracts), they may prove to be more disruptive to some of the existing European software companies than many expect.

To explain briefly what these AI startups (and software companies) are trying to do when it comes to AI agents, a brief elaboration is perhaps helpful.  Think of an AI agent as a piece of software that can understand a goal and will make decisions to achieve that goal with limited human intervention.  For example, in the old world of Dassault Systemes, engineers would design a component, then run the simulation, then analyse the manufacturing constraints etc.  In the new world with an agent, the engineer tells the AI agent “reduce the weight of this aircraft bracket by x% while maintaining strength and ensuring it can be manufactured using similar processes.”  The AI agent will then go away and review the CAD model, identify weight reduction, generate multiple different designs and run them all through the simulation software.

The point is that for every software company to develop these “agents” is not only very costly but also not their core competency.  Lots of private markets from Silicon Valley have better AI coders and bigger budgets - including the soon to be public ones. 

It is also a massive corporate distraction.

Just think of any business you work for building an in-house AI lab.  Even for software companies this is complicated when not your core competency.  Being good at holding financial information (LSEG), transactional data (Rightmove) or engineering and design info (Dassault Systemes), does not make you experts in building AI agents.

It's (probably) a better approach that these companies try and develop data sets that are not only proprietary but also useful and maintain proximity to their customers rather than to try and compete with deep pocketed Silicon Valley AI labs.  Far better to adopt the Apple approach to AI and use anybody else's 3rd party agents within your “secure” walled garden to deliver your information in a more useful way to customers.  Especially as agents might turn out like LLMs to be relatively commoditised.

A lot of European software names are also still parroting lines that were true about their competitive advantages a decade ago but are becoming less true. 

Legacy codes become easier to move with AI, data extraction easier via natural language interfaces, and the impetus to potentially lift and shift vast data sets greater if the upside to efficiency and usage is multiples rather than percentages (which it probably will be with properly utilised agents / AI).

Against all this is the fact that enterprises can be slow and cumbersome and rightly worry about data protection and integrity.  AI is clearly moving beyond linear chatbots to autonomous agents, and it is unlikely that every company building their own agents is going to be a workable model given the time, resources and capital required.

There is still definitely a place for European software firms, many of which are essentially data aggregators like LSEG and Rightmove, who are using clearly defined deterministic data / workflows. At the moment, however, it feels as though some European management teams are underestimating quite how large the threat is, and that really became apparent because of this dual public and private market track at this particular conference.

Conclusion

The Paris Conference 2026 was interesting on several counts. 

We got to check in with our companies who are broadly making progress in their aims (that as shareholders we got behind when we invested) and generated some new investment ideas we shall talk about in our future Investor Letters. 

AI remains alive and kicking in Europe but the debate on how much is currently monetised vs. what might be in the future, remains as fierce as ever.  It was also fascinating to witness the private markets track, crucial to allow reflection on some of the European software companies that may not always play to their strengths (creating useful and proprietary data sets) and are instead trying to compete with AI start-up labs where the risk of failure is high. 

For more information on our European Funds, visit the Fund Centre via the links below or contact Sanjana Morjaria Ramrakha.

Waverton European Capital Growth Fund F GBP Inc

Waverton European Dividend Growth Fund F GBP Inc