Region: Europe and the United States
Period: May & June 2026

Download PDF version here.

Over three weeks in late May and early June I attended four investor conferences across Europe and the US and held management meetings, attended fireside chats and conference presentations with more than 50 companies spanning technology and AI infrastructure, semiconductors and semiconductor capital equipment, software and internet, media and live entertainment, financials, electrical and power infrastructure, and public safety / defence technology.

The dominant topic of almost every technology meeting was artificial intelligence, but the conversation was more nuanced than a simple debate about which company is an AI winner. The key questions are now whether the pace of spending can be sustained, where the bottlenecks sit, which parts of the stack capture economic value, and which business models can convert demand into durable free cash flow. The most consistent signal was that the binding constraint is migrating down the stack, away from the GPU alone and towards power, memory, grid connection, advanced packaging and the physical plumbing of the data centre.

Outside AI, the trip reinforced the importance of dispersion. Consumer commentary was mixed but not uniformly weak (the famous K-shaped economy). Demand remains resilient where the value proposition is clear, the experience is distinctive or the use case is mission critical. By contrast, lower-quality or more discretionary businesses are being forced to rely more heavily on promotion, cost cutting or hope of macro relief.

The most actionable conclusion from the trip is therefore not simply to 'own AI', but to redirect the next stage of work towards the less crowded enablers where management commentary pointed to real bottlenecks and near-term earnings visibility. In addition to owned positions in Schneider Electric, Eaton Corp and Legrand SA, Prysmian and GE Vernova moved up the priority list because several meetings pointed to power availability, grid connection, cable supply, turbines and in-rack electrical architecture as more immediate constraints on data-centre growth (as much as accelerator demand itself).

Among existing tech holdings, ASML remains the highest-quality way to access the semiconductor layer after a reset in expectations, while ServiceNow deserves further debate on portfolio sizing. The ServiceNow meeting strengthened the case that its AI opportunity is not just a thematic overlay, and the company is monetising AI by embedding it in workflow automation across real enterprise processes, with customer outcomes measurable in productivity, service levels and cost savings. That makes AI more likely to deepen ServiceNow’s platform advantage than dilute it. Motorola Solutions was also a positive reiteration, with the meetings providing new evidence of an expanding total addressable market (TAM) in drone defence, resilient backlog, pricing power, mission-critical demand and rising AI-related content across command-centre and public-safety workflows.

Outside Technology, Media, and Telecommunications (TMT), Marex and Robinhood both caught our attention because both are platform businesses with clear value propositions and company-specific growth drivers. For example, Marex via clearing balances, agency and execution volumes, market-making opportunities and a broader institutional client base; Robinhood via net deposits, product velocity, AI-enabled operating leverage and optionality in prediction markets, tokenisation and international expansion.

The overall portfolio implication is to prioritise power infrastructure, mission-critical software and hardware, capital-markets and consumer-fintech infrastructure, and select monopoly-quality technology assets, while resisting the temptation to chase crowded AI-semiconductor multiple expansion.

Key Takeaways:

1. ‍AI capex is not slowing, but the market is getting more complicated. AI infrastructure spending remains robust and broadening. The recent sell-off in parts of the AI complex looked more like a sentiment, positioning and expectations reset than evidence of a demand cliff. The most attractive risk-adjusted exposure may still be the ‘picks-and-shovels’, i.e., semicap, advanced packaging, networking and data-centre infrastructure, where the payoff does not depend on predicting the ultimate model, GPU, ASIC or optical winner. But the valuation bar is higher and estimates are already rising, so the focus should be on backlog visibility, critical process steps and realistic next 2 years’ earnings power. The risk is less of a sudden collapse in demand and more of digestion, timing and the market capitalising on peak earnings too quickly.‍
2. The bottleneck is moving from chips to power and physical infrastructure. The most actionable message was that electricity, grid connection and in-rack power architecture are increasingly the binding constraints on data-centre growth. NVIDIA-native 800V HVDC racks are anchoring first delivery in 2H27, the power-equipment ecosystem is staging products through the coming months, and only a small minority of facilities appear ready for that architecture at launch. Hyperscaler custom ASICs are also pursuing separate high-voltage paths that add to the tightness of the supply chain. The investable conclusion is that the power picks-and-shovels such as cables, grid equipment, turbines, electrical systems and data-centre power distribution, can provide AI exposure that is less correlated with crowded semiconductor factor risk.‍
3. Memory remains an important swing factor. The chip layer is still attractive, but the opportunity is increasingly about process complexity, mix and pricing power rather than simply unit growth. HBM remains the tightest memory sub-segment, while NAND has also become a much more important variable after a spectacular year-to-date rally in NAND pricing and NAND-linked equities. The key nuance is that near-term EPS upside in Memory is largely driven by price increases, tight supply, and mix shift toward enterprise SSDs and AI data-centre storage. This is attractive while supply remains constrained, but it also makes the call more dependent on where we are in the pricing curve, whereas genuine greenfield memory capacity appears to be a CY27 story.
4. ‍AI is a productivity and product tailwind for the right software platforms. Enterprise software teams increasingly described AI less as a standalone product and more as a productivity layer for product development, customer support, workflow automation, engineering velocity and sales efficiency. This creates both risk and opportunity. AI can pressure undifferentiated seats and point solutions, but it should strengthen platforms with proprietary workflow data, embedded distribution, governance, auditability and measurable customer ROI. ServiceNow and Autodesk are both incumbent software platforms that are well positioned where AI appears more likely to deepen the moat than erode it.‍
5. Consumer demand is bifurcated, but value still wins. Consumer commentary was mixed, with pockets of pressure in discretionary categories, but the better operators continue to compound because their value proposition is obvious to customers. Commentary from valued-led formats such as Texas Roadhouse, Dollarama suggests consumers remain willing to spend where the experience is consistent, the offer is compelling and price increases are justified by value. Live entertainment demand also remains structurally supported by the consumer preference for experiences. The key is to avoid broad statements about the consumer and focus on traffic, mix, brand relevance, value perception and unit economics.‍
6. Europe showing selective green shoots, but the opportunity is stock-specific rather than simply “cheap Europe”. The surprise of the European leg was how steady the underlying businesses were performing. Management tone was generally upbeat, and the outlook appeared more positive than recent history would have suggested. Italian banks and financials described asset quality, deposit behaviour and fee trends as stable to improving, while several industrial and technology-adjacent companies (Prysmian and Maire) pointed to more visible order books and improving end-market conditions. The meetings also reinforced that Europe can still provide differentiated, non-US-dollar exposure to structural themes such as AI infrastructure, electrification, energy transition and semiconductor complexity. The risk in Europe remains indirect (geopolitics, regulation, disposable income, cost of capital and export controls), but the trip supports a more constructive view on the region.‍
7. Defence, public safety and mission-critical infrastructure remain durable growth markets. The public safety and defence technology meetings stood out for visibility and budget resilience. Demand for secure communications, counter-drone capability, electronic warfare, resilient connectivity, video, command-centre software and AI-assisted workflows appears structural rather than cyclical. These are areas where government and agency customers are prioritising capability upgrades even in a mixed macro environment. Motorola Solutions remains the clearest expression of this theme in the portfolio.‍
8. Valuation discipline is once again the differentiator. Many of the best businesses are already well understood by the market. The trip increased conviction that our edge will come from separating genuine multi-year earnings visibility from narrative-driven short-term multiple expansion. A number of high-quality AI and software names belong on the watchlist, but the buying discipline needs to remain strict; prioritise reasonable valuation, balance-sheet strength, clear free-cash-flow conversion and management teams that can compound through a less friendly macro tape.

New Stocks in Focus

Applied Materials (AMAT US)

The broadest single-name expression of the semicap process-complexity thesis, with direct exposure to leading-edge logic, DRAM / HBM, NAND and advanced packaging.

• The meeting reinforced that the semicap opportunity is increasingly about materials engineering, process integration and manufacturing complexity rather than simply more wafer units. That maps well to Applied Materials, whose strongest exposures sit in the parts of the flow where AI chips and memory are becoming harder to manufacture.

Action point: Initiated a position in early June.

Prysmian (PRY IM)

World number one in cables, with an oligopolistic HVDC / submarine franchise and an approximately €17bn Transmission backlog that gives utility-grade revenue visibility.

• FY2025 was a record year, with adjusted EBITDA of approximately €2.4bn and net income of approximately €1.3bn. The valuation is the constraint rather than franchise quality, with the shares trading around 17x EV/EBITDA and roughly 30x P/E.

Action point: Re-examine the fibre optics price ceiling. Likely a quality BUY on weakness.

Robinhood Markets (HOOD US)

A US consumer-fintech compounder, with a broadening multi-product ecosystem and genuine AI operating leverage.

• The platform has reached substantial scale with over $350bn of assets, a roughly 20% annualised net-deposit growth profile, and approximately 4.5m Gold subscribers, creating a monetisable base at low incremental cost. AI is already delivering operating leverage, with around 75% of support tickets automated, nine-figure software savings and roughly 50% faster engineering commit velocity. The platform also has optionality in agentic trading / commerce, prediction markets through the Susquehanna JV, tokenised private markets and international expansion.

Action point: Prioritise for deeper work and watch for an entry on volatility. Main risks are valuation, revenue cyclicality and take-rate compression.

Marex Group (MRX US)

UK-domiciled, Nasdaq-listed clearing, execution and market-infrastructure platform met; a potential capital-markets compounder with both cyclical and structural earnings drivers.

• The financial record supports doing the work. FY25 revenue grew 27% to approximately $2.0bn, adjusted profit before tax grew 30% to approximately $418m and adjusted ROE was around 30%. Q1 2026 was another record quarter, with revenue up 48% and adjusted profit before tax up 59%, helped by elevated volatility, higher exchange volumes and continued client wins.

Action point: Add as a new idea and prioritise for deeper work. Build a normalised earnings bridge that separates rate / volatility tailwinds from structural share gains, stress-test clearing balances and client-default risk, assess acquisition discipline and compare valuation against exchanges, brokers and other capital-markets infrastructure peers.

Important information

This document has been prepared by Bell Asset Management Limited (BAM) and is provided on a confidential basis solely for information only purposes of selected “professional investors” (as the term is defined in the Corporations Act 2001 (Cth)) and investors in other jurisdictions which, if they were located in Australia, would qualify as professional investors.

The distribution of this document may be restricted by local laws in certain jurisdictions and BAM accepts no responsibility for recipients of this document in those jurisdictions – recipients in those jurisdictions should inform themselves as to the requirements of, and observe accordingly, local laws. To the maximum extent permitted by law, none of BAM and its directors, employees or agents accepts any liability for any loss arising, including from negligence, from the use of this document or its contents nor does BAM assume any obligation to update the information. Recipients of this document agree to maintain the document confidential and to only use it to evaluate a potential investment in a BAM product or strategy. Recipients of this document also agree not to share this document in whole or in part with any person who is not a professional investor.

This document may contain forward looking statements and such statements are made based on information BAM holds as reliable; however, no guarantee is given that such forward looking statements will be achieved. BAM has made every effort to ensure the accuracy and currency of the information contained in this document; however, no warranty is given as to the accuracy or reliability of the information.

Information about specific investments is included for illustrative purposes, in order to assist prospective investors in better understanding the investment strategies and processes used by BAM, and is not intended to be indicative of actual future investments or performance results that will be achieved in the future. There is no assurance that similar investment opportunities will be available in the future, and the results of actual investments in the future may differ significantly.