Pictet Asset Management Holding 13F Portfolio

Pictet’s biggest adds by dollars read like a checklist of what you need to actually deploy AI at scale: metrology, testing, power, cooling and networks. They are not backing away from the AI theme; they are upgrading its industrial backbone.

• KLA: A massive +572.8% increase and a roughly $1.25B capital add turns a small line into a 1.40% book position. Importantly, the stake sits at a loss versus average cost, suggesting they are averaging down into semiconductor process control as a non-negotiable bottleneck for advanced nodes.
• Vertiv: A +298.8% add and about $470.7M more capital signals high conviction that data center power and thermal infrastructure will be a structural, not cyclical, AI winner.
• Cisco: Doubling the stake (+100.5%) and adding roughly $456.2M pushes networking hardware firmly into the AI thesis — you don’t get model throughput without switching and routing.
• Snowflake and ServiceNow: With Snowflake up +56.7% and ServiceNow up +18.1%, Pictet is leaning into workflow and data platforms that monetize AI usage rather than pure compute, even as both are around or below cost on their disclosures.
• Visa and Autodesk: The +33.5% and +30.6% increases show renewed preference for durable, software-like margin streams tied to payments and design, classic compounding franchises that can benefit from AI without needing bubble-like expectations.
• Agilent and Intuitive Surgical: Adds here, alongside a larger Johnson & Johnson position, show a quiet but deliberate build in health care tools and procedures — less story, more recurring demand.

The funding list is just as revealing: Pictet is taking real chips off the table where the market has already paid them handsomely, and where positioning risk has grown crowded.

• Nvidia and the AI chip complex: Nvidia is cut -16.9% (about $869.2M in value), while Marvell is slashed -57.6% and Applied Materials -25.6%. All three sit on triple-digit gains versus average cost; this looks like disciplined profit-taking and risk management in the hottest corner of the market.
• Cybersecurity and high-beta software: Palo Alto Networks is effectively halved (-52.9%, about $867.4M), and CrowdStrike is cut -32.6% while showing a loss relative to cost. Together with trims in Dynatrace and Twilio, Pictet is walking back from the priciest, sentiment-driven software subsectors.
• Megacap platforms: Microsoft and Amazon see -17.7% and -10.2% reductions, with Alphabet and Meta also gently trimmed. These look like position-size controls after strong runs, not a repudiation of the businesses.
• Health care and life science tools: Thermo Fisher is taken down sharply (-47.9%, roughly $495.3M), while Lilly and AstraZeneca are trimmed but still sit on healthy gains. Capital is being recycled from expensive, rate-sensitive science tools into more reasonably priced health care and AI-adjacent infrastructure.
• Cyclical and industrial exposure: Cuts to Parker-Hannifin, Trane and Ferguson show a willingness to reduce classic industrial cyclicals in favor of more structural growth and balance-sheet strength elsewhere.

On the surface, sector weights look stable: technology barely budged (65.92% to 65.22%). Underneath, the composition of that tech block — and its adjacencies — has changed meaningfully.

Telecom and networking exposure, via Cisco and Arista, jumped from 2.21% to 3.44%, a clear statement that bandwidth is now core to the thesis rather than a side bet. Finance grew from 4.19% to 4.68% as UBS and JPMorgan were increased, while the payments rails bucket (Visa and Mastercard, reported under Real Estate here) moved from 2.82% to 3.19%, effectively a stealth overweight to transaction toll collectors.

Industrials and Consumer Discretionary nudged down, reflecting trims in Trane, Parker-Hannifin, Tesla, Amazon and Ecolab. Utilities and Basic Materials are roughly flat, but within them Pictet is upgrading quality — maintaining Xcel and Republic, and holding Linde as a stable, quasi-infrastructure play.

Health care crept up from 4.98% to 5.01%, but the direction of travel is more important than the small number: less in high-multiple tools like Thermo Fisher, more in procedure growth (Intuitive Surgical) and diversified pharma (Johnson & Johnson). The aggregate picture is a portfolio still anchored in AI, yet spreading its bets across power, bandwidth, banks and biologics instead of relying on a narrow set of GPU and mega-cap narratives.

Taken together, Pictet is positioning as if the AI cycle is entering a more capital-intensive, infrastructure-heavy phase — and that the easy multiple expansion in headline names is largely behind us. They’re still long the theme, but in a way that should be less hostage to single-stock hype and more tied to installed base and recurring usage.

The big upsizing in KLA, Vertiv and Cisco says they expect bottlenecks to shift from chips to yield, power and networking. If that view is right, earnings growth in these names can persist even if GPU pricing or unit growth normalizes. At the same time, reducing exposure to cybersecurity high-flyers and trimming mega-cap platform stocks suggests they see a worse skew in crowded, story-driven trades.

Building positions in Agilent, Intuitive Surgical and Johnson & Johnson hints at a secondary thesis: health care tools and treatments as another long-duration compounder that doesn’t need a perfect macro tape. And the incremental capital to UBS, JPMorgan, Visa and Mastercard underscores confidence in resilient, capital-light earnings streams that benefit from nominal growth and are less sensitive to the exact path of AI capex.

If the next leg of the market is less about multiple expansion at the top and more about cash flows from enabling infrastructure and services, this book is aligned. If, instead, the tape keeps rewarding pure AI beta and speculative software, Pictet’s Q2 2026 moves trade some upside for resilience and a broader base of earnings drivers.