Swedbank AB 13F Portfolio

The most striking move is the massive capital redeployment into semiconductor capital equipment and related infrastructure. Swedbank lifted KLA by +763.9% to 1.43% of the book, adding about $1.47B despite the position sitting roughly -68.4% versus their own average cost – a classic averaging-up-in-size trade keyed to process control’s centrality in leading-edge fabs.

They paired that with heavy adds to Lam Research (+29.6%, +$459.7M) and Applied Materials (+2.0%, +$40.5M), while also increasing Taiwan Semi and Micron. That cluster says they are betting the bottleneck in AI shifts from chips to capacity: more etch, deposition, and yield management at foundries, and far more high-bandwidth memory to feed GPUs.

On the compute side, they grew AMD by +26.8% (+$521.2M) and Broadcom by +5.9% (+$276.0M), complementing a still-massive Nvidia at 7.51% of the book. This is a deliberate barbell: Nvidia remains the profit engine, but they are building out exposure to alternative accelerators, custom silicon, networking ASICs, and interconnect.

Conviction is also rising in mission-critical enterprise software. ServiceNow was almost doubled (+98.7%, +$298.9M), Cadence rose +17.9% (+$123.6M), and Oracle ticked higher (+4.7%). These are the workflow, EDA, and database platforms that monetize AI productivity directly in corporate IT budgets, rather than relying on ad cycles.

Outside pure tech, adds in UnitedHealth (+55.9%, +$258.0M) and Eli Lilly (+10.2%, +$197.9M) quietly push health care exposure up to 8.29%. The mix – obesity drugs, specialty pharma, and managed care – suggests they want structural health spend, not just defensive yield.

The funding side of the ledger is telling: they are not abandoning tech, they are pruning around the edges of prior winners to free up cash for the semicap and enterprise-software push. Alphabet’s GOOG line was cut -4.6% (about -$165.6M), while Meta was trimmed -8.3% (-$144.3M). These are still big, profitable platforms, but incrementally less attractive than the upstream enablers they’re buying.

Old-guard networking is clearly a source of funds. Cisco was reduced by -15.7% (-$160.7M), and Arista was nudged down -2.1%. That capital is effectively being recycled into KLA, Lam, and high-speed connectivity names like Qualcomm (+70.0%, +$197.1M) and Amphenol (+47.6%, +$272.2M), which are more levered to AI-era bandwidth and signal integrity.

Health care cuts are targeted, not broad. McKesson was reduced by -21.1% (-$160.1M) and Royalty Pharma by -7.4%, even as Lilly, J&J (+24.1%, +$178.6M), AbbVie (+9.9%) and UnitedHealth were increased. That rotation within the sector shifts them from lower-margin distribution and royalty structures toward branded innovation and payor scale.

Cyclicals and consumer exposure are also being eased back at the margin. Home Depot was trimmed -13.5% (-$108.8M), Netflix -7.0%, Spotify -4.8%, and Prologis only modestly increased. Combined with small trims in Mastercard, Visa, Berkshire, and Microsoft, this reads as profit-taking in economically sensitive and already-rich names to load up on the parts of the stack they think still underprice AI’s throughput demands.

At the sector level, the changes look small in percentage points but meaningful in composition. Technology’s share of the book climbed from 67.79% to 69.2%, but it’s a different tech mix: more semiconductors and wafer-fab tools, more EDA and workflow software, less social, less legacy networking.

Consumer Discretionary edged down from 8.66% to 8.19% as Swedbank chipped away at Home Depot, Netflix, Spotify, and held Costco effectively flat. They are still long Amazon, now 3.39% of the portfolio after a +3.6% add, signaling they want the structural ecommerce and cloud exposure, not the more cyclical housing and media legs of consumption.

Health care ticked up to 8.29% from 8.18%, but the tilt matters more than the magnitude: out of distributors and royalty vehicles, into innovators and diversified payors. Finance slipped slightly to 5.36% as they modestly added to large banks but left payment rails as quasi-“real estate” underweights.

Telecom and legacy comms equipment declined from 2.86% to 2.48% with the Cisco and Arista trims, while Industrials fell as they cut Thermo Fisher by -12.3%. Utilities, represented by Waste Management, actually rose from 0.72% to 0.79% as they added +15.8% – a small but telling desire for regulated, inflation-linked cash flows.

Net-net, the portfolio is migrating from ad- and consumer-driven tech to infrastructure and tools. They are accepting a bit more capital-intensity risk in exchange for a claim on the physical and logical underpinnings of AI compute growth.

Put together, this quarter reads as Swedbank explicitly re-underwriting a thesis: AI’s next leg is about capacity, not discovery. By plowing nearly $1.47B into KLA at a substantial mark-to-cost loss and scaling Lam, Applied, Broadcom, AMD, Micron, Cadence and ServiceNow, they are betting that the market still underestimates how much silicon, tooling, and software plumbing needs to be built.

They are financing that shift not by slashing winners, but by dialling back the parts of the book where upside now looks more beta-like: big platforms (Alphabet, Meta, Microsoft), legacy networking (Cisco), distributors (McKesson), and rate- and cycle-sensitive consumer exposure. Importantly, their trims are measured – often in the mid-single-digit to low-teens percent – suggesting a view that these remain solid but no longer scarce assets.

The subtle build in health care – especially Eli Lilly, J&J, AbbVie, and UnitedHealth – adds a second pillar: structural health spend and GLP-1–driven therapeutics as a long-duration growth and defensive buffer alongside tech. Waste Management and Linde contribute similar “never-turn-it-off” characteristics in services and industrial gases.

Looking ahead, unless the macro or regulatory backdrop breaks the AI supply chain narrative, expect more of this: incremental rotation inside tech toward higher operating leverage to wafer starts, bandwidth, and AI productivity software. If the cycle turns or capacity overshoots, the moves into resilient health care, banks, and hard-to-disrupt services show they are at least partially hedging that call – but for now, Swedbank’s book is very clearly constructed around the idea that the picks, shovels, and factories of AI remain underpriced relative to the gold rush they support.