Bank of New York Mellon Corp

The “biggest buys” list makes it clear: they are upgrading from broad semiconductor beta to the specialist tools and connectivity that sit deepest in the AI capex stack. KLA and Marvell are the poster children — both aggressively scaled and still sitting on modest gains versus cost, implying these are being built as ongoing growth engines rather than just momentum trades.

• KLA: An 841.1% share jump to $2.52B (0.41%) screams structural conviction in process control and inspection as AI fabs get denser and more complex; at only 61.3% above their average buy price, they seem to think the runway is still long.
• Marvell: A 153.3% increase to $2.37B (0.39%) is a straight bet on high-speed networking and custom silicon as AI workloads move off the whiteboard and into distributed datacenters.
• Microsoft: A 3.5% add, lifting the stake to $19.9B (3.27%), underscores preference for cloud and software monetization of AI rather than just GPU units shipped.
• IBM: A 29.3% add to $2.61B (0.43%) suggests they still see upside in hybrid cloud and enterprise AI services from a relatively undemanding starting valuation.
• UnitedHealth and Thermo Fisher: Incremental adds here show they’re still buying structural health-care growth — medical services and life-science tools — as secular compounders alongside the AI narrative.
• Coca-Cola and Philip Morris: Double-digit percentage increases into these cash-gushers show an appetite to lock in predictable, high free-cash-flow franchises as a counterweight to tech volatility.

Taken together, the buys argue for a barbell: AI infrastructure and platforms on one side, and durable cash-flow machines in staples and health care on the other.

The trims are not a repudiation of semis or cyclicals so much as a refinement: they’re cashing in on the broadest, most cyclical winners to fund a more surgical AI toolchain. The heaviest cuts in dollar terms hit memory, foundry, and generalist chip names rather than the highest-margin enablers.

• Applied Materials: An 18.4% cut and roughly $1.14B off the table is a classic “harvest the workhorse” move after a huge run, recycling capital from a broad wafer equipment bellwether into more targeted bets like KLA.
• Micron and Taiwan Semi: Mid-teens and mid-single-digit reductions here suggest they’re less interested in riding the full memory and foundry cycles now that AI enthusiasm is well-capitalized in prices.
• Apple and Alphabet (both share classes): Modest low-single-digit trims across AAPL, GOOGL, and GOOG look more like disciplined risk management in oversized, high-multiple winners than any break with the platform thesis.
• Cisco and Texas Instruments: Double- and high-single-digit trims look like a quiet de-emphasis of slower-growth, more mature hardware names.
• Caterpillar, Exxon, Chevron: Meaningful cuts — including a 14.6% reduction in Exxon and 13.1% in Caterpillar — signal waning enthusiasm for capital-intensive, macro-sensitive cash flows as AI and quality defensives soak up more of the risk budget.

The pattern is consistent: lighten up where earnings are more hostage to the economic cycle and commodity prices, and recycle into higher perceived structural growth or safer, non-cyclical income.

At the sector level, the headline weights barely move — but under the surface the portfolio gets sharper. Technology inches up from 55.7% to 55.9% of the disclosed book, and that tiny net change masks a meaningful internal reshuffle from broad semis and legacy hardware into AI-enabling equipment and networking.

Health care climbs from 7.67% to 7.89%, helped by adds to UnitedHealth and steady pharma exposure, underscoring its role as a structural-growth and defensive anchor. Consumer staples rises from 0.77% to 0.93% on the back of a 20.4% increase in Coca-Cola, while consumer discretionary nudges up to 9.27% via Amazon, even as they trim Walmart, Costco, and Home Depot.

Energy drops from 2.85% to 2.55% after significant trims in Exxon and smaller reductions in Chevron, signaling less appetite for hydrocarbon leverage at this stage of the cycle. Industrials ease from 3.95% to 3.8% as Caterpillar and Tesla are cut, while telecommunications exposure via Cisco falls from 1.78% to 1.59%.

Unclassified holdings — broad ETFs plus Berkshire and GE Vernova — stay roughly stable around 8.9%, acting as core market and quality conglomerate ballast. Finance edges slightly higher, from 5.76% to 5.82%, driven by a modest add to JPMorgan amid tiny trims elsewhere, keeping a toe in rate- and credit-sensitive upside without making it the core story.

This quarter’s repositioning says BNY Mellon wants to own the infrastructure of AI and the cash flows that can survive any macro tape. They’re not chasing the newest AI narratives; they’re concentrating into the hard-to-replicate process control, networking, and cloud platforms that turn AI enthusiasm into recurring capex and software revenue streams.

The clear funding sources — memory, foundries, mature hardware, cyclicals, and energy — suggest they see diminishing risk/reward where capital expenditures are high and pricing power more contested. At the same time, larger positions in stalwart consumer and health-care names indicate a desire to pair that AI cyclicality with boring-but-beautiful compounding cash flows.

With broad ETFs and Berkshire still meaningful, they’re keeping a diversified core under a more opinionated set of satellite bets. Going forward, expect more of the same: incremental upgrades within tech toward higher-quality, more specialized AI exposure, occasional profit-taking in macro-levered cyclicals, and selective adds to global, high-margin consumer and health-care franchises whenever volatility presents a chance to buy long-duration cash flows at a discount.