Price T Rowe Associates Inc MD 13F Portfolio

The biggest adds cluster tightly around one idea: the AI bottleneck is migrating from headline GPUs to capacity, memory, and tools.

• Micron (MU) is the standout: a 165.3% position increase and about $10.2B of incremental capital says they see high-bandwidth memory as the next choke point in AI systems, not an afterthought.
• Alphabet GOOGL is up 25.8% with roughly $7.4B added, even as the GOOG line is trimmed. That is a deliberate consolidation into a single share class and a statement that hyperscale AI infrastructure and first-party models still earn a premium slot.
• Lam Research (LRCX) and Applied Materials (AMAT) both more than double in share terms (up 85.8% and 108.3% respectively in shares), with a combined near-$9.7B increase. That is a straight read-through: fabs, not front-end apps, are where the incremental dollar of AI capex gets spent.
• AMD and Marvell (MRVL) see aggressive adds — AMD up 24.6%, Marvell up 714.3% in shares and about $3.6B more capital — which rounds out a bet that alternative accelerators, networking silicon, and custom silicon will carve out real profit pools alongside Nvidia.
• Taiwan Semi (TSM) and ASML are both boosted (TSM shares up 38.7%, ASML up 25.7%), tying the thesis together: if you think AI volumes sustain, you must own the bottleneck nodes and the lithography monopoly.

Around the edges, they also add to linchpin analog/connectivity names like Amphenol and to utility Southern, suggesting they see AI as a power and bandwidth story as much as a chip story.

To fund that second-wave AI build-out, Price T Rowe is taking real money off the table in the most obvious winners of the last cycle.

• Apple and Microsoft are both meaningfully cut (Apple shares down 20.3%, Microsoft down 25.6%), removing roughly $11.8B and $11.2B of exposure at gains north of 200%–400% versus their average costs. These aren’t thesis breaks; they’re valuation and concentration releases.
• Amazon and Meta are trimmed as well (Amazon shares down 11.5%, Meta down 6.6%), indicating less appetite for ad and e-commerce cyclicality relative to more direct AI infrastructure leverage.
• Within Alphabet, they rotate: GOOG is reduced by 12.7% while GOOGL is sized up. That looks like a governance and liquidity preference rather than a change in view on the underlying business.
• On the networking and bandwidth side, Arista (ANET) is cut by 20.8% and Cisco modestly trimmed, just as Marvell and equipment names are boosted. They’re effectively saying the best risk-reward in AI networking now sits in silicon and wafer tools, not in box vendors.
• Consumer and financials are quiet funding sources: Netflix, Booking, Procter & Gamble, Charles Schwab, Bank of America, and Visa all see net selling, marking them as tradable cash reservoirs rather than engines of future excess return in this framework.

At the sector level, the wheel barely turns — but the spokes are different. Technology creeps from 61.63% to 62.55% of the book, yet inside that bucket they shift from platform-heavy FAAMG exposure toward semiconductors, foundries, and equipment.

Consumer Discretionary ticks down from 9.94% to 9.0% as Amazon, Netflix, Booking, and Carvana bleed capital, even while Home Depot is quietly added. That looks like a move away from online ad and demand-volatility toward more tangible-cycle plays tied to housing and construction.

Health care edges up from 7.92% to 8.43%, led by sizable increases in Johnson & Johnson, UnitedHealth, Gilead, Stryker, and AbbVie. This is classic ballast: cashing in hyper-growth tech gains and upgrading the defensive spine in large-cap pharma, managed care, and medtech.

Finance and Real Estate both slip a touch (Finance from 5.01% to 4.81%, Real Estate from 4.91% to 4.56%) as banks, brokers, and card networks get trimmed while Welltower is increased. They’re keeping exposure to financial plumbing and real-asset income, but clearly not paying up for incremental cyclicality.

Industrials, Energy, Basic Materials, and Utilities all nudge higher, with adds to Caterpillar, Exxon, Linde, and Southern. That combination reads as a quiet macro overlay: if AI drives capex and power use, heavy equipment, chemicals, and electricity should not be an afterthought.

Taken together, the quarter describes an investor that believes the easy money in "AI beneficiaries" at the application layer has been made — but the hard money in fabs, memory, tools, and power is still ahead. The modest increase in overall tech weight hides a very intentional migration from glossy narratives to capital-intensive bottlenecks.

The pattern of trades suggests they expect AI demand and related capex to prove more durable than consumer or advertising cycles. They’re willing to own more Micron, Lam, Applied, TSM, and ASML precisely because those are the names that get hurt first in a downturn; in exchange, they dial down richly valued, broadly owned platforms like Apple, Microsoft, Amazon, and Visa.

Simultaneously, they’re upgrading the defensive core: more diversified pharma and managed care, plus a bit more utilities and real assets. That mix — higher operational leverage to AI plus sturdier health care and infrastructure — looks like preparation for a world where index-level multiples might compress even as select industrial ecosystems still compound.

If you want to understand where this book goes next, watch two clusters: the semiconductor equipment and memory complex, and the big diversified health care names they have been building. If those keep growing at the expense of consumer internet and financials, Price T Rowe will have finished turning a broad "quality growth" book into a highly opinionated bet on the physical backbone of the AI economy.