Mitsubishi UFJ Asset Management 13F Portfolio

The “biggest buys” widget doesn’t show a spray of small tweaks; it shows a handful of conviction swings into the AI supply chain and its platform beneficiaries. KLA, Apple, Alphabet, NVIDIA, Amazon, Microsoft, Marvell, and Broadcom dominate the dollar adds — that’s deliberate concentration in the plumbing of modern compute.

• KLA: A nearly 10x position expansion (up 968.0%) to $1.01B, even though it sits 65.1% below the fund’s average cost, is the quarter’s loudest signal. They’re averaging down in one of the most cyclical, but structurally advantaged, wafer inspection and process-control franchises, effectively betting that AI fab capex stays elevated through volatility.
• Apple, Microsoft, Alphabet: Adding $764.7M to Apple, $374.4M to Microsoft, and $583.1M to Alphabet reinforces a view that cloud, device, and ad platforms will be the durable bottlenecks for AI distribution, not easily displaced by new entrants.
• NVIDIA, Broadcom, Marvell, Micron, AMD, Lam Research, Applied Materials: Across these names, the direction is one-way: more capital. NVIDIA and Broadcom, both already extremely profitable AI winners, still see large incremental dollars. Marvell’s +121.8% share jump and Micron’s and AMD’s high embedded gains suggest MUFG is willing to chase where earnings revisions are strongest, not just where multiples are cheap.

The through-line is clear: they are willing to lean into drawdowns (KLA, Marvell) and extend winners (NVIDIA, Alphabet) so long as the asset sits on the AI-critical path — chip design, manufacturing, networking, or hyperscale deployment.

On the sell side, the trims are few but telling: this is not a broad de-risking, it’s surgical reallocation away from stable, slower-growth franchises toward higher-operating-leverage AI exposure. The biggest dollar reduction is Texas Instruments, followed by modest cuts in Coca-Cola and S&P Global.

• Texas Instruments: A -21.6% cut and -$331.9M estimated dollar reduction, despite the position still sitting about 65.6% above cost, looks like a funding source decision. TI is a fantastic analog franchise, but less directly levered to the datacenter and AI cycle than the NVIDIA/Micron/KLA cluster they’re building.
• Coca-Cola: The -3.2% trim in a Coke stake that’s up 43.2% versus cost frees cash from a pure defensive without signaling any real macro view shift; consumer staples exposure overall actually falls slightly from 1.61% to 1.45%.
• S&P Global: The -1.8% trim here is tiny in dollar terms, but notable in that it takes money from a high-quality, fee-based information monopoly to help fund a more cyclical, capex-driven AI build-out.

Notably absent are broad reductions in the mega-cap tech winners: NVIDIA, Apple, Microsoft, Amazon, Alphabet, Meta, and the core semi equipment names are all being added to, not pared back. That tells you the conviction hierarchy: legacy quality compounders can be tapped to fund what they see as a structurally superior earnings trajectory in AI infrastructure.

The sector bar chart confirms what the name list implies: this is a tech-first, AI-centric equity book, with everything else playing a supporting role. Technology inches up to 59.14% from 58.74%, even as the fund gets more selective within semis — recycling from Texas Instruments into tools like KLA, Lam Research, and Applied Materials, plus data-center-facing chips like NVIDIA and Marvell.

Outside tech, net moves are incremental but directionally consistent. Health care edges up from 7.29% to 7.34% as they add to Eli Lilly, Johnson & Johnson, AbbVie, Merck, UnitedHealth, and Philip Morris — a subtle statement that drug pricing and GLP‑1 dynamics are opportunities, not reasons to avoid the space. Consumer staples slips as Coca-Cola is trimmed, while consumer discretionary stays roughly flat at 11.71%, but with more love for scale retailers and platforms like Amazon, Walmart, Costco, and Home Depot.

Financials ease back from 6.10% to 5.95%, even as they add to JPMorgan, Goldman, Bank of America, Aon, and ICE, offset by the S&P Global trim — a sign they still like capital markets and fee platforms but are unwilling to expand the sector’s overall risk budget. Energy and industrials weights are largely steady; adds to Exxon, Chevron, Caterpillar, Tesla, and Thermo Fisher look like inflation and reshoring hedges rather than primary performance engines.

Taken together, the quarter’s moves say Mitsubishi UFJ believes the AI build-out is still in its middle innings — and that the bottlenecks will live in semicap tools, high-bandwidth chips, and hyperscale platforms, not in commodity cyclicals or bond-like defensives. The willingness to average down aggressively in KLA and add meaningfully to Marvell, while extending already huge winners like NVIDIA and Alphabet, suggests they see earnings power compounding faster than sentiment can catch up.

At the same time, they’re not abandoning ballast: health care majors, consumer staples, big-box retail, and diversified financials remain meaningful sleeves, just not the growth engine. Those positions look designed to keep drawdowns tolerable while the AI cluster does the heavy lifting on returns.

For observers, the key takeaway is that this is no longer just a “quality global equities” portfolio; it’s morphing into an AI infrastructure barbell surrounded by cash-generating moats. If that thesis is right, the book is set up to benefit from several more years of elevated capex and cloud profit growth. If it’s wrong, the quarter’s high-conviction adds in cyclical tools and rich platforms will be where the pain shows up first — and where Mitsubishi UFJ will have to decide whether this was a one-cycle bet or a decade-long worldview.