FMR LLC 13F Portfolio

The biggest adds cluster around one idea: own the indispensable nodes of the AI economy. Nvidia, already the book’s anchor at 8.91%, still saw the share count lifted by +3.2%, a notable move at that scale and a signal they think upside remains despite a +1774.6% gain vs their average cost.

The more revealing moves are away from the headline leaders toward under‑owned or catch‑up beneficiaries deeper in the stack:

• Broadcom was boosted by +16.3%, a $7.65B increase, reinforcing the thesis that custom accelerators, networking silicon, and edge connectivity are long‑duration assets in AI and cloud.
• AMD and Intel were lifted sharply (+67.2% and +64.2% in shares, respectively), a textbook “upgrade the laggards” bet that the AI compute pie is big enough for more than one GPU/CPU vendor.
• ASML, Lam Research, Applied Materials, Monolithic Power, and NXP all saw incremental adds, underscoring a belief that the bottleneck is not just GPUs but the lithography and process tools that enable leading‑edge capacity.

Outside pure semis, conviction is quietly rising in durable, cash‑rich growth. Alphabet’s GOOGL and GOOG lines were both scaled up, while UnitedHealth (+132.9% in shares, a $6.64B add) and Eli Lilly (+8.1%) show a conscious build‑out of health‑care compounders as a second leg of the book. Adds to Tesla, Thermo Fisher, and a modest increase in Coca‑Cola round out a pattern: Fmr is paying up for platforms where pricing power and data, not just units, drive the P&L.

If the buys tell you what Fmr wants to own through the next leg of the cycle, the trims show what they now see as funding sources. The largest cuts are not expressions of fear; they’re profit‑taking in names that already crystallized the early AI and digital‑consumer wave.

The most aggressive reductions came in semis that ripped on the first phase of the AI build‑out:

• Marvell was slashed by -56.6% in shares, a $22.14B reduction, and SanDisk by -41.3% ($12.57B). Western Digital and Micron were also cut double‑digits by shares. These were massive winners — Micron is up +1116.0% vs Fmr’s cost, SanDisk an eye‑popping +3600.9% — and are now being treated as cash registers.
• Meta’s position was chopped by -17.8%, a $11.67B trim, and Netflix by -33.6% ($4.91B). Both have run hard since their lows; Fmr is signaling that monetizing eyeballs is less interesting than owning the rails those eyeballs run on.
• GE Vernova and Berkshire Hathaway A were reduced by -11.9% and -14.5% in shares, respectively, freeing “quality but fully‑valued” capital for higher‑conviction themes.

Financials are also clearly being used as liquidity. Wells Fargo, Bank of America, and JPMorgan all saw mid‑single to mid‑teens percentage share reductions, despite healthy gains vs cost. That is less a macro call on credit and more an acknowledgment that, in Fmr’s opportunity set, big banks simply don’t compete with AI chips or GLP‑1 pipelines for incremental dollars.

The sector chart shows a book that is firmly anchored in technology, but with a slow, deliberate broadening into health care and industrial quality. Tech’s share of the top‑50 barely budged, from 70.89% to 70.64%, despite some eye‑catching trims — evidence that proceeds from memory, storage, and older communications bets were largely recycled into other tech subsectors rather than taken off the table.

The genuine net shifts are elsewhere:

• Health care climbed from 4.32% to 5.06%, powered by the scale‑up in UnitedHealth and steady adds to Eli Lilly and Philip Morris’s medicinal‑chemicals profile. Fmr is clearly building a second growth spine anchored in reimbursement‑resilient, IP‑rich franchises.
• Industrials expanded from 2.64% to 2.95%, driven by adds in Tesla, Boeing, and Thermo Fisher — all levered to long‑cycle capex (EVs, aerospace, life‑science tools) rather than short‑cycle GDP beta.
• Finance edged down from 2.81% to 2.58%, Energy from 2.14% to 2.07%, and the unclassified bucket from 3.48% to 3.18%, reflecting the use of banks, oil majors, and diversified holding companies as funding sources.

Consumer Discretionary and telecom‑adjacent networking remain largely stable in aggregate, even though individual names diverge sharply. Cutting Netflix while adding Lowe’s, and trimming Arista while adding Cisco, suggests a preference for cash‑rich, more reasonably valued cyclicals and incumbents over pure multiple expansion stories.

Taken together, this 2026‑Q2 snapshot reads like a manager that believes we are in the middle innings of an AI and data‑infrastructure super‑cycle, not the tail. They are content to let some early winners go — memory, storage, and social‑media ad platforms — in order to concentrate capital in the most durable choke points: compute, lithography, networking, and security.

At the same time, Fmr is not running a one‑factor bet. The build‑up in health‑care compounders and industrial platforms suggests they’re preparing for a world where AI‑driven productivity, rising capex, and aging demographics matter as much as ad impressions and e‑commerce volumes. Performance over the past three years (29.87% annualized, 119.05% cumulative) gives them the balance sheet and the confidence to keep pressing that edge.

Going forward, expect more of this “upgrade within the theme” behavior rather than dramatic sector whipsaws. If AI infrastructure and health‑care IP continue to absorb flows while banks, old‑economy energy, and fully‑valued internet platforms get chipped away, the message will be consistent: Fmr wants to own the rails and the reagents of the next decade, not just the apps on top.