Point72 Asset Management L P 13F Portfolio · Steven Cohen

The biggest dollar adds show Point72 is willing to pay for software operating leverage and boring-but-beautiful cash flows, using the AI windfall as ammunition.

On the growth side, they turned Snowflake from a toehold into a real position. The stake is up +2448.7% in shares to $716.1M (1.09%), with the book sitting about +54.6% above their average buy. That is not averaging down; it is paying up for a data platform they clearly see as a structural winner in the AI era.

The same pattern appears in enterprise software:

• Oracle: shares up +188.1%, position now $499.8M. They are effectively underwriting a durable transition to cloud and AI workloads from a legacy vendor with real pricing power.
• Texas Instruments: a new $410.5M position in a mature analog‑chip franchise. This is a vote for steady free cash flow over lottery‑ticket AI silicon.
• Seagate: shares up +50.6% to a $794.1M stake, with the position showing nearly +192.0% versus average cost. They are pressing a high‑conviction storage call into the AI data deluge.

The other side of the ledger is pure cash generation:

• Keurig Dr Pepper: share count up +1319.3%, now a $456.5M stake. That is Point72 buying a low‑beta beverage cash machine at roughly fair value, using tech profits.
• Procter & Gamble and Mondelez: PG is up +105.8% in shares to $633.9M, and Mondelez +66.8% to $475.2M. They are explicitly upgrading the quality of their consumer book.
• Utilities and infrastructure: AEE (+207.5% shares), PPL (+47.6%), VST (+60.9%), and a new $234.5M ATO position show a clear appetite for regulated, rate‑base‑anchored earnings streams.

Rising conviction is not about chasing the frothiest AI names; it is about using AI‑era cash to own the tollbooths and necessities that keep compounding through multiple cycles.

The sell tape is blunt: Point72 is ruthlessly carving down some of the best AI‑cycle performers and crowd favorites to free capital for more durable stories.

The clearest message comes from the semiconductor and test-equipment complex:

• Teradyne: share count slashed -79.9%, taking the position down to $221.9M after an estimated -$883.5M reduction. With the stake still showing about +269.6% versus cost, this is a textbook harvest of test‑equipment beta after a furious upcycle.
• Applied Materials: -65.1% in shares, -$807.7M in estimated dollar exposure. They’re de‑risking from wafer‑fab capex cyclicality while pocketing roughly +89.1% vs average buy.
• Broadcom, Micron, Credo: AVGO (-54.8% shares, -$469.4M), MU (-45.1%, -$295.4M), and CRDO (-17.7%, -$360.7M) all see sizable trims, yet remain profitable positions (with MU and CRDO up more than +300% and +236.1% vs cost, respectively).

Outside semis, they’re lightening up on expensive digital infrastructure and megacap platform risk:

• Equinix: -41.0% in shares, -$228.7M. After a strong run, datacenter REIT exposure is being swapped for cheaper, more regulated yield in utilities.
• Arista Networks: -21.5% in shares, -$238.8M, even as AI networking remains a hot narrative. Point72 is signaling more comfort owning AI’s data layer (Snowflake, Oracle) than every last box in the rack.
• Meta Platforms: -37.5% in shares, -$191.8M, with the position modestly underwater vs cost. That looks less like risk‑off and more like a judgment that ad‑driven megacaps no longer offer the best risk‑reward versus the software and payments rails they’re adding.

Across these sells, the pattern is consistent: lock in rich semicap and infra profits, shrink exposure to crowd‑favorite AI beneficiaries, and recycle into names with better forward skew on both valuation and cyclicality.

At the sector level, this isn’t a tech exodus so much as a recalibration from high‑beta AI plays toward steadier earners and non‑correlated cash flows.

Technology drops from 48.3% to 43.95% of the book, but the remaining exposure is better balanced. Cyclical semicaps like AMAT, AVGO, MU and equipment names like Teradyne give ground, while more durable chip franchises (TXN, LSCC, STM, VSH) and data/software platforms (Snowflake, Oracle) pick up the baton.

The real shift is into classic defensives:

• Consumer staples jump from 2.73% to 5.69%, driven by big adds in Keurig Dr Pepper, Mondelez, and Constellation Brands. These are margin‑rich, brand‑heavy names that can pass through inflation.
• Utilities surge from 2.43% to 5.52%, via PPL, Ameren, Vistra, Atmos Energy, and American Electric Power. This is a deliberate move into regulated rate‑base assets with long‑dated cash flow visibility.
• Health care (UnitedHealth) rises from 0.82% to 1.54%, and finance (Capital One) from 1.55% to 2.27%, adding exposure to credit‑cycle and policy‑driven earnings that behave differently from pure growth.

Consumer discretionary actually edges up from 15.76% to 17.65%, but the mix is nuanced: trims in Amazon, Home Depot, and Spotify are offset by aggressive adds in Coupang, Sea, and travel (United Airlines) plus more idiosyncratic names like Somnigroup. Overall, sector rotation is about turning a tech‑dominant, AI‑beta book into a barbell: AI plumbing and data on one side, regulated and consumer cash cows on the other.

Taken together, the moves sketch a manager positioning for a late‑cycle AI market where multiples are frothy, macro is uncertain, and dispersion inside tech will widen.

On one side of the barbell, Point72 is keeping its hand firmly in the AI story: stakes in Credo, AMD, TSMC, ASML, Seagate, Snowflake, and Texas Instruments ensure the portfolio still benefits from rising compute intensity and data growth. But position sizing and trims show they want that exposure in more idiosyncratic, less consensus‑crowded names rather than in every high‑beta semicap or ad platform.

On the other side, the build‑out in staples, utilities, healthcare, and transaction rails (Visa, Mastercard, Oracle) suggests they are preparing for a world where rates may stay higher for longer and growth leadership could broaden. These are businesses with pricing power, regulated returns, or network effects that can grind out mid‑teens EPS in a wide range of macro scenarios.

The more speculative signal lies in the consumer internet upgrades: large adds to under‑earning platforms like Coupang and Sea, even though both sit below cost, indicate a willingness to lean into early‑stage operating leverage where the market is skeptical. That complements the expensive, high‑quality software they just paid up for.

Going forward, expect Point72 to keep trading around the edges of its AI infrastructure winners while steadily promoting compounding cash franchises up the book. The 2026‑Q2 13F shows a fund that knows it has already been paid handsomely for calling the AI build‑out — and is now quietly re‑underwriting the next three years of returns in less obvious, more durable places.