Bridgewater Associates LP Dumps Index Beta To Double Down On AI Supply Chain

The biggest adds table reads like a blueprint for an AI hardware supercycle paired with a physical-economy upturn. The single largest dollar add is Amazon, where Bridgewater boosted the stake by +125.3% to 4.08% of the book, effectively declaring that hyperscale cloud, retail logistics, and AI-native services are still mispriced in aggregate indices.

Semiconductors are the centerpiece. They opened a new 1.62% position in Taiwan Semiconductor worth $364.0M, a textbook move if you believe leading-edge foundry capacity is the real choke point in AI. Existing chip bets were aggressively scaled: Broadcom up +57.4%, Micron up +65.9%, and Nvidia up +21.4%. All sit comfortably in the green versus average cost, with Micron showing a gain vs buy of +275.1%, suggesting they’re pressing winners rather than fading them.

The story doesn’t stop at chips. Alphabet’s primary line (GOOGL) was lifted +25.6%, even as the non-voting GOOG line was slashed (more on that later), effectively upgrading to the cleaner exposure while staying long AI software and ad recovery. Further down the list, they nearly doubled Marvell (+97.4%), raised Applied Materials (+17.2%), and added to ASML, KLA, Teradyne, Credo, and ARM — a coherent bet on the full stack from lithography and wafers to interconnect and chip design.

Outside tech, they made striking new or outsized cyclicals. Union Pacific was ramped +877.5% to $129.3M, Nucor arrived as a new $104.6M position, and CSX jumped +243.8%. That trio, alongside a new PACCAR stake and a +284.0% increase in Baker Hughes, says they’re positioning for more freight, more steel, more trucks, and more energy services in a capex-heavy world.

On the macro-hedge side, they nudged up gold via Newmont and Barrick, and added slightly to emerging markets (VWO, IEMG), preserving some diversification even as the core of the book leans harder into growth and cyclicality.

The funding sources for this shift are crystal clear: broad index beta out, crowded winners partially harvested, and defensive staples cut back. IVV is the biggest ATM — shares are down -36.0%, freeing an estimated $984.1M while the position remains a hefty 7.81%. SPY, still the single largest line at 12.67%, was nudged down -2.0%, unlocking another ~$58.9M without touching its strategic role.

Inside tech, Bridgewater is rotating within the winners, not abandoning them. Lam Research was trimmed -48.3% after a massive run, even though it’s still showing a +236.8% gain vs cost, and AMD was cut -22.0% despite a +297.1% gain. GOOG (the non-voting Alphabet line) was slashed -75.7%, while the GOOGL line was increased — a classic clean-up trade: consolidate exposure, simplify governance, and recycle capital into higher-conviction structures.

They also took chips out of GEV, cutting GE Vernova by -34.7% after a spectacular +237.2% gain vs cost, hinting that the power-equipment trade has moved closer to fair value in their models. On the defensive side, Johnson & Johnson (-47.0%) and Altria (-48.0%) were both aggressively reduced, nearly halving health-care and "bond proxy" exposure in one go. Smaller trims to South Korea via EWY and to ex-China EM via EMXC suggest a subtle preference for broader EM baskets (VWO, IEMG) over single-country risk.

Net-net, the sell list reads less like fear and more like discipline: monetize factor-like beta and mature defensives to fund more idiosyncratic, cycle-sensitive growth exposures.

The sector bar chart tells an unusually clean story for a firm of Bridgewater’s scale. Technology has stepped up to 40.5% of the top-50 book from 35.79%, while the catch-all “unclassified” bucket — almost entirely ETFs — sank nearly 10 percentage points. They are trading the comfort of diversified wrappers for the sharper edges of single-stock risk.

Within tech, the mix has migrated down the value chain. There’s still platform exposure via Microsoft, Alphabet, and Meta, but the real weight increase is in semiconductors and equipment — Nvidia, TSMC, Broadcom, Micron, Marvell, Applied Materials, ASML, KLA, ARM, Credo, and Teradyne — plus electronic components like Amphenol, Seagate, and Celestica. The new Synopsys position completes that arc, giving them a foothold in the EDA software that effectively taxes every advanced chip design.

The other big move is a quiet embrace of the real economy. Consumer discretionary rises to 8.22%, but this isn’t just ecommerce: Amazon sits alongside PACCAR and Baker Hughes, meshing digital and industrial demand. Industrials at 3.78% now include rails (Union Pacific, CSX), steel (Nucor), construction services (Comfort Systems), and test gear (Teradyne) — exactly the kind of names that benefit from reshoring, infrastructure spend, and data center build-out.

Meanwhile, health care is cut back to 1.94%, mostly J&J and Altria, signaling that traditional defensives are no longer a priority. Basic materials (gold miners Newmont and Barrick) tick up modestly, and utilities (NRG, Talen) inch higher, adding a thin layer of inflation and power-price optionality. Finance remains a rounding error at 0.62%, with a single Citigroup line, underscoring that this is not a banks-led macro trade.

Put together, the 2026-Q1 book looks like a deliberate barbell: AI infrastructure and digital platforms on one side, real-economy beneficiaries of capex and reshoring on the other, tied together with a modest sleeve of macro hedges. The -6.43% quarter likely reinforced the lesson that owning generic beta into factor whiplash is less attractive than selectively leaning into structural winners.

On the tech front, Bridgewater is effectively underwriting the thesis that the AI cycle is capacity-constrained, not demand-constrained. Heavy adds in TSMC, Nvidia, Broadcom, Micron, Marvell, ASML, and Synopsys signal belief that the bottlenecks — leading-edge wafers, memory, networking, design tools — will enjoy durable pricing power beyond a single hype phase. Alphabet, Amazon, and Microsoft sit atop that stack as monetization engines.

The complementary bet is that physical infrastructure has to catch up. Bigger stakes in rails, trucks, steel, construction services, and oilfield equipment suggest an expectation of elevated freight volumes, industrial investment, and energy services demand — not a quick reversion to a low-growth, low-capex world. Utilities and data-center-adjacent names (like Vertiv in the tail) hint at an appreciation of the grid and cooling constraints behind AI.

Gold miners and investment-grade credit ETFs (LQD, MBB) provide just enough ballast to keep the portfolio from becoming a one-way growth bet, especially if rates or inflation surprise. But the direction of travel is clear: less anonymous exposure, more targeted risk in the choke points of AI and industrial capacity. Unless the macro regime changes abruptly, expect future quarters to rhyme with this one — incremental trims to broad ETFs and mature defensives, and steady reinforcement of the AI-and-infrastructure spine they’re building.