Citigroup INC

The biggest adds show a manager who thinks the AI cycle is nowhere near done — and is now about both the headline winners and the plumbing underneath.

On the marquee side, Nvidia was boosted by +23.0% in shares to a 5.70% position, with the stake now sitting about 93.8% above the fund’s average cost. Microsoft and Apple also saw double‑digit share increases, adding roughly $887.9M and $874.4M respectively, reinforcing the view that hyperscale cloud and premium devices are durable toll booths on AI adoption.

More telling is where they’re moving down the stack. KLA exploded by +1093.6% in share count and roughly $1.28B of incremental capital, despite the position being about -67.2% versus the fund’s average buy price — this is classic averaging down into a mission‑critical wafer‑inspection name. Lam Research, Applied Materials, TSMC, and Analog Devices all saw mid‑teens to mid‑20s percentage share adds, signaling a broad bet that semiconductor equipment and specialty silicon are the real bottlenecks.

The AI thesis extends beyond chips. Tesla’s share count jumped +53.0%, Amazon +26.5%, and Meta +21.9%, effectively treating EVs, logistics, and ad platforms as leveraged plays on AI‑driven demand and efficiency. At the same time, Elevance Health was ramped by +777.9% in shares and Eli Lilly and AstraZeneca were scaled up, pointing to rising conviction that managed care and pharma can compound earnings even if AI‑rich multiples wobble.

If the buys are precise, the sells are blunt: this quarter, Citigroup clearly chose to stop paying for undifferentiated beta and low‑return financial leverage.

On the funding side, the largest dollar reductions were in broad ETFs and a big money‑center bank. The iShares Russell 2000 position was cut -51.4%, freeing about $1.31B; SPY was trimmed -20.9%; and the manager eased off IGV and IBB. That is capital pulled from generic U.S. equity and sector beta and re‑deployed into hand‑picked single names.

In financials, Bank of America took a -56.5% share hit and roughly $1.29B in estimated dollar reduction, even as JPMorgan and Goldman Sachs were boosted by +34.6% and +59.7% in shares. The message is straightforward: if they want exposure to higher rates and capital markets, they prefer best‑in‑class franchises over more rate‑sensitive, politically exposed balance sheets.

Within tech, they are pruning around the edges rather than exiting the theme. Qualcomm, Marvell, and Oracle all saw single‑ to mid‑teens percentage trims despite sitting on gains versus average buy price in most cases. That looks less like a thesis change, more like risk budgeting — taking chips off the table in crowded or less differentiated parts of the stack to pay for high‑conviction adds like KLA and Nvidia.

At the sector level, the book is still dominated by tech, but the mix is evolving meaningfully. Technology’s reported weight edged up to 62.14% from an estimated 61.30%, but that small headline change masks a big internal rotation from broad tech exposure into specific AI infrastructure and security names.

Unclassified holdings — largely broad ETFs and macro products — fell from 17.55% to 15.13%, consistent with the heavy trims in IWM, SPY, IGV, and IBB. Instead of owning “the market,” Citigroup is increasingly picking the winners it wants: Nvidia and Broadcom in semis, Palo Alto Networks in cybersecurity, and an enlarged stake in Microsoft and Alphabet for cloud and software.

Health care rose from 3.05% to 3.95%, driven by larger positions in Eli Lilly, Johnson & Johnson, AstraZeneca, and the step‑change add in Elevance Health. That’s a noticeable tilt toward earnings‑resilient, pricing‑power businesses that can hold up if growth multiples compress.

Outside growth, the fund nudged up its Energy (0.56% to 0.68% via Exxon Mobil) and Industrials (3.72% to 4.37% via Tesla and Caterpillar) exposure, while Financials slipped from 5.30% to 4.35%. Add in more GLD and TLT, and the picture is of a manager willing to run an aggressive AI‑heavy core but buffered by health care, energy, and explicit interest‑rate and inflation hedges.

The 2026 Q2 moves say Citigroup expects the AI build‑out to behave less like a hype cycle and more like a multi‑year capex super‑cycle — and is structuring the book accordingly.

By fortifying Nvidia, Microsoft, Apple, Amazon, Alphabet, and the chip‑equipment complex, the manager is effectively underwriting years of elevated data‑center, device, and networking spend. The aggressive averaging down in KLA, and consistent adds in Lam Research, Applied Materials, TSMC, Analog Devices, and Intel, imply a view that semiconductor capital intensity and complexity keep ratcheting higher, not mean‑reverting.

Yet this is not a one‑way macro gamble. Bigger positions in Elevance Health, Eli Lilly, Johnson & Johnson, and AstraZeneca introduce earnings streams less tied to the cycle or AI enthusiasm. Simultaneously, scaling into GLD and TLT adds a direct hedge against rate volatility, geopolitical shocks, or a growth scare.

Going forward, expect this book to keep trading around the same playbook: concentrate further into perceived structural winners across AI, consumer platforms, and industrials; downgrade generic ETFs and weaker financials; and maintain a layer of health‑care and hard‑asset ballast. The quarter’s 25.51% reported performance suggests the strategy has earned the right to take a bit more idiosyncratic risk — and Citigroup is clearly choosing to spend that risk budget on the AI supply chain.