Alibaba ASML Split Marks Greenwoods Asset Management's Retreat From Tech Giants

Greenwoods Asset Management cut its reported U.S. equity portfolio by 43.64%, creating a stark Alibaba ASML split in its second-quarter holdings.

The Hong Kong manager exited Alibaba, Meta, Amazon, and Nvidia while opening positions in ASML and Applied Materials. It also added Applied Optoelectronics and Chinese data center operator VNET.

That is more than a routine reshuffle among technology stocks. Greenwoods moved away from companies selling platforms, cloud services, advertising, and finished processors. It instead bought businesses supplying the equipment and physical connections behind expanding computing capacity.

The disclosed portfolio fell from $3.878 billion at the end of March to $2.186 billion on June 30. That $1.692 billion contraction suggests capital preservation mattered as much as the new positions.

The central tension is therefore not Alibaba versus ASML as operating companies. It is ownership of AI demand versus ownership of the equipment required to serve that demand.

Greenwoods Cut More Than Its Largest Technology Positions

The defining second-quarter action was a major reduction in reported U.S. exposure, followed by a selective move into infrastructure suppliers.

Greenwoods Asset Management Hong Kong filed its quarterly holdings report on August 7. The quarterly filing covers securities held on June 30, not positions held on the filing date.

The reported portfolio contracted by $1.692 billion from the previous quarter. Its 43.64% decline cannot be explained by one or two stock-price changes alone.

Greenwoods fully exited Meta, Amazon, Nvidia, Alibaba, New Oriental, TAL Education, UnitedHealth, and Lumentum, according to the disclosed changes. The exits crossed several sectors, including internet platforms, education, health insurance, graphics processors, and optical components.

That breadth matters. A manager simply changing its preferred AI stock might sell Nvidia and buy another chip designer. Greenwoods instead reduced its overall book while replacing only selected parts of the technology exposure.

The first-quarter portfolio had already shown caution around prominent AI beneficiaries. Greenwoods reduced Meta and Nvidia during that period, while initiating Amazon and adding heavily to Intel.

The second quarter completed several of those moves. Amazon disappeared after one quarter, while Meta and Nvidia went from reduced positions to full exits.

Alibaba followed a different path but reached the same destination. The company combines commerce, cloud computing, AI models, and extensive infrastructure spending. Greenwoods nevertheless removed its U.S.-listed shares from the reported portfolio.

The filing does not disclose purchase prices, sale dates, hedges, or the manager's reasoning. It also excludes many securities that do not appear on the Form 13F list.

That limitation is crucial. A 13F shows certain long U.S.-listed securities at quarter-end. It does not provide a complete account of a manager's global exposure, cash, short positions, swaps, or Hong Kong-listed holdings.

Greenwoods could retain economic exposure to some companies through instruments outside this disclosure. It could also own their Hong Kong shares through accounts not captured in this filing.

Still, the direction inside the disclosed portfolio is unusually clear. Greenwoods reduced its long U.S. equity book and removed multiple large technology names simultaneously.

The new purchases were much narrower. ASML supplies lithography systems used to print chip patterns. Applied Materials sells equipment for deposition, materials engineering, and other fabrication processes.

Applied Optoelectronics supplies optical networking products used to move data. VNET operates carrier-neutral data centers, facilities where customers can connect through multiple telecommunications providers.

These additions cover four physical layers: chipmaking tools, semiconductor processes, optical links, and data center capacity. That combination creates the article's real reversal.

Greenwoods did not abandon technology altogether. It moved down the supply chain.

Why the Alibaba ASML Shift Matters Now

The Alibaba ASML rotation transfers exposure from companies funding AI expansion to businesses that receive orders when physical capacity expands.

Alibaba illustrates the demand-owner side of the trade. It operates cloud services, develops Qwen models, and spends heavily to build computing infrastructure.

Its Cloud Intelligence Group generated 41.6 billion yuan in the March quarter, up 38% from one year earlier. AI-related product revenue had recorded triple-digit growth for an eleventh consecutive quarter.

Those figures show that Greenwoods did not exit Alibaba after its AI business stopped growing. It sold the reported position while the cloud operation was accelerating.

Alibaba's broader economics were less straightforward. Group revenue grew only 3% to 243 billion yuan in the March quarter. Operating results also absorbed higher investment across AI infrastructure and consumer businesses.

This creates a familiar problem for platform owners. Strong AI demand can support revenue while the required spending pressures near-term cash generation and margins.

ASML occupies the other side of that spending decision. A chip manufacturer expanding advanced capacity needs lithography equipment before it can produce additional processors or memory.

Extreme ultraviolet lithography, commonly called EUV, uses short-wavelength light to print extremely small circuit patterns. ASML remains the only commercial supplier of EUV systems.

The company's second-quarter results strengthened the infrastructure thesis. ASML reported €9.3 billion in net sales and €2.9 billion in net income for the period ended June 30.

It also raised its 2026 sales outlook to between €43 billion and €45 billion. Its updated outlook cited customers accelerating capacity plans tied to AI demand.

Applied Materials provides a broader equipment exposure. Its tools support logic, memory, and advanced packaging, which combines multiple components into higher-performance chip systems.

The company reported fiscal second-quarter revenue of about $7.9 billion. Its equipment results included a 49.9% gross margin and $2.52 billion in operating income.

Management directly connected demand with the global buildout of AI computing infrastructure. That statement is a company view, but the reported revenue and profitability provide measurable support.

The rotation therefore separates two different ways to invest in the same capital cycle.

Alibaba and other platform operators must convert expensive capacity into durable customer revenue. ASML and Applied Materials receive orders earlier, when manufacturers decide to build that capacity.

This does not make equipment suppliers low-risk. Their orders remain cyclical, concentrated among a few major customers, and vulnerable to export controls.

It does change the immediate question. For a cloud platform, investors ask whether AI revenue will eventually justify spending. For an equipment supplier, investors first ask whether spending plans will become machine orders.

Greenwoods appears to have preferred the second question at the June 30 snapshot. The position changes put capital closer to announced construction and fabrication budgets.

That shift also reduced direct exposure to consumer demand, advertising markets, and model-level competition. ASML does not need to predict which chatbot attracts the most users.

Its customers still need confidence in long-term semiconductor demand. However, orders can reflect capacity plans across foundries, memory producers, and multiple chip designers.

The Alibaba ASML contrast is therefore about where Greenwoods wants to encounter uncertainty. It exchanged product adoption risk for capital-cycle, customer-concentration, and geopolitical risk.

The Portfolio Moved From AI Winners to AI Bottlenecks

Greenwoods targeted bottlenecks that must expand before additional computing power can reach cloud customers.

Nvidia, Meta, Amazon, and Alibaba represent different parts of the technology market. Yet each sits relatively close to chips, applications, or services already delivered to end users.

Nvidia designs accelerators and sells computing platforms. Meta and Amazon operate vast infrastructure while monetizing advertising, commerce, and cloud services. Alibaba combines commerce, cloud infrastructure, and AI development.

ASML and Applied Materials sit further upstream. Their systems determine how much advanced manufacturing capacity chipmakers can install and how efficiently they can produce complex devices.

Applied Optoelectronics targets another constraint. AI clusters contain thousands of processors that must exchange data quickly, making optical connections increasingly important inside and between data centers.

An optical transceiver converts electrical signals into light and back again. Faster transceivers help servers transmit more data without relying only on short copper connections.

Applied Optoelectronics said after its first quarter that it expected an 800-gigabit product ramp beginning in the second quarter. The company anticipated stronger growth after additional manufacturing capacity came online.

Its optical roadmap remained a forward-looking company projection when Greenwoods established its position. Execution still needed to follow.

The addition of VNET extends the same logic from components to facilities. Data centers need land, power, cooling, network access, and operating expertise before customers can deploy servers.

VNET describes itself as a carrier-neutral provider in China. Carrier neutrality lets customers obtain connectivity from different network operators rather than depending on one carrier.

That structure can appeal to cloud providers and enterprises requiring flexible connections. However, it also exposes VNET to capital requirements, financing conditions, and utilization risk.

Its first-quarter reporting showed a 24.2% utilization rate for ramp-up retail capacity. Ramp-up capacity refers to recently opened space still attracting customers.

That figure reveals both opportunity and risk. Unused capacity can support future revenue, but it produces weaker economics until customer deployments increase.

VNET's capacity update gives investors a measurable indicator for the infrastructure thesis. Rising utilization would show that physical capacity is meeting real demand.

Together, ASML, Applied Materials, Applied Optoelectronics, and VNET form a coherent sequence.

Chipmakers need manufacturing tools. Accelerators need high-speed optical connections. Cloud operators need powered data center space where the systems can run.

This is not a complete AI supply chain. Greenwoods did not disclose new stakes in every category, such as memory, electricity generation, cooling, or networking silicon.

The selection nevertheless concentrates on assets whose demand depends on capacity installation. That differs from holding a broad group of technology leaders.

The simultaneous exit from Lumentum adds an important complication. Lumentum also supplies optical and photonic products, so Greenwoods did not simply buy every optical networking company.

Replacing Lumentum with Applied Optoelectronics suggests company selection still mattered within the infrastructure theme. The filing cannot explain whether valuation, customers, product timing, or position size drove the choice.

Greenwoods also exited Nvidia despite the chipmaker's central role in AI infrastructure. That indicates the manager did not treat every hardware beneficiary equally.

Nvidia captures demand through accelerators, networking, software, and complete systems. Its position is closer to the finished computing platform and carries expectations reflecting that leadership.

Equipment suppliers receive capital before newly planned fabrication capacity produces finished chips. Optical suppliers and data center operators can benefit as those chips enter deployed systems.

The portfolio therefore moved toward bottlenecks, but not toward hardware indiscriminately. It favored selected enablers around fabrication, transmission, and hosting.

The Filing Does Not Prove Greenwoods Abandoned Alibaba

A 13F is a delayed portfolio snapshot, not a live declaration of strategy or a complete map of Greenwoods' assets.

The most tempting interpretation is that Greenwoods lost confidence in Alibaba and the largest U.S. technology companies. The evidence does not support that conclusion by itself.

The filing shows reportable long positions held on June 30. Greenwoods submitted it on August 7, leaving more than five weeks when positions could have changed.

It does not show trades completed after quarter-end. It also does not identify the prices at which Greenwoods bought or sold securities.

The reported market value combines trading activity with market-price changes. Comparing $3.878 billion and $2.186 billion therefore establishes portfolio contraction, but not the exact cash raised through sales.

The disclosure also covers only designated U.S.-listed securities. Alibaba trades in both New York and Hong Kong, creating an especially important reporting gap.

An exit from Alibaba's American depositary shares does not establish that Greenwoods eliminated every Alibaba position. The manager might own Hong Kong-listed shares outside this specific filing.

American depositary shares represent foreign-company equity traded in the United States. Their removal from a 13F only proves that the reportable U.S. position was absent at quarter-end.

The same caution applies to hedging. A long position can form one side of a more complex trade involving options, swaps, or short exposure not fully visible here.

Greenwoods provided no public investment letter explaining the second-quarter changes. Any claim about its precise motivation remains an inference from disclosed holdings.

There is also no evidence that the new positions are permanent. ASML, Applied Materials, Applied Optoelectronics, and VNET were new at one quarter-end snapshot.

A manager can establish a position for valuation, event timing, or portfolio balancing without making a multi-year industry call. The next filing might show additions, reductions, or complete exits.

The phrase "cash rotation" therefore needs care. The large portfolio reduction suggests Greenwoods raised liquidity inside the disclosed book, but Form 13F does not report cash balances.

Nor does the filing show that Greenwoods moved every dollar sold from technology giants into the four new infrastructure names. The reported portfolio shrank far more than the new positions could offset.

That asymmetry is central to the story. Capital preservation appears at least as important as sector rotation.

The Alibaba ASML narrative can also overstate a direct comparison. Alibaba sells cloud capacity and invests in AI infrastructure, while ASML sells manufacturing equipment to chipmakers.

They are not conventional competitors. Their economic exposures occupy different stages of a shared investment cycle.

Export controls create another source of uncertainty. ASML faces restrictions on shipping certain advanced equipment to China, and policy changes can affect customer demand and product mix.

Applied Materials faces related regulatory exposure. Its customers can delay orders when trade policy, fab construction, or semiconductor demand becomes uncertain.

Applied Optoelectronics carries execution and customer-concentration risks. A forecast product ramp matters only when production, qualification, and customer orders follow.

VNET must translate demand into occupied data center capacity. New facilities can depress returns when power, construction, and financing costs arrive before customer revenue.

These risks are different from the pressures facing Meta, Amazon, Nvidia, or Alibaba. They are not automatically smaller.

Greenwoods exchanged exposure to high-profile technology leaders for companies with more direct operational bottlenecks. It did not exchange uncertainty for certainty.

Alibaba and ASML Still Depend on the Same AI Spending Cycle

The rotation changes where Greenwoods sits in the cycle, but every selected company still depends on sustained infrastructure spending.

Alibaba needs computing capacity to train models, run inference, and serve cloud customers. Inference is the process of using a trained model to generate an answer or prediction.

ASML and Applied Materials benefit when chip manufacturers expand production to meet that demand. Applied Optoelectronics benefits when deployed systems require faster connections.

VNET benefits when Chinese cloud and enterprise customers occupy additional data center space. Each company depends on a different purchasing decision, but those decisions remain connected.

The chain begins with expected usage. Cloud operators estimate future demand and commit capital to servers and facilities.

Chip designers translate that demand into processor and networking orders. Foundries and memory manufacturers then decide whether existing production can handle the expected volume.

Equipment orders follow when manufacturers need new or upgraded fabrication lines. Data center and optical spending arrives as servers move toward deployment.

This sequence creates timing differences. ASML can report strong orders while cloud companies still face questions about return on invested capital.

It can also reverse. If cloud demand disappoints, manufacturers can delay capacity, equipment orders can weaken, and suppliers can experience abrupt revenue cycles.

ASML's latest results showed the favorable side of that mechanism. Its €9.3 billion quarterly sales and higher annual outlook indicated firm customer demand through June.

Applied Materials also connected its performance with investment in advanced logic, DRAM memory, and advanced packaging. These categories support AI accelerators and their surrounding systems.

Alibaba supplied evidence from the demand side. Its cloud growth and repeated AI-product expansion showed real customer activity, even as group-level spending weighed on operating results.

That means the Greenwoods exits should not be read as proof that AI demand is collapsing. The new holdings actually require continued demand.

The portfolio says something narrower. Greenwoods preferred selected suppliers whose revenue connects with physical expansion while sharply reducing its total reported exposure.

This distinction matters for investors comparing Alibaba and ASML. A platform can produce faster growth but face heavier reinvestment and product competition.

An equipment vendor can hold a narrower technical position but face order volatility. Its results can also depend heavily on a small group of semiconductor manufacturers.

A data center operator can own scarce power and space but struggle with financing or low utilization. An optical supplier can serve a growing market but lose share during product transitions.

No layer captures the entire economics of AI. Value can move between model providers, cloud operators, chip designers, foundries, equipment companies, and infrastructure owners.

Greenwoods' filing captures one attempt to reposition within that contest. It favors the enabling layers at one specific moment.

The exits also reduce exposure to several unrelated earnings drivers. Meta depends heavily on advertising, Amazon spans retail and cloud services, and Alibaba combines commerce with cloud computing.

ASML and Applied Materials offer cleaner semiconductor capital-spending exposure. That focus can improve upside when fabrication investment rises, but it reduces diversification across business models.

The Alibaba ASML split should therefore be viewed as a change in risk location. Greenwoods shifted from monetization and market-share questions toward capacity, orders, regulation, and utilization.

That is a meaningful reversal, but not a one-way bet on infrastructure.

Three Signals Will Test the Infrastructure Rotation

The next evidence must show that equipment orders, optical deployments, and occupied data center capacity can support the portfolio's new direction.

The first signal is Greenwoods' next Form 13F, covering positions held on September 30. It will reveal whether the manager expanded or reversed its new holdings.

Additional purchases of ASML and Applied Materials would strengthen the case for a sustained move upstream. Reductions after one quarter would suggest a tactical trade or rapid reassessment.

The filing will also show whether Greenwoods rebuilt positions in Alibaba, Meta, Amazon, or Nvidia. Any return would complicate the current interpretation of a broad retreat.

The second signal is supplier execution. ASML has raised its 2026 sales outlook, so bookings, shipments, and customer capacity plans must support that confidence.

Applied Materials must convert demand across logic, memory, and advanced packaging into continued operating performance. Order timing will matter as customers adjust fabrication budgets.

Applied Optoelectronics faces a more specific test. Its projected 800-gigabit ramp must produce measurable shipment and revenue progress after new capacity comes online.

A delay would weaken the idea that Greenwoods successfully targeted an optical bottleneck. Successful qualification and volume growth would strengthen it.

The third signal is infrastructure utilization, especially at VNET. New data center capacity creates value only when customers deploy equipment and begin paying for occupied space.

VNET's ramp-up retail utilization was 24.2% at the end of March. Subsequent results must show whether customer demand is absorbing that available capacity.

Alibaba's next results provide useful demand-side context. Continued cloud and AI growth would support the infrastructure cycle even after Greenwoods exited its reported U.S. position.

Weaker growth combined with heavy capital spending would strengthen the manager's apparent preference for upstream suppliers. Strong growth and improving returns would challenge that choice.

Readers should resist turning one filing into a universal recommendation. The document is most useful as evidence of how a major manager redistributed visible risk.

Greenwoods cut a large portion of its reported U.S. book before selecting a small group of infrastructure companies. That order matters.

It suggests the firm first reduced exposure, then chose specific points where AI spending meets physical constraints. This was not a simple exchange of Alibaba for ASML.

For technology leaders, the portfolio applies pressure to prove that capital spending creates durable earnings. For suppliers, it creates a different obligation to deliver capacity without losing pricing or utilization.

The next quarter will show whether this Alibaba ASML divide becomes a durable portfolio structure. Watch the filing, supplier execution, and occupied infrastructure together.

Those three signals will distinguish a long-term supply-chain thesis from a temporary move into cash and selected hardware names.