Summary

  • NECK began trading on 23 September with a 0.75% annual expense ratio and a launch brief naming 20 target holdings across five sleeves. Those weights are targets, not an index, and can move at the adviser’s discretion.
  • The prospectus requires at least 80% exposure to broadly defined current AI bottleneck companies and permits up to 20% in businesses expected to become or relieve future shortages. Investors are paying for classification and rotation, not purchasing proof that today’s scarcity will persist.

On NECK’s first trading day, two versions of the portfolio were visible. The product brief offered a crisp launch map: 30% memory, 25% optics, photonics and networking, 17.5% power and infrastructure, 17.5% semiconductors and compute, and 10% future bottlenecks. It named four target companies in each sleeve, 20 in total. The live product page, when reviewed, displayed the headings of a holdings table but no positions; its performance and premium-or-discount fields were still marked TBD.

That contrast needs restraint. An empty web table is not evidence that the fund owned nothing. It is a first-day disclosure gap between a marketing portfolio and an observable portfolio. The gap is useful because it forces the investor back to the legal instrument. Before there is a track record, the 17 September prospectus is the best description of what the buyer has actually delegated.

A shortage is an input; a mandate is the product

The xETFs pitch begins from a defensible industrial observation. AI systems require more than models. They require memory, optical links, switching, foundry capacity, packaging, electricity, cooling and buildings. A missing component can hold up a larger system, while a supplier at that point of constraint may win longer backlogs or better prices. The 23 September launch announcement says NECK is designed to seek those pressure points as they emerge and ease.

But a physical bottleneck is not automatically a profitable security. Five bridges still have to hold. The shortage must be real rather than inferred from an order headline. It must last long enough to influence delivered volume or price. The supplier must retain the benefit after customers, workers and upstream vendors bargain for their share. The equity must not already price in a longer or more lucrative shortage. Finally, the fund must leave before new capacity, technical substitution or weaker demand destroys the scarcity premium.

NECK therefore sells two decisions. The first is classification: which company is sufficiently exposed to a binding constraint? The second is timing: when has the market stopped underestimating that constraint, or when has the constraint itself moved? Both decisions belong to the managers. The word “bottleneck” describes the search field; it does not settle the trade.

The 80% rule is firm at the edge and broad inside it

Under normal conditions, the fund must put at least 80% of net assets, plus borrowings for investment purposes, into securities of “AI Bottleneck Companies” and instruments providing exposure to them. That is a genuine boundary. Changing the policy requires 60 days’ notice to shareholders.

Inside the boundary, the opportunity set is wide. The current definition spans four sub-themes. Memory includes HBM, DRAM, NAND and related storage. Optics, photonics and networking includes lasers, transceivers, switches and interconnect fabrics. Power and infrastructure includes cooling, electrical distribution, generation, grid connection, construction and other data-centre equipment or services. Semiconductors and compute includes processors, accelerators, neoclouds, foundries, packaging, manufacturing tools, materials and components.

The prospectus generally anchors each sub-theme to a 50% test based on revenue, operating profit or attributable assets. That keeps a casual mention of AI from qualifying every industrial company. It does not create a single comparable metric. Revenue, profit and assets answer different questions, and a company can cross the threshold through any of them. A capital-heavy utility supplier and a high-margin optical component maker may both qualify even though scarcity reaches their accounts through different channels.

Exposure may also come through shares, depositary receipts, swaps, options and other derivatives. Total-return swaps count at notional value toward the 80% policy. The fund says these tools may help where direct ownership faces local-market access, foreign-ownership, tax or liquidity constraints. They also make the portfolio harder to read from a list of cash equities alone: economic exposure, collateral, counterparty and notional value can diverge.

The fund expects 15 to 25 companies, significant exposure to Asian issuers, and concentration of at least 25% in an information-technology industry or group. It is also non-diversified, which permits greater issuer concentration than a diversified fund. “One ticker” simplifies the purchase instruction. It does not simplify the underlying geography, settlement, currency, customer concentration or supply-chain risk.

The launch weights describe a thesis, not a constitution

The product brief’s initial target portfolio makes the thesis concrete. Memory’s 30% target names SK hynix, Samsung, Micron and Kioxia. The 25% optical and networking sleeve names Coherent, Lumentum, Marvell and Arista. Power and infrastructure’s 17.5% names Vertiv, Eaton, GE Vernova and Bloom Energy. Another 17.5% goes to ASML, TSMC, Broadcom and CoreWeave under semiconductors and compute. Cerebras, Oklo, Sivers and Everspin occupy the 10% future-bottleneck sleeve.

Those allocations are informative precisely because they are provisional. The brief says the holdings and weights are initial targets and will change with market activity and adviser discretion. The prospectus expects more than 10% in each of memory and optics, but does not freeze the launch weights. A future factsheet could therefore look very different while remaining faithful to the mandate.

This flexibility solves a real problem. Bottlenecks migrate. If memory supply expands, the binding constraint may move to networking, transformers, power generation, permitting or customer deployment. A fixed index risks preserving yesterday’s scarcity. Active management can rotate before a scheduled index review.

Flexibility also removes a simple external answer key. A shareholder cannot evaluate NECK merely by asking whether an issuer appears on a published index. The relevant questions become whether the manager’s definition was consistent, whether the evidence for a constraint was contemporaneous, whether the valuation test was applied and whether exits occurred before or after the shortage became obvious to everyone else.

The 20% sleeve can anticipate scarcity—or manufacture narrative room

Up to 20% of net assets may go into “Future AI Bottleneck Companies.” These may supply a part of the AI chain that is not short today, may be expected to fall into shortage, or may offer a product designed to relieve an existing or expected constraint. The prospectus gives on-site generation, alternative memory and new light sources as examples, while warning that the anticipated shortage may arrive late, fail to arrive or be removed by competing technology.

This sleeve is economically important. Waiting for a shortage to be visible can mean buying after valuation has adjusted. Anticipation is where active management might earn its fee. Yet it is also where the thesis becomes least falsifiable. A company can qualify because its input will become scarce, or because its product may make another input less scarce. One position owns the tollbooth; another finances the bypass.

Neither is inherently wrong. They are different payoffs. A shortage beneficiary needs scarcity to persist. A shortage reliever needs adoption and execution to destroy part of that scarcity. Combining them can diversify the causal path, but the fund should eventually show which role each position plays. Without that classification, “future bottleneck” risks becoming a narrative reserve for any AI-adjacent security the manager likes.

The fee buys judgment before it buys evidence

NECK’s management fee and total annual operating expense ratio are 0.75%. The prospectus illustration puts the cost at $77 for a $10,000 holding after one year and $240 after three years, assuming a 5% annual return and unchanged expenses. Brokerage commissions and intermediary charges sit outside that illustration. Turnover costs do too.

The prospectus says the strategy may produce high turnover, but a new fund has no turnover record. It also has no performance history. At launch, investors cannot measure whether rotation added value, whether taxable distributions were created, or whether the fund outperformed a cheaper collection of the same companies. The 0.75% is paid from the beginning; the evidence arrives later.

That does not make the fee excessive by definition. The fund crosses markets, may use derivatives, and asks managers to compare constraints with different clocks. Memory capacity is measured in fabrication and qualification cycles. Optical supply turns on component yields and customer design wins. Power infrastructure can be governed by multi-year factory capacity, utility queues and permits. Treating all of them as one static factor would be false precision.

The fair test is narrower: does the manager publish enough evidence to show that active classification and rotation are doing work that a rules-based basket could not? Holdings, changes, exposure by sleeve and an explanation of why a bottleneck entered or left the portfolio would make that judgment auditable. A slogan cannot.

Scarcity contains its own expiry mechanism

The deepest risk in the strategy is not simply that the AI buildout slows. It is that the strategy succeeds. High prices, large backlogs and rich equity valuations attract capital. Suppliers add factories, customers redesign systems, rivals enter, and buyers seek substitutes. The bottleneck becomes an investment signal; investment then helps remove the bottleneck.

That feedback is healthy for the industry and dangerous for an undisciplined scarcity trade. A company can report a record backlog while the marginal order is already weakening. Capacity can remain sold out while investors have capitalised several years of favourable pricing. A technology can relieve one constraint by shifting cost to another layer. The manager must distinguish an engineering constraint from an equity mispricing, then update both faster than the market.

NECK is thus a wager on institutional learning. The managers need to observe how scarcity moves across the stack and be willing to abandon the story that attracted the assets. If they cannot, “periodically rebalanced” becomes a way to keep the theme current after the original claim has expired.