xETFs AI Bottlenecks ETF · AMEX:NECK
xETFs AI Bottlenecks ETF — Investor Presentation (September 2026)
28 slides from xETFs AI Bottlenecks ETF's investor presentation, September 2026
Slide 1: NECK: AI's Critical Constraints. One Ticker.
NECK xETFs AI Bottlenecks ETF AI’s Critical Constraints. One Ticker. September 2026 © 2026 WallStreetX ETFs, Inc.
Slide 2: Disclaimer
Disclaimer Carefully consider the Funds’ investment objectives, risk factors, charges and expenses before investing. This and additional information can be found in the Funds’ Prospectus and Summary Prospectus, which may be obtained by visiting www.xETFs.com/NECK. Read the Prospectus and Summary Pro
Slide 3: Disclaimer Continued
Disclaimer Continued Concentration Risk. The Fund is susceptible to an increased risk of loss, including losses due to adverse events that affect the Fund’s investments more than the market as a whole, to the extent that the Fund’s investments are concentrated in the securities and/or other assets o
Slide 4: AI Is No Longer a Sector. It Is Embedded in the Market.
AI Is No Longer a Sector. It Is Embedded in the Market. The largest companies are essential to spending on Artificial Intelligence ("AI") infrastructure. The ten largest U.S. companies are all technology related The ten largest U.S. companies by market capitalization are all technology ~53% business
Slide 5: Capital is Committed to Meet the Demand for AI
Capital is Committed to Meet the Demand for AI The buyers of AI hardware are funding the physical buildout directly, and NVIDIA is funding some of its own suppliers. $500bn+ ₩1,100tn ~$700bn+ $60-64bn NVIDIA-partnered financing SK hynix long-term Korean 2026 hyperscaler capex TSMC 2026 capital budge
Slide 6: What It Takes to Run an AI Model
What It Takes to Run an AI Model Every AI token generated is read out of memory, moved across a network, paid for in electricity, and only then computed. 1 2 3 4 The model has to sit in The chips connect and The data center has to And only then does it memory communicate power it compute An AI model
Slide 7: Demand Compounds. Supply Is Physical.
Demand Compounds. Supply Is Physical. Across multiple layers of AI inputs, indications of constrained capacity. MEMORY OPTICS & NETWORKING “We still forecast that customer demand will remain higher 800G optical transceivers supply is expected to fall 40% - 60% than our supply capacity even beyond 20
Slide 8: Multiple Physical Constraints, One Portfolio
Multiple Physical Constraints, One Portfolio NECK will focus on four key areas of constraints for 80%+ of the portfolio and will also invest in potential future bottlenecks MEMORY OPTICS, PHOTONICS, POWER & SEMICONDUCTORS POTENTIAL FUTURE & NETWORKING INFRASTRUCTURE & COMPUTE BOTTLENECKS Holds the m
Slide 9: What Makes Something a Bottleneck?
What Makes Something a Bottleneck? Four assessments. Being critical is not enough. Demand exceeds supply Pricing power Sold-out capacity, order backlogs, allocation, multi-year lead Tight supply supports higher prices, and those gains translate into times and lengthening qualification cycles. strong
Slide 10: Get to Know NECK
PART TWO Get to Know NECK An actively managed ETF holding 20 companies at launch across the layers where AI demand meets a physical limit.
Slide 11: Introducing NECK
Introducing NECK xETFs AI Bottlenecks ETF Fund Snapshot What NECK Offers Fund ticker NECK Investment objective Capital appreciation • Focused exposure to key constraint layers: memory; optics, photonics & networking; power & infrastructure; semiconductors & Structure U.S.-listed, actively managed ET
Slide 12: Multiple Bottlenecks. One Ticker.
Multiple Bottlenecks. One Ticker. NECK seeks exposure across the AI infrastructure stack. Areas of portfolio allocation at launch include: % of fund by holding; grouping totals below MEMORY OPTICS, PHOTONICS POWER & SEMICONDUCTORS FUTURE & NETWORKING INFRASTRUCTURE & COMPUTE BOTTLENECKS HBM, DRAM &
Slide 13: The Portfolio at Launch: 20 Names
The Portfolio at Launch: 20 Names Grouped by constraint, then ordered by weight within each. Company Ticker Weight Why It Is in NECK MEMORY 30.0% SK hynix 000660 KS 10.0% HBM and dynamic random-access memory (DRAM) leader; NVIDIA’s lead memory partner Samsung Electronics 005930 KS 8.5% The only firm
Slide 14: Memory: The Wall Every Model Hits
Memory: The Wall Every Model Hits The binding constraint on AI performance has moved from processing speed to memory bandwidth and capacity. A fast chip with slow memory is a slow chip 3.6x Modern accelerators can perform arithmetic far faster than memory can feed them data, so a chip spends much of
Slide 15: Memory is a Scarce Resource
Memory is a Scarce Resource Memory scarcity is already visible in reported earnings, not just in order books. 557% YoY 85% Over 60% $38.8bn SK hynix Q2 2026 operating Micron gross margin Kioxia annual cap 2026 DRAM equipment sales income change vs Q2 2025 ex spending vs 2025 On YOY Revenue increase
Slide 16: Memory: Initial Target 30% of NECK
Memory: Initial Target 30% of NECK Three dynamic random-access memory (DRAM) makers and the NAND flash memory (NAND) pure-play, chosen for how directly memory scarcity reaches earnings. 10.0% SK hynix 8.5% Samsung Electronics 000660 SK · HBM & DRAM 005930 KS · Memory, foundry & devices The purest ex
Slide 17: Optics, Photonics & Networking: Why AI Clusters Need It
Optics, Photonics & Networking: Why AI Clusters Need It Past a few meters, copper cannot carry the data rates AI networking now runs at. Every connection beyond the rack becomes optical. At 800G and 1.6T per lane, copper runs out at about three meters COPPER usable signal degrades: too much loss and
Slide 18: Inside the Optical Supply Chain Constraint
Inside the Optical Supply Chain Constraint Constraints exist at multiple points in the chain, and no single stage relieves them all Indium phosphide wafer Laser die Optical engine Transceiver module Switch & fabric The specialist crystal a laser The light source itself. A Laser, modulator and detect
Slide 19: Optics, Photonics & Networking: Initial Target 25% of NECK
Optics, Photonics & Networking: Initial Target 25% of NECK From the laser die to the fabric, four companies that carry the signal between accelerators. Coherent Corp Lumentum Holdings 8.0% 8.0% COHR US · Lasers & optical components LITE US · Transceivers & laser chips Vertically integrated from indi
Slide 20: Power & Infrastructure: One of the Hardest Resources to Buy
Power & Infrastructure: One of the Hardest Resources to Buy A finished data center with no connection to the grid is useless. The scarce asset is a delivery date. 2030-2031 ~945 TWh 9-17% 0.3% Order pipeline for new gas- Estimated global data-center Esitmates for data centers’ Northern Virginia vaca
Slide 21: What Actually Delivers the Power to AI
What Actually Delivers the Power to AI A megawatt reaches a chip through multiple methods, and NECK seeks exposure to these. Held in NECK Dispatchable generation at utility scale. Slots are booked years out: As of Q2 2026, 116 Gigawatts of gas Gas turbines backlog including slot reservations at GE V
Slide 22: Power & Infrastructure: Initial Target 17.5% of NECK
Power & Infrastructure: Initial Target 17.5% of NECK Four companies standing between a signed data center lease and an AI running chip. 4.5% Vertiv Holdings 4.5% Eaton Corp VRT US · Power & thermal systems ETN US · Electrical distribution & switchgear A near pure-play on digital infrastructure: powe
Slide 23: Semiconductors & Compute: AI Runs on Leading-Edge Silicon
Semiconductors & Compute: AI Runs on Leading-Edge Silicon One lithography supplier, one foundry with nearly all leading-edge AI volume, and a packaging step that is currently limiting customers. One supplier for the hardest step One foundry for the leading edge And packaging may be what runs out fir
Slide 24: From a Design to a Working Rack
From a Design to a Working Rack Four steps between an idea and delivered compute, each controlled by a handful of suppliers. STEP 1 STEP 2 STEP 3 STEP 4 Design the chip for AI Print the logic Join it together Build the rack Broadcom co-designs custom ASML’s machines print the Logic and memory are bo
Slide 25: Semiconductors & Compute: Initial Target 17.5% of NECK
Semiconductors & Compute: Initial Target 17.5% of NECK The tool, the fab, the custom chip and the rented capacity. 6.5% ASML Holding 4.5% TSMC ASML US · EUV lithography TSM US · Foundry & CoWoS packaging There is no second source for EUV. Q2 2026 net sales of €9.3bn with FY2026 No credible alternati
Slide 26: Potential Future Bottlenecks: Cerebras & Oklo
Potential Future Bottlenecks: Cerebras & Oklo Not every constraint gets relieved by building more of the same thing. Some could get routed around. Cerebras Systems 3.5% Oklo Inc 2.5% CBRS US OKLO US Routes around: wiring thousands of chips together Routes around: waiting years for a grid connection
Slide 27: Potential Future Bottlenecks: Sivers & Everspin
Potential Future Bottlenecks: Sivers & Everspin Not every constraint gets relieved by building more of the same thing. Some could get routed around. Sivers Semiconductors 2.5% Everspin Technologies 1.5% SIVE SS MRAM US Routes around: the shortage of laser chips Routes around: the trade-off between t
Slide 28: AI's Bottlenecks. One Ticker.
AI's Bottlenecks. One Ticker. NECK · xETFs AI Bottlenecks ETF · NYSE September 2026 © 2026 WallStreetX ETFs, Inc.
xETFs AI Bottlenecks ETF — Investor Presentation (September 2026) on 6ix