Roundhill Photonics & Optics ETF ($LYTE): the first concentrated US bet on optical interconnect

Roundhill filed a registration statement on May 20, 2026 for a new thematic ETF - the Roundhill Photonics & Optics ETF, ticker $LYTE. Concentrated holdings, single sleeve, no diversification overlay. It's the first US-listed exchange-traded fund built specifically around the photonics layer of AI infrastructure, and one of the concentrated sleeves on Roundhill's foundational-AI-layers shelf (next to $DRAM for memory and $NCLD for neoclouds).

The filing isn't effective yet - no expense ratio, no NAV, no holdings disclosed - but the prospectus framing ("concentrated", "Photonics & Optics") plus Roundhill's prior playbook on $CHAT, $HUMN, and $WTAI tells you most of what the basket will look like. This piece walks through what $LYTE is, why the photonics theme is a structurally interesting AI-infrastructure trade in 2026, which stocks are likely to be in scope, and how the new fund compares to the existing thematic-AI ETF universe.

What is the Roundhill Photonics & Optics ETF (LYTE)?

The Roundhill Photonics & Optics ETF ($LYTE) is a concentrated, actively-managed thematic ETF that will hold US-listed equities tied to photonics and optical technologies. Roundhill's standard concentrated-thematic format means roughly 25-40 holdings, weighted toward pure-plays rather than mega-cap diversifiers - closer in shape to $CHAT (generative-AI) or $HUMN (humanoid robotics) than to $SOXX or $SMH.

"Photonics" in the prospectus context covers any company materially exposed to:

• Optical transceivers - pluggable modules that convert electrical signals to light inside datacenter racks (400G, 800G, 1.6T)
• Silicon photonics - chips that integrate optical components onto silicon dies, enabling co-packaged optics with switches and accelerators
• Photonic components - lasers, modulators, photodetectors, optical fibers, packaging
• Photonic instruments and test equipment - burn-in, characterization, wafer-level optical test
• Adjacent enablers - connectivity ASICs, retimers, DSPs, and photonic-electronic co-design tooling

This is a narrow thematic in absolute terms - the global photonics-pure-play universe is perhaps 20-30 listed names - which is exactly why a concentrated wrapper makes sense. A diversified ETF can't get you clean exposure to this slice without diluting it with $NVDA, $AVGO, and $TSM. $LYTE, by design, will not.

Why photonics matters for AI right now

The structural pitch: optical interconnect is the next bandwidth bottleneck after HBM.

The 2023-2025 AI capex cycle was carried by GPU compute ($NVDA) and high-bandwidth memory ($MU, SK Hynix, Samsung). Both of those layers are now scaling in the price/wafer/yield plane the industry knows how to scale. The next bottleneck - the one the hyperscalers are openly worried about - is moving data between accelerators, both inside a rack (chip-to-chip) and between racks (top-of-rack to spine).

Copper interconnect runs into hard physics:

• Above ~200 Gbps/lane copper traces lose signal integrity over centimeter-scale distances
• Power per bit transmitted scales unfavorably above 1.6T per port
• Cable bulk and connector density become physical-design constraints in 100kW+ AI racks

Optical interconnect - moving bits as photons over fiber - sidesteps all three. The transition has been on the roadmap for a decade; the AI capex cycle has compressed the timeline. NVIDIA's 2024-2025 announcements around co-packaged optics, Broadcom's CPO switches, Marvell's connectivity portfolio, and the rapid 800G → 1.6T transceiver ramp are all the same story: the interconnect layer is being rebuilt around photonics now, not in 2030.

That's the thesis $LYTE will sell. It's also empirically visible in the price action of the underlying names through 2025-2026.

Likely $LYTE holdings

Roundhill hasn't disclosed the basket. But QuantAbundance tracks 16 photonics-theme stocks on /themes/photonics, and the Roundhill index methodology + the existing photonics ETF universe (THNQ, ARKQ partial exposure, IPAY etc.) constrain the candidate list. Probable holdings, split by role:

The core five - $LITE, $COHR, $AAOI, $FN, $POET - are the most exposed pure-plays and the most likely to anchor the index at high weights. $MRVL and $ALAB are connectivity-not-photonics names, so Roundhill may include them at trimmed weights or skip them entirely depending on how strictly the index methodology defines "photonics".

What we do not expect in $LYTE: $NVDA, $AVGO, or $TSM. Those names are too large to be carried as pure-photonics exposure - the entire concentrated-thematic format exists precisely to give an investor the photonics signal without diluting it with the AI mega-caps.

How $LYTE compares to existing AI-thematic ETFs

The closest comparables on the US tape today:

The differentiation: $LYTE will be the only US-listed ETF where photonics is the whole basket rather than an incidental slice of a broader semi/AI fund. The closest analog in QA's coverage is the relationship between $QTUM and the Quantum bubble - a single-thesis vehicle that lets an investor express a view on one bottleneck layer without trading the names individually.

This is also why $LYTE is interesting from a flow-rotation standpoint. When an AI-infrastructure narrative shifts - say, from compute (NVDA-led) to interconnect (LITE/COHR/AAOI-led) - the rotation is currently expressed by selling NVDA and buying the photonics names individually. A single ticker that aggregates that rotation makes the trade easier and, if it gathers AUM, makes the rotation itself more violent.

What to track until $LYTE launches

The filing-to-launch interval for Roundhill ETFs has typically been 60-120 days. Concrete checkpoints:

1. SEC effectiveness notice - when the 485 amendment posts as effective, $LYTE can list. Watch the Roundhill press feed.
2. Disclosed expense ratio - Roundhill thematic peers run 50-75 bps. Anything materially higher would be a flag.
3. Listing exchange and inception date - most Roundhill funds list on NYSE Arca or Nasdaq.
4. Day-one holdings disclosure - will confirm which photonics names made the cut at which weights.
5. The first 30 days of net creations/redemptions - early flow tells you whether the marketing reached the audience that already trades $LITE/$COHR/$AAOI individually.

We'll track all five on the QuantAbundance ETFs page as the data lands. The 11 currently-live Roundhill funds are already on-platform with AUM, expense ratio, and 30-day net flow where coverage exists.

Risks and the bear case

Three structural risks before getting long:

1. The photonics universe is small. A 25-name concentrated ETF in a 30-listed-name universe is functionally the universe. Liquidity and crowding risk are real - if $LYTE gathers $500M+ AUM, it becomes a meaningful price-setter in some of the smaller names ($AAOI, $POET, $LASR float is modest).
2. Cyclicality. Optical-transceiver demand tracks hyperscaler capex with a lag. A capex pause - voluntary or otherwise - would compress earnings across the basket simultaneously. The same concentration that makes $LYTE useful on the way up makes it punishing on the way down.
3. Substitution risk at the architecture level. Linear pluggable optics (LPO), co-packaged optics, and emerging optical-switch architectures redistribute economics across the basket unevenly. The names that win the next-gen interconnect race are not necessarily the same names that own the current 800G socket.

The bull case rests on the structural thesis: if AI rack-scale buildout continues, optical interconnect is non-substitutable, and the photonics names own a disproportionate share of the bottleneck. The bear case rests on either the buildout pausing or the bottleneck moving somewhere else (back to memory, forward to optical compute) faster than the basket can rotate.

Where this fits in the QA bubble taxonomy

Within QuantAbundance's bubble framework, $LYTE maps cleanly onto the Networking / Optical bubble - the bloc we've been tracking as the residualized-correlation peer of Memory / HBM and a layer-down complement to AI Compute Accelerators. It's also the canonical ETF wrapper for the Photonics theme, which currently lists 16 stocks across the optical-AI value chain.

If you want context on why this layer of the AI stack matters at all, the Datacenter Power bubble piece covers the power-and-thermals leg of the same buildout, and the 12 AI bubbles ranked by empirical realness places photonics inside the broader supercycle taxonomy.

$LYTE is, in short, the cleanest single-vehicle expression of "I think optical interconnect is the next bottleneck" that the US market has ever offered. Whether that thesis is right is a separate question; that the thesis is now ETF-wrappable is the structural fact, and Roundhill, as usual, was first to file.

The full Roundhill lineup - AUM, expense ratio, 30-day net flow, bubble mapping - is tracked live on /etfs. We'll add $LYTE to the catalog the day the listing prints.

To hold $LYTE or its optical constituents from a US-retail or LLC account, see /stack/ibkr. Bubble shifts and rule-based alerts on the networking-optical cohort are part of /pro.