The above button links to Coinbase. Yahoo Finance is not a broker-dealer or investment adviser and does not offer securities or cryptocurrencies for sale or facilitate trading. Coinbase pays us for certain activity generated through this link. Prices displayed are informational.

The Roundhill Memory ETF (DRAM) has been one of the most remarkable ETF stories of the year. The fund launched in early April and now has nearly $7 billion invested in it a little over a month later, putting it neck and neck with iShares Bitcoin Trust (IBIT) for the fastest-growing ETF launch of all time.I wrote about it shortly after launch and called it one of the smartest ETF launches I've seen.DRAM had two things going for it. The first was a real underlying investment thesis. Memory had become one of the hottest corners of the stock market, driven by booming demand for high bandwidth memory from AI data centers (Micron Technology (MU) is on track to generate profits next year in the same ballpark as Apple and Amazon!)The second factor was that some of the biggest beneficiaries of the boom were difficult for U.S. investors to buy. Two of the three largest HBM producers, Samsung Electronics and SK Hynix, trade in South Korea and aren't held in any major US semiconductor ETFs.Investors had been buying the iShares MSCI South Korea ETF (EWY) to get exposure to those stocks, but that fund also includes dozens of unrelated Korean stocks. DRAM gave investors much cleaner exposure to the theme.That combination—an extremely compelling investment thesis plus the access gap—is what made DRAM so successful.

DRAM's success coincided with the broader rise of the bottleneck trade as a framework for AI investing. The original was GPUs, which turned Nvidia (NVDA) into the most valuable company in the world.Power followed, with utilities, nuclear stocks, gas turbines, and electrical infrastructure all benefiting as data center buildouts started running into grid constraints.Memory, of course, was another, and more recently, CPUs have gotten attention as AI shifts to inferencing from training.Then there's photonics, which has also gained momentum over the past year.

According to Ansys, photonics is "a multidisciplinary domain that involves the generation, control, manipulation, and detection of light."The technology has been used for decades across imaging systems, industrial lasers, fiber-optic telecom networks, medical devices, sensing applications, and displays. Anywhere light needs to be generated, shaped, or detected to do something useful.In AI data centers, photonics shows up as the use of light to move data between chips, racks, and servers, replacing the copper wiring that's hitting its physical limits.As AI clusters scale to hundreds of thousands of GPUs, the bandwidth needed to keep them in sync is overwhelming what copper can handle. Optical interconnects, transceivers, photonic integrated circuits, and co-packaged optics are how the industry is solving that.The capital flowing into the industry reflects how seriously this is being taken. Nvidia has invested billions of dollars into photonics companies to lock down supply, and the stocks have skyrocketed. Applied Optoelectronics (AAOI) is up roughly 440% year to date, while Lumentum (LITE) and Coherent (COHR) have both more than doubled.

The Corgi Lithography & Semiconductor Photonics ETF (EUV) launched last week, the first U.S.-listed ETF to specifically target the photonics theme. The fund has been slow out of the gate, sitting at about $3 million in assets after four days, well below the trajectory DRAM was on at the same point.There could be a number of reasons for that. For one, the photonics winners are mostly already US-listed and tradeable. Stocks like Lumentum, Coherent, and Applied Optoelectronics all trade on US exchanges.There's no Samsung or SK Hynix-style access gap to solve, so demand for an ETF targeting the space is naturally going to be lower.But another issue pertains to the design of EUV itself. The top five holdings as of May 11 are Taiwan Semiconductor (TSM) at 9.5%, ASML Holding (ASML) at 8%, Corning (GLW) at 5.3%, Lam Research (LRCX) at 5%, and Applied Materials (AMAT) at 4.9%.Of those, you could say Corning is the clearest AI photonics play. Its Optical Communications segment is growing rapidly, fueled by AI data center fiber demand, and that segment is the reason investors are buying the stock right now.The rest are general chip and semicap names. They're benefiting from AI, but not from the photonics bottleneck specifically. TSM is benefiting because of demand for AI chips at leading-edge nodes. ASML is benefiting because more of those AI chips need EUV lithography.That's a different cycle from the photonics bottleneck driving Lumentum, Coherent, and Applied Optoelectronics—stocks that sit further down the holdings list.

EUV invests in companies involved with "photonics and light-based technologies, including extreme ultraviolet ('EUV') lithography and related semiconductor manufacturing and inspection tools," according to the fund's prospectus.On some level, the theme makes sense. Lithography is a light-based process. ASML makes the EUV machines that use light to print the most advanced chips, and TSM uses those machines to manufacture the AI chips driving the boom.But while those companies are using light and benefiting from the AI boom, the investment thesis is different from what's driving the photonics names.The AI photonics trade has been about a bottleneck in the data center buildout, a severe demand-overwhelming-supply story for transceivers, optical components, and fiber as hyperscalers race to wire up GPU clusters.The semicap and foundry story is a longer-term capacity build for leading-edge node manufacturing, with AI as one driver among several.Not to mention, those semicap and fab names already have significant exposure in the broad semiconductor ETFs like the VanEck Semiconductor ETF (SMH).TSM, ASML, LRCX, AMAT, and KLA Corp (KLAC) together make up over a quarter of SMH.Someone who wants to play the photonics bottleneck might want something more targeted, the same way DRAM was highly targeted to just memory.

When I tweeted about this recently, several replies came back asking when there would be a pure-play photonics ETF. Well, three other ETFs have been filed that target the space, each with a different take on what it means to be a photonics ETF.The first, the Tuttle Capital Pure Play Photonics ETF, is the one that on paper sounds most like what investors are asking for. Pure play is right in the name. But the potential holdings are way broader than just AI photonics, covering industrial and scientific lasers, LiDAR, infrared and spectrometry, and optoelectronic devices. It also aims to focus on small and mid cap names. So once again, this seems different than the AI infrastructure trade investors are clamoring for.Tuttle does emphasize flexibility, though. The fund is actively managed, and the adviser has explicit discretion to include companies based on subjective factors. So in practice, the portfolio could lean toward AI photonics over time, since that's where the demand is.The second filing, the Tema Optical and Photonic ETF, uses a three-prong test for inclusion. It includes stocks of companies that either generate at least 50% of their revenue from photonics activities, pre-revenue companies whose primary business is photonics equipment, or companies with less than 50% of their revenue from photonics that have publicly disclosed it as a strategic focus.The filing also includes a list of potential company types, most of which are data center focused, but some, like makers of lithography machines and LiDAR sensors, suggest a somewhat broader mandate. The third filing, the KraneShares Optical AI Infrastructure ETF, is the one that most closely matches what investors are seemingly asking for. It uses a strict 50% revenue or gross profit test on four buckets: transceivers and pluggable optics including co-packaged optics, photonic integrated circuits and silicon photonics, optical fiber and assemblies, and optical networking systems. None of the chip and semicap names that dominate EUV's holdings would clear that test. KraneShares' stricter strategy has a tradeoff, however. Names like Corning, where AI photonics drives the bull thesis but isn't a majority of revenue, could potentially fall outside the fund.On the other hand, Tuttle and Tema's flexibility could let them include those same kinds of names if the adviser sees strategic fit. That comes with its own tradeoff, though, since looser standards could also pull in companies where photonics exposure is more incidental.

Like all thematic ETFs, the rules and judgments built into the fund shape the exposure investors end up with.

Though the category has plenty of potential, none of the photonics ETFs are likely to match the success of DRAM. That story wasn't just about the compelling investment thesis. It was also about the access gap, and the photonics trade doesn't have that.Still, investors clearly want exposure to the AI photonics trade in an easy-to-buy ETF. So one or more of these funds could find a real audience, even if not at the scale DRAM has reached.

Permalink | © Copyright 2026 etf.com. All rights reserved