LAZR: The ETF That Invests in the AI's Optical Infrastructure - and Also Holds a Slice of Anthropic

On June 30 a new ETF launched on the New York Stock Exchange called the Tema Photonics & Optical ETF, trading under the ticker LAZR. It deals with a niche we have covered here a fair amount in recent weeks - the optical infrastructure that moves data in the world of AI - and it also contains a component drawing attention: an exposure of a little over 11% to Anthropic, one of the largest private AI companies in the world.

(On the ticker: LAZR was previously the ticker of Luminar, a lidar maker that entered insolvency proceedings in December 2025, was delisted from Nasdaq and moved to over-the-counter trading under LAZRQ. The ticker was freed, and the ETF received it. There is no connection between the two beyond the four letters.)

What Photonics Is - and Why It Relates to AI

Photonics is simply the engineering of light: producing light (lasers), transporting it (optical fiber) and converting it to electrical signals and back (transceivers). In the computing world, light replaces electricity anywhere a great deal of data must move quickly and without losing energy to heat along the way.

The field's centrality stems from the structure of the computation itself. Training a large AI model does not run on one chip but on tens of thousands of accelerators working almost like one giant processor, and they must talk to each other at an enormous rate. As models grow, the bottleneck moves from the question "how much compute is there" to "how fast is the data moved between the compute elements" - and the practical answer to the second question is light.

This is exactly the pattern we wrote about in the Nokia review, which repositioned itself as an optical-infrastructure supplier for data centers and is even acquiring a manufacturing campus in Arizona to be converted to indium-phosphide semiconductor production. And it is the same family of theses as power scarcity: the AI wave does not enrich only the accelerator makers, but the entire physical-infrastructure chain around them.

The fund itself defines its mandate exactly that way: investing in "companies enabling faster data movement between the chips, servers, and data centers powering AI." Per Goldman Sachs data the fund cites, the business opportunity around advanced AI data movement is forecast to grow roughly ninefold between 2026 and 2028 - a third-party forecast, which deserves the usual caution owed to market forecasts.

What the Fund Actually Holds

These are the ten largest holdings, as of July 24, 2026:

Company	Weight	What it does
Lumentum	14.89%	Lasers and transceivers for data centers and telecom
Anthropic exposure (SPV)	11.24%	Private AI company, not publicly traded
Aixtron	8.85%	German equipment for growing compound-semiconductor layers
AXT	7.88%	Indium-phosphide and gallium-arsenide substrates for optical components
Applied Optoelectronics	4.69%	Transceivers and lasers for data centers
Tower Semiconductor	4.29%	Israeli chipmaker with a silicon-photonics platform
Furukawa Electric	4.27%	Japanese optical fiber and cable giant
Landmark Optoelectronics	3.90%	Epitaxial wafers for optical components, Taiwan
Aehr Test Systems	3.75%	Test and burn-in systems, including for photonics
Zhongji Innolight	3.29%	Large Chinese transceiver manufacturer

In total the fund holds 26 positions. The sector breakdown: 83.56% information technology, 4.27% industrials, 11.23% "other" (this is effectively the Anthropic exposure) and 0.94% cash.

Three insights emerge from the list.

First, this is not an American list. There is a German, Japanese, Taiwanese, Chinese and Israeli company here. That makes sense, because the supply chain for optical components is global by nature - but it also means exposure to different currencies and to different regulatory and geopolitical risks, including tensions between the U.S. and China in semiconductors.

Second, this is not a fund of "brand names." Apart perhaps from Tower and Lumentum, most of the names here are unfamiliar to the average Israeli investor. That is not necessarily a drawback - these are precisely the component companies that sit deep in the chain - but it does mean they are harder to follow and tend to be more volatile than the giants.

And third, there is a point of Israeli interest here. Tower Semiconductor of Migdal Haemek, one of Israel's veteran chip companies, is in the fund at a weight of 4.29%. Tower does not manufacture at the nanometer edge but specializes in dedicated processes - and one of them is silicon photonics, meaning the integration of optical components inside a silicon chip. That is exactly the junction the fund focuses on.

And above all, one figure that determines the product's character: Anthropic is 11.24% of the fund, meaning almost 89% of it is something else entirely - publicly traded optical-component companies, mostly small to mid-cap. Whoever buys the fund because of the Anthropic headline is buying, with most of the money, a photonics portfolio. That is the distinction that determines what really moves the return here.

Another connection emerges from the list: AXT, the fourth-largest holding, manufactures indium-phosphide substrates - the same technology for which Nokia is buying the manufacturing campus in Arizona, as we covered last week. When a company like Nokia buys a plant to produce that component, and a dedicated fund holds 7.88% of its substrate maker, those two data points point in the same direction.

The Anthropic Exposure: How It Works

The component drawing most of the attention is, naturally, the 11.24% Anthropic exposure. It is important to understand exactly what that means, because it differs from holding an ordinary stock.

Anthropic is a private company. Its shares cannot be bought on an exchange. Per the company's announcement, it completed a $65 billion funding round in April 2026 at a valuation of $965 billion post-money, and per reports it filed a confidential IPO prospectus on June 1.

So how does an ETF hold it? Through a structure called an SPV - a Special Purpose Vehicle. This is an entity established for one purpose only: to hold a particular asset. In this case, the SPV holds Anthropic shares, and the fund in turn holds a stake in the SPV. This is an accepted structure in private investing, and it is almost the only practical way a tradable product can hold a non-tradable company.

And how is it valued? Here the fund publishes an explicit figure, in wording worth quoting: it says the Anthropic position "was acquired and is marked at an implied $1.8 trillion valuation, a premium to the company's last publicly disclosed funding round but a discount to known secondary market transactions" (as of July 21, 2026).

That is a transparent statement, and it is worth unpacking: the mark is nearly double the valuation of the last funding round ($965 billion), but per the fund it is below the prices of secondary-market transactions

• a market where employees and early investors sell private shares in individual deals. The result: there is no single agreed market price here, but a range - and the fund chooses a point within it. That is legitimate and accepted, and it is also exactly where the risk sits, because 11% of the fund is priced on an estimate rather than on a traded price.

One last point: the fund notes that on the SPV position it charges no brokerage, management or additional performance fees (what it calls "0/0/0"), and that only the general 0.75% expense ratio applies. That differs from the norm in many private funds, where separate management and performance fees are charged.

The Precedent: What Happened with SpaceX

We have already seen this script in full, and not in a parallel sector but at this exact same manager. Tema launched on March 30, 2026 an ETF called Tema Space Innovators under the ticker NASA - a space fund centered on SpaceX exposure through an SPV, exactly the same structure.

The move worked well in terms of raising money. Per the coverage, the fund crossed a billion dollars in assets within just 37 trading days. As of July 24 it manages about $1.06 billion, and SpaceX is its largest holding at 16.39%, marked at an implied valuation of about $1.37 trillion.

And then SpaceX really did go public - on June 11, at $135 per share, in the largest IPO in history (raising $75 billion). On the first trading day it closed at $161, a jump of 19.3%. On June 16 it touched an intraday high of $225.64.

And here is what has happened since: as of today the stock trades around $115 - about 15% below the IPO price, and about 49% below the peak. The pressure comes, among other things, from the lockup release schedule: per reports, a staged release began at the end of July, and the full 180-day lockup expires on December 8, 2026.

And now the lesson that bears directly on LAZR. Whoever bought the NASA fund to capture the SpaceX IPO discovered something important: even after the IPO, the SPV position remained locked. The fund explains this explicitly - like every inside shareholder, the SPV is subject to a lockup expected to last at least six months, and only afterward will the private holding convert to freely traded shares. The fund could not sell into the surge to $225, nor can it react quickly to the decline.

And in fairness, two important caveats to this whole story:

The first - the fund itself actually delivered a return. Despite everything, NASA posted a NAV return of about 22% over three months, because it also holds Rocket Lab, AST SpaceMobile, Viasat and others. Holding a basket is not holding a single stock - and that is exactly the point relevant to LAZR as well.

The second - the fund's mark actually looks reasonable. If we cross-check the implied valuation the fund assigns to SpaceX (about $1.37 trillion) against the market value derived from the actual share price, the numbers are roughly the same order of magnitude. Meaning, at least in this case, there is no evidence of an artificially inflated private mark.

The Interesting Trades: Four Ways Investors Tried to Capture a Private Company

The idea of buying a private company through a tradable security is not new, and in 2024-2026 it became one of Wall Street's hottest trends. And beyond long-term holding, trading activity grew around it too: investors tried to ride the anticipation in waves - entering the exposure vehicle at early stages of the story and exiting as the enthusiasm inflated toward the IPO, not necessarily holding through it. The examples below show that some succeeded and some got burned, sometimes in the same security.

This must be said plainly: such trading demands a high level of skill. It rests on understanding the product's structure, the timetables for lockups and share releases, and the gap between price and net asset value - not on the story itself. Whoever does not master those details is left, in effect, with a bet on a headline. That said, for the investor willing to research in depth what happened in similar IPOs - and especially the SpaceX precedent reviewed above - there is practical and interesting study material here.

Above all, a structural point: the same "exposure" can be an entirely different product, depending on the pipe through which it arrives.

1. An ETF with an SPV (the LAZR and NASA model). The fund holds a special-purpose entity that holds the private shares. The advantage: an ETF's creation-and-redemption mechanism keeps the trading price pinned to net asset value - and indeed LAZR traded at a gap of just 0.36% from it. The disadvantage: the net asset value itself is based partly on the manager's estimate, not on a traded price.

2. A closed-end fund (the DXYZ model). Here the number of units is fixed, so there is no mechanism pinning the price to the value of the assets. The result can be extreme: the Destiny Tech100 fund, which holds a basket of private technology companies including SpaceX, traded in April 2024 around $105 - a premium of more than 2,000% above its asset value. Meaning whoever bought paid more than twenty times what the assets were actually worth. Even today, per the coverage, it trades at a significant premium. This is a live demonstration that the structure can be more dangerous than the asset.

3. A public-private "crossover" fund (the XOVR model). A fund that holds both public and private companies, accompanying them through the transition. Per the coverage, SpaceX was its largest holding at a weight of about 23% - far higher concentration than LAZR.

4. A traditional active fund (the RONB model). Ron Baron's fund held SpaceX at a weight of only about 2%. A tiny exposure, and therefore a tiny effect on the return - for better and for worse.

The practical conclusion: that same "exposure to a private company" can be 2% or 23% of the portfolio, at a fair price or at a 2,000% premium. Before examining the company, it is worth examining the pipe.

And one caveat that must be said in fairness: the ETF structure is indeed structurally preferable here - the creation-and-redemption mechanism prevents the extreme distortion seen in the closed-end fund. But the risk does not disappear, it merely moves. The price is indeed close to net asset value, but the net asset value itself includes a private asset marked by estimate. The question moves from "how much am I paying above the value" to the question of whether the value itself is accurate.

The Risks - In Detail

These are not theoretical risks; some have already materialized in practice in similar products.

1. Valuation risk: what happens if the estimate is too high

This is the central risk unique to this product. 11.24% of the fund is not priced by the market - it is priced by estimate. The fund marks Anthropic at an implied valuation of about $1.8 trillion, while the last publicly disclosed funding round was at $965 billion. That gap - nearly double - is legitimate and explained (the fund says it reflects secondary-market transactions), but it rests on an irregular, thin market where each deal closes separately.

The scenario to understand: if Anthropic ultimately goes public at a valuation below $1.8 trillion, the mark in the fund will have to converge downward - and that falls directly on the net asset value. An IPO at $965 billion, for example, means a decline of nearly half in the value of that component, which is about 11% of the fund - meaning a hit on the order of 5% to the fund's value before a single photonics stock has moved. And the reverse, of course, if the IPO comes higher.

2. Lockup risk: the exposure is locked exactly when it matters

Here the SpaceX precedent speaks for itself. Even after the IPO actually happened, the SPV position remained locked from trading, per the fund for a period of at least six months. The meaning: the fund cannot sell into a surge, and cannot escape a decline. The investor in the fund is exposed to swings in the mark, without the manager having any ability to act.

Regarding Anthropic, the logic is identical: if and when it goes public, it is reasonable that the same structure will repeat - a lockup, and only afterward conversion to tradable shares. Whoever buys today buys an exposure that will open to realization only far in the future, and not necessarily at the price they see today.

3. Sector risk: the historical lesson of 2000

This point concerns the largest holding in the fund, of all things. Lumentum, which makes up 14.89% of it, was created in 2015 as a spinoff from JDS Uniphase - a name that says a great deal to anyone who remembers the year 2000: JDSU was the flagship stock of the telecom bubble. Per the record, its share price tripled three times within half a year, and its market value at the peak, in early 2000, passed $80 billion. Then the bubble burst, and the entire optical-components sector collapsed - one of the most destructive crashes of that era.

What does that mean for today? Not that history repeats. The current demand is real and backed by orders from paying customers, unlike 2000, when infrastructure was built far ahead of demand. But it does say something about the character of the sector: the optical-components industry is unusually cyclical and volatile, tends to build excess capacity at the peak of enthusiasm, and consists of suppliers competing for orders from a handful of giant customers. The fund itself discloses that photonics companies carry "higher levels of risk and volatility."

4. Concentration risk

The two largest holdings - Lumentum and Anthropic - are together over a quarter of the fund. The five largest are over 47%. This is not a broadly diversified fund; it is a focused bet. A negative event at one large supplier can move the whole portfolio.

5. Geopolitical and currency risk

The fund holds companies in Germany, Japan, Taiwan, China and Israel. That brings three separate risks: currency exposure that is not necessarily hedged; regulatory risk around export restrictions on semiconductor technologies between the U.S. and China, which can hurt suppliers on both sides directly; and concentrated geographic risk - a significant part of the optical-components chain sits in East Asia, including Taiwan.

6. Risk of the fund itself

Assets of about $17.8 million are a very small scale. That has three implications: the average 30-day bid-ask spread stood at 0.64% - a cost added to the 0.75% expense ratio; thin trading volumes may make it harder to enter and exit in large amounts; and in funds that fail to raise assets, issuers sometimes simply decide to close the fund and return the money - an event that forces realization on a timetable that is not the investor's. The fund has existed for less than two months, and the NAV return since inception stood at 1.65% - too short a span to learn anything from.

7. Story risk

And the last risk is behavioral, and perhaps the most common of all: buying a product because of one line in a headline. Anthropic is 11% of the fund. Whoever buys "exposure to Anthropic" and actually receives a portfolio of nine optical-component companies may discover that their return is determined by things they never examined. That is exactly what happened with the NASA fund: it delivered a nice return, but not because of SpaceX.

Summary

LAZR is a young, small ETF doing two things in parallel: it gives concentrated exposure to the photonics chain

• lasers, substrates, transceivers and fiber that move data quickly between AI components - and within that same product it also holds a stake of about 11% in a private AI company that has filed a prospectus. For an investor looking for a way to touch the optical link of AI infrastructure, this is an interesting starting point for examination, mainly because of the composition of the holdings - including Israel's Tower and AXT, which connect directly to the theses we have covered here.

And against them stand the facts that must be known: a fund less than two months old, small assets, non-trivial trading spreads, and part of the value set by estimate rather than by a traded price. The SpaceX precedent is the most practical reminder - there the story materialized in full, and even then the road from exposure to outcome was longer and more winding than expected. As always, whoever examines such a product would do well to read the fund's page and the prospectus themselves, not the headline.

Sources: the official fund page of the Tema Photonics & Optical ETF (temaetfs.com/lazr) and the Tema Space Innovators ETF page (temaetfs.com/nasa), including holdings, weights, marks, fees, spreads and the SPV risk disclosures - accurate as of July 24 and July 21, 2026 as stated therein; Anthropic's announcement of its Series H round at a $965 billion valuation (anthropic.com) and CNBC and TechCrunch reporting on it; CNBC reporting on the SpaceX IPO and its prices, and Axios, Investing.com and Motley Fool reporting on the lockup release schedule; Benzinga, Morningstar and Acadian reporting on the DXYZ fund and its premium to net asset value; TheStreet and CNBC coverage of the NASA fund's asset gathering. Data accurate as of the time of writing (July 27, 2026) and changes continuously - fund holdings change daily. Nothing herein constitutes a recommendation, advice or an offer to purchase any security - see the full disclaimer at the bottom of the page.