-2.png)
Most of the world's leading AI labs are private, which the ETF industry has decided is a paperwork problem rather than a real one. One of them is not private, which turns out to be the more awkward case.
Harbor AI Lab Ecosystem ETFs (ANTW, OAIW, DEPW, MTAW, XAIW; NYSE)
Five funds arrived at once, one each for Anthropic, OpenAI, Google DeepMind, Meta AI, and a fifth I will come back to. Four of them cannot hold a single share of the company on the label, because no such share is available to buy. What each fund holds instead is the commercial ecosystem orbiting that lab: the cloud providers running its inference, the chipmakers filling its training clusters, the enterprise software firms that have embedded its models and now issue press releases saying so.
This is, if you squint, quite a reasonable idea. The economic reality of the AI boom is that the money flows outward from a handful of private labs into a very public supply chain. Building a fund around that flow is a more honest description of the trade than most thematic products manage.
It is also, if you unsquint, a portfolio defined by the company it is not allowed to hold. These are actively managed, so a human being is deciding what counts as sufficiently adjacent to each lab. Somewhere there is a portfolio manager whose job is to answer the question “is this company Anthropic enough?" and who has to answer it again every quarter. I would like to see the meeting notes.
The overlap problem is the one that will get discussed. Four of these five labs buy from broadly the same suppliers. There is a real possibility that an investor who buys three of these funds to express three different views ends up owning the same handful of names three times over, at three separate management fees. The differentiation is genuine at the edges and thins considerably toward the middle.
And then there is the fifth ticker, which went into the filing as an xAI fund and arrived on the tape as SpaceXAI. That is not a marketing flourish. SpaceX bought xAI in the largest merger ever transacted, on the stated logic that orbital data centres require you to own both the rockets and the models, and the lab now operates under the combined name. The fund is simply keeping up with the paperwork.
Keeping up with it in one respect, at least. Because the whole premise of this product family is that you cannot buy the lab, and in this one case you absolutely can. The parent went public in one of the largest debuts ever staged, deliberately routed an unusually large slice of the offering to retail, and now trades on the Nasdaq at a valuation with a T in front of it. Any investor who wants exposure to that lab can simply type the ticker.
So XAIW is an ecosystem fund wrapped around a company that is already available at one click, sold alongside four sibling funds whose entire reason for existing is that their labs are not. It is the one fund in the set that has to justify itself on stock selection rather than on access, and it is the one most likely to be bought by someone who thinks they are getting the hard-to-reach thing. The name says SpaceXAI. The holdings, necessarily, say everyone else.
VegaShares AI Inference ETF and VegaShares AI Thermal, Cooling and Power Management ETF (CGPT, COOL; NYSE)
Two products from a suite that also includes funds for AI networking, rack hardware, fab equipment and raw compute. The AI trade has been sliced so finely that you can now express a view on one layer of the data centre stack without accidentally taking exposure to any of the others.
CGPT is the AI Inference fund, and the ticker is doing a great deal of unpaid labour. Inference is the sensible half of the AI infrastructure story: training is the capital expenditure, inference is the recurring bill, and a fund built around the latter is pointed at the part that has to be paid every single day. The ticker, meanwhile, is pointed at a chatbot the fund does not own, cannot own, and has no relationship with. Marketing and mandate, working in different buildings.
COOL is the one I admire. It buys the companies that keep the data centres from cooking themselves: thermal management, cooling systems, power distribution. This is HVAC. Extremely well-branded HVAC, sold at the valuation multiple of an AI company, but HVAC. The pitch is sound, because heat and power are the actual binding constraints on the whole build-out, and the people who solve them get paid regardless of which lab wins. Getting the ticker COOL for a cooling fund is the kind of small victory that makes a product launch worth showing up for.
Evolve Nasdaq Technology UltraYield ETF (TECY; TSX)
Take the 46 technology companies inside the Nasdaq-100. Apply leverage. Then write covered calls on top of the leverage. Then distribute the proceeds at least twice a month.
Read that sequence slowly, because the two halves are pulling in opposite directions on purpose. The leverage amplifies the upside. The call writing sells the upside away for income. You are borrowing money to buy growth stocks and then agreeing, for a fee, to cap how much growth you keep. The prospectus calls the leverage modest, and the strategy does have an internal logic: the calls dampen the volatility that leverage introduces, and the income offsets the borrowing cost. It nets out to a yield product wearing a technology fund's clothing.
It will be bought for the distribution schedule, which is fair enough. Twice-monthly cash has a gravitational pull that no amount of prospectus language can counteract. Just be clear that “UltraYield" describes the cash flow rather than the total return, and that in a year when Nasdaq technology rips, this will be the fund quietly explaining why it did not.
Harbor Short Term Treasury ETF (HBIL, NYSE)
Which brings me to the quietest fund in the batch, and the detail I cannot stop enjoying.
HBIL tracks the three-month US Treasury bill market. That is the whole strategy. It holds the safest, most liquid, least interesting instrument in global finance, and it does so on behalf of investors who have decided that the correct amount of thematic conviction is none.
It comes from the same issuer as the five AI lab funds. The same firm, the same trust, filed a three-month T-bill fund and then, within days, filed five funds named after private artificial intelligence laboratories. Both sets of paperwork walked onto the tape in the same batch.
That is not hypocrisy. It is a barbell, and it is arguably the most accurate view of the current market anybody has published. Own the ecosystem of the labs, because that is where the growth is. Hold the bills, because you may need the money back at a moment of your choosing rather than the market's.
One of those two products will still be trading in a decade. I am fairly confident which.
Comments