NODE Is a Barbell, Not a Basket: The Miner-to-AI Bet Hiding in the Top 10

VanEck's NODE Onchain Economy ETF added 2% in August. The Nasdaq 100 added 3.9%. The S&P 500 added 2.6%. NODE trailed both. That is a small gap, and on its own it is not a signal — a one-month slip says more about who was in the tape than about what the fund holds.

The reason the slip happens at all is the more useful thing to understand, because it is the clearest read on what NODE actually is and where its risk lives. It is not a broad basket of "the onchain economy." Open the top positions and the label doesn't fit. The fund's five largest holdings are VanEck's own in-house digital-asset vehicle at about 13%, then four bitcoin miners in the middle of converting into AI data centers — Cipher at 5.9%, Hut 8 at 5.85%, IREN at 5.8%, and TeraWulf at 5.6%. Those four names are roughly a quarter of the fund, and together with that in-house digital-asset position they account for close to two-fifths. Everything else — stablecoin issuers, software, infrastructure — is the long tail.

So NODE is less a diversified onchain bet and more a high-beta barbell: a direct digital-asset sleeve on one side, and a concentrated bet on a specific, unproven business transition on the other. The August miss is just the low-beta side of that barbell showing itself on a calm tape. The risk is in the miner-to-AI half.

Here is why. Look at the two biggest of those four, and put the valuation next to the financials.

IREN is the largest at a $17.6 billion market cap. It trades at about 24.9 times trailing revenue. But its operating margin is negative — roughly -54% — its return on invested capital is negative, and its free cash flow is deeply negative. It is burning capital at an industrial scale: about $4.4 billion in capital expenditure over the trailing twelve months. The market is pricing IREN as if the AI data-center deal has already won. The financials still read like a mining company spending to find its next identity. That gap between the multiple and the margin is the whole thesis sitting in one stock.

CIFR is sharper. Cipher trades at about 38.5 times trailing revenue — the highest revenue multiple of the group — and at 12.5 times book. Its revenue actually fell 28.7% over the most recent quarter while it spent roughly $1.2 billion in capex. In other words, the most expensive name in the complex on a revenue basis is the one whose revenue is shrinking. It has bounced hard off its lows — up roughly 17% over the past four months, with nearly all of that gain landing in the last five sessions — but it is still trading below its 200-day line, and momentum is a timing tool, not a business model. When the tape is slow, the multiple has nowhere to hide from the underlying report card.

The pattern across the group is the real story: these are all pre-profitability names being marked to market on revenue, not earnings. APLD at least shows accelerating revenue — up about 167% year over year — and is the cheapest at roughly 12.6 times trailing revenue, but it is also still loss-making. None of them prints a clean earnings multiple you can underwrite. You are buying on the revenue trajectory and the AI promise, not on cash flow.

And here is the catch that makes the financing leg of the barbell the live risk. To turn mines into data centers, these companies need money, and they are spending it fast. IREN's roughly $4.4 billion capex, APLD's roughly $2.9 billion, CIFR's $1.2 billion — that is real industrial spend for companies that are not yet generating positive free cash flow. The gap has to be funded, and funding it usually means issuing more equity or convertible debt — which dilutes the very holders buying on the multiple. VanEck's own August commentary lands on exactly this point: it singles out the miners attempting to convert to AI data centers as facing significant regulatory and financing pressure. That is the bear case, in the manager's own framing, on the single transition the fund is most exposed to. It is not a hedge. It is the center of gravity.

There is a second, quieter risk in the tail. Circle (CRCL) — the stablecoin name — took a hit in June when the Open Standard consortium launched "Open USD", a competing stablecoin model, and it has since been marked down as a top detractor. That is the risk that the onchain economics get compressed by competition, which is a different kind of risk than the execution risk on the miners, and it lives in the part of the portfolio the top-ten list doesn't fully show you.

So what does the factor stack say to do? This is a sleeve, not a core holding. If you want onchain-economy exposure, NODE is one way to get it — and it is cheap to size because it is small, a little over $60 million in assets across 50 names. But understand what you are owning: a digital-asset leg plus a concentrated, execution-dependent bet on whether a handful of unprofitable miners can out-execute on a multibillion-dollar AI pivot before dilution and financing pressure eat the multiple. The trade works when AI data-center demand stays hot and the tape is moving, which is why it rallies hard in good months and quietly underperforms in calm ones.

The single variable that would change the read is the gap I flagged in IREN and CIFR: do the revenue multiples stay stretched while free cash flow is still negative and capex is still compounding? If the financials ever start to flex toward positive cash flow, the "mining company at a data-center multiple" becomes "data center company still being underpaid," and that is the inflection. Until then, the multiple is doing the work that the earnings have not yet earned. Give the winners room, but keep the size honest to a bet that has not yet paid for itself.