"Communication Services" is a label. It is not a promise.
Last week I pulled the Vanguard Communication Services ETF holdings disclosure. Ticker VOX. Asset class: US equity sector fund. Top three positions: Alphabet, Meta, and โ in the reporting I was reading โ SpaceX. A rocket company sitting inside a telecom-and-media basket. The filing math suggests the top three could clear half the fund.
I stopped reading the marketing and went to the mechanism.
The index VOX tracks carries "25/50" in its name. That is not a slogan. It is a constraint. Under the Investment Company Act of 1940, a registered fund must pass a diversification test: no single issuer above 25% of assets, with the concentrated remainder capped at 5% per name. The number in the index name is the compliance code written into the product's DNA.
Twenty percent. That is the reported SpaceX weight. It sits close enough to the ceiling that the mechanism โ not the manager โ now decides what this fund actually is.
Let me slow down for anyone who does not live inside index methodologies.
Vanguard runs VOX. VOX is an ETF. It is registered under the Investment Company Act of 1940, which classifies it as a regulated investment company, or RIC. RICs receive structural and tax privileges. In exchange, they accept rules. One of those rules โ the diversification test โ is exactly what the "25/50" name encodes. No single issuer may exceed 25% of the portfolio. The remaining half of the book, after the first concentrated block, must hold each issuer at 5% or below.
This is not decoration. It is a kill switch. When a position breaches the ceiling, the index reconstitutes and the fund must rebalance. The manager does not choose. The formula does.
Now the weighting. The index does not use full market capitalization. It uses free-float weighting. Only the shares actually available to trade count. Locked shares, insider positions, and pre-IPO restricted stock are excluded from the float. Lower float, lower weight. More float, more weight.
That single design choice is the entire story.
SpaceX, in the scenario I was reading, went public with roughly 5% of its shares trading. The other 95% sit under lock-up, released in tranches over time. Under free-float weighting, that 5% float produces a small initial weight. As tranches unlock, the float grows. The weight grows with it. The fund does not buy a single share. Its exposure climbs anyway.
This is where a "sector fund" quietly becomes something else.
I have audited vesting contracts on-chain. The pattern is identical. A token launches with a thin circulating supply. An index or a pool weights it by what is tradeable. The unlock schedule then does the buying for everyone. No vote. No decision. Just a formula executing a thesis nobody wrote.
Let me do the arithmetic, because arithmetic does not care about narratives.
The existing top two โ Alphabet plus Meta โ are reported at 42.4% combined. Add SpaceX at a reported 20%. That is 62.4% in three names. Two of them are the same company's extended family: digital advertising and social platforms, both squarely in the growth factor. The third is a capital-intensive aerospace and satellite operator.
Here is the part most people miss. Alphabet, Meta, and SpaceX are not diversifiers of one another. They load on the same factor: long-duration growth. When discount rates move, they move together. When the growth trade unwinds, it unwinds in all three. So the fund does not hold a 62% concentrated book with internal hedging. It holds a 62% single-factor book with no hedging. That correlation is structural, not coincidental.
Walk the test line by line. Group one is every issuer held above 5%. That group, combined, cannot exceed 50% of assets. Within the group, no single issuer can exceed 25%. Everything else โ the non-concentrated half โ must sit at 5% or below per issuer. VOX's reported top two already occupy 42.4 points of the 50-point concentrated budget. SpaceX at 20 would blow through the group cap and nearly reach the single-name ceiling. That is why the arithmetic matters: the fund is not drifting toward its limits. It is already pressed against them.
Now watch the mechanical amplification.
The unlock schedule sits on a calendar. Each unlock expands the float. Each expansion raises the index weight. The fund's mandate says: replicate the index. So the fund, or its authorized participants, must match the new weight. In a rising market, this means buying more of the name that just went up. In a falling market, it means selling into weakness. Free-float weighting against a staged-unlock IPO is a pro-cyclical lever wearing the costume of a neutral rule.
Follow the float. Ninety-five percent of the shares are locked, which means the marginal buyer and the marginal seller are trading a tiny slice of the cap table. A small flow moves the price disproportionately. The float expands on a calendar. Every unlock adds supply precisely when the price โ and therefore the index weight โ has been rising, because the float-and-price product is what the weighting formula sees. The formula is rewarded for momentum it did not create and cannot manage.
Most IPOs float 15% to 25% on debut. The lock-up releases are gradual and the float is broad enough that index weighting behaves itself. SpaceX at roughly 5% is an outlier, and free-float weighting is not stress-tested for outliers. The formula is neutral. The calibration is not. When you run a formula outside the range it was built for, you do not get a slightly wrong answer. You get a qualitatively different one.
I have seen this exact failure mode in DeFi. Liquidity-mining programs weight rewards by existing liquidity. Early liquidity providers accumulate. Their positions auto-compound. The mechanism rewards those already positioned and converts a "fair launch" into a quiet concentration. Nobody coded malice. The formula did it.
The float here is the liquidity. The unlock is the emission schedule. The index is the pool. And the holder โ the person who thought they bought a sector fund โ is the late depositor, arriving after the concentration has already been priced in.
Now the liquidity risk, which is subtler and worse.
SpaceX trades with roughly 5% float. Thin book. Wide spreads. High market-impact cost. When the index rebalances โ and it will, on the schedule and on any 25/50 breach โ the trade has to clear through that thin pool. The impact cost is real. Who pays it? Not Vanguard. Vanguard charges a basis-point fee on assets under management. That fee is insensitive to concentration. The cost lands in the fund's tracking error. It lands on the holders.
The ledger remembers what the wallet forgets. Every rebalancing tick through a thin order book is a small tax the wallet never sees and the ledger never forgets. It shows up as slippage, as a widening bid-ask, as an intraday premium and discount that should not exist in a supposedly plain-vanilla sector ETF.
The rebalance makes the holder pay twice. First in the trim: the index caps the offending name, and the fund sells it โ often on a schedule dictated by float expansion, not by price. Second in the impact: those sells clear through the thin pool, so the exit is worse than the mark. You get the concentration on the way up and the mechanical selling on the way down, both mandated by a rule the holder never read.
Then there is the authorized participant chain. ETFs rely on APs to create and redeem shares, arbitraging away premiums and discounts. That arbitrage needs a liquid underlying basket. When one name is 20% of the book and trades with a 5% float, the AP's basket is only as liquid as its least liquid leg. Arbitrage gets expensive. The premium and discount widen. The passive wrapper shows its mechanical seams.
The DeFi analogue is exact. I have watched an on-chain index product rebalance into a pair where one leg had near-zero depth. The formula said the target weight. The market said there was no bid. The rebalance executed at a price that existed for exactly one transaction. That is the same trade VOX will run if SpaceX's float stays thin while its weight climbs. The index never asks whether the market can absorb the order. It only asks whether the weight is correct.
Now classification, which is the most under-discussed lever here.
SpaceX is being sorted, in this scenario, into Communication Services โ the theory being that Starlink is satellite internet, which is telecom-adjacent. That is defensible under GICS logic. But look at what classification does. It routes 100% of SpaceX's index weight into a single sleeve: the sector fund. If SpaceX were classified as Industrials or Aerospace, it would sit in a different fund, and VOX would never touch it.
Classification is the routing table. Change one entry, and the entire concentration picture rewrites itself. If MSCI re-sorts SpaceX in a future cycle, VOX's exposure does not decay โ it evaporates, and the fund is forced to sell on the index's schedule, not the holder's. A classification edit becomes a forced transaction.
I keep returning to a line I use in audits. Code is law, but bugs are the human exception. Index methodologies are code. The 25/50 test is a constraint in that code. The free-float formula is a function. And the human exception โ the edge case the designer never intended โ is the low-float, staged-unlock IPO. The rulebook was written for companies whose traders already own most of the float. It works beautifully there. Point it at a company where 95% of the equity is locked, and the rule produces the opposite of its intent: not diversification, but slow-motion concentration with the manager's hands tied.
There is a category of failure I meet constantly in systems work: the purely mechanical process everyone assumes is neutral because no human touches it. Automation removes discretion. It does not remove bias. The bias moves into the specification. Free-float weighting is a bias toward whatever the lock-up schedule releases. It is a specification choice, made once, compounding forever. The fund is not managed. It is specified. And a specification runs until someone edits the spec โ which is exactly what a classification change or an index inclusion does.
Now the incentive question. This is where the forensic lens matters most.
Vanguard earns a fee on AUM. Concentration does not hurt the fee. It can help it, when the concentrated bet wins, because performance attracts flows, and flows raise AUM, and AUM raises revenue. When the bet loses, redemptions amplify the drawdown โ and fee collection shrinks with AUM โ but the manager bore none of the concentration risk on the way up or down. The manager is paid on the size of the ship. The holders are the ones in the water.
I have built and audited protocols where the same misalignment lived in the code. A vault that takes a management fee on total value locked, regardless of strategy risk. The vault designer eats no drawdown. The depositors eat all of it. The code compiles cleanly. The incentive is still broken.
And note the product-line temperature. The reporting I read suggested VOX uses free-float weighting more aggressively than some Vanguard peers. Same manager, different sleeve, different appetite for concentration. That is not a bug in one fund. It is a strategy differential inside one house โ which means someone at the firm already knows concentration is a performance lever, and is choosing when to pull it.
The retail holder does not read the float. They read the label. "Communication Services." A reasonable person believes that means advertising, media, telecom โ mature cash flows, modest volatility. They are instead holding a book where the marginal position is an aerospace name trading on a 5% float, weighted by a formula that treats thin liquidity as a feature. The gap between the label and the ledger is the whole product. That gap is not disclosed in one sentence. It is embedded in a methodology document most holders will never open. That is the quiet part. A holder can be fully informed about SpaceX and still be misinformed about their own fund.
Here is where I diverge from the reflexive panic.
The obvious read is: "Concentration. Danger. A sector fund is secretly a rocket-stock bet." Correct, as far as it goes. But the reflexive read misses two things.
First, the mechanical relief valve. The 25/50 test is not only a constraint. It is a circuit breaker. If SpaceX's weight climbs toward 25%, capping and forced rebalancing kick in. That is not a warning. It is an automatic trim. The rule that created the exposure is also the rule that caps it. The index will not let one name run away past the ceiling. So the doomsday scenario โ a sector fund becoming a 40% single-stock fund โ is structurally fenced. The code will not allow it. In that sense, the mechanism is self-policing, even if it is self-policing by accident.
Second, the dissolution path. If SpaceX matures and gets added to a broad-market index โ the reporting floats a timeline around 2027 โ the weight stops being trapped inside VOX. It spreads across every index tracking the broad market. The concentration is not resolved by a hero. It is resolved by breadth. A problem that looks like a time bomb is, on a longer horizon, a time-release capsule of its own resolution.
But โ and here is the blind spot I actually care about โ both relief mechanisms are scheduled events, and the danger lives in the gap between now and then. The kill switch only fires near 25%. The dissolution only happens after broad-index inclusion. In the interim, the fund can sit at a 20% single-name weight with 5% float and a 62% single-factor top three. The mechanisms protect the endpoints, not the path. The protocol guards the ceiling and the exit. It does nothing about the corridor in between. That corridor is where the widest gap between expectation and reality gets recorded.
And the kill switch fires late. A circuit breaker that only trips at the last quarter of the range does not protect the first three. By the time capping engages, the holder has already absorbed the amplification from every prior unlock. Protection at the edge is not protection through the middle.
The crypto-adjacent angle nobody is naming: the fact that a crypto outlet broke this story about a Vanguard ETF is not a coincidence. The structural problems of token launches โ thin float, staged unlocks, index weighting doing the buying โ have migrated into traditional asset management. The two worlds are running the same playbook. The code that ran ICOs now runs index funds. If you understand unlock schedules, you already understand this ETF better than its own marketing does.
So watch two numbers, not the narrative.
First, the top-three combined weight. The moment it clears 50%, the diversification argument is formally dead, and everyone still holding the "sector fund" label is carrying a bet they never placed.
Second, the SpaceX weight against 25%. The closer it drifts to the ceiling, the closer the index's own kill switch comes to firing โ and the closer VOX comes to being forced to sell the very thing its recent performance depends on.
The floor question is not whether SpaceX is a good company. It is not whether Vanguard is a good manager. It is whether anyone holding VOX to get "communication services" knows that a rocket company, a free-float formula, and an unlock calendar are now deciding their returns.
The ledger will show it. The wallet just will not feel it until the rebalance.
