The Notional Is Not the Risk: Autopsying the $9.6 Trillion Options Expiry

The Number That Travels

I ran the decomposition twice on a Tuesday afternoon in Copenhagen, because the first pass looked like the kind of error a junior would make — a misplaced decimal, a multiplier applied twice. The figure in circulation was $9.6 trillion: the notional value of U.S. options expiring on September 18, attributed to Citadel Securities, relayed through a crypto news outlet, and by the time it reached my timeline it had become a weather event.

Nine point six trillion dollars. Written that way, it reads like a shockwave. Written the way a risk system actually stores it, it reads like a boundary condition.

Notional is what you get when you multiply contracts by multiplier by underlying price. It is the size of the sandbox, not the size of the child. Of that $9.6 trillion, the portion that obliges anyone to buy or sell anything is the delta-adjusted exposure — and if you apply any defensible delta distribution to a real index option book, you land somewhere between eight hundred billion and one and a half trillion dollars. The headline figure overstates the hedgeable exposure by roughly an order of magnitude, and that ratio is not a nuance. It is the entire story.

The remainder is a legal artifact. Cash-settled European-style index options deliver no shares. They resolve to a number, that number is netted against a clearinghouse ledger, and the net cash movement is a rounding error next to the gross figure that made the headline.

I have spent long enough inside these structures to know the rest of the news cycle. The number gets quoted. The number gets amplified. The one measurement that would tell you whether September 18 actually matters — the sign of dealer net gamma — never appears, because it is not in the press release and cannot be derived from open interest alone.

So let me do the work the headline skipped.

The Relay Chain Tells You More Than the Number

Before the arithmetic, the provenance. The claim originates with Citadel Securities. It is picked up by trade press. It is then relayed by a crypto outlet. By the third hop it has become a standalone fact with no parentage.

The provenance is the insight. Citadel Securities is not a research house that happens to trade. It is one of the largest equity options market makers on the planet. When the largest options market maker publishes the observation that options markets are becoming more influential, you are not reading a neutral measurement of a system. You are reading a description of a balance sheet written by the balance sheet.

That is not a conspiracy claim. It is a reflexivity claim, and it is much harder to dismiss. A market maker's entire economic function is to warehouse the gamma that the rest of the market does not want. Any statement about the growing importance of the options market is simultaneously a statement about the growing importance of the firm making it. The observer is inside the system being observed. There is no external vantage point from which to check the claim, because the claim is about the vantage point.

The second-hop detail matters too, and almost nobody flagged it. A crypto publication covering a U.S. equity options expiry is not a category error. It is a confession. It means the desks that read that publication now treat U.S. equity volatility as their own primary risk factor. Crypto did not decouple. It re-denominated. The correlation regime that ran through 2022 and 2023 — when digital assets traded as the high-beta tail of a single global risk appetite — never reversed. It just stopped being newsworthy.

There is a third layer that the source does not disclose at all: the year. The reporting gives a date — September 18 — and no year. September 18 falls on a Friday only in certain years, and the quarterly derivatives expiry (the third Friday of March, June, September, and December) is the only occasion on which a $9.6 trillion concentration is mechanistically plausible. So the reasonable inference is that this refers to a quarterly quadruple-witching window. But an inference is not a disclosure. A number without a year attached is not a datum. It is a mood.

I want to be precise about how much that matters. It does not change the mechanics I am about to describe, because those mechanics repeat four times a year with metronomic reliability. It does change whether the event is a forecast, a present-tense condition, or a historical footnote. A reader who cannot place the event on a calendar cannot act on it, and a reader who cannot act on it is not being informed. They are being entertained.

What a Quarterly Expiry Actually Is

Strip away the jargon and the structure is simple enough to sketch on a napkin.

Four times a year, on the third Friday of March, June, September, and December, four classes of instrument expire simultaneously: index futures, index options, single-stock options, and single-stock futures. The market calls it quadruple witching, which is a colorful name for a mundane event — the largest scheduled liquidity concentration of the quarter, known in advance, anticipated by every desk on the street, and priced accordingly.

Index options in the U.S. are predominantly European-style and cash-settled. There is no assignment, no delivery, no scramble for borrow. What expires is a set of contingent claims. What settles is the difference between a strike and a settlement value, computed off a special opening quotation. The gross notional is enormous because the multiplier is large and the index level is high. The net cash flow is small.

This is where the language of the headline breaks down. Calling $9.6 trillion "expiring" implies that $9.6 trillion of something is about to leave a room. Nothing leaves a room. What happens is that a very large set of contracts reaches its terminal date simultaneously, and the counterparties who sold them must arrive at that date with their hedges in a state that produces no residual exposure.

That obligation is where the real market impact lives. Not in the notional. In the hedging.

Three numbers, and only one of them moves markets

The first number is notional. It is computed as contracts × multiplier × underlying. It is the number in the headline. It is invariant to moneyness, direction, expiry, and volatility. It is, in the strict sense, information-free.

The second number is delta-adjusted notional. Delta measures the sensitivity of an option's price to a one-unit move in the underlying, and by extension the equivalent number of underlying units the option synthetically represents. A deep in-the-money call has a delta approaching one. A far out-of-the-money call has a delta approaching zero. An at-the-money option sits near 0.5, modulated by time to expiry and implied volatility.

Real index option books are not concentrated at the money. They have heavy wings, large short-dated populations, and a persistent structural tilt from overwriting and hedging programs. A blended absolute delta of 0.08 to 0.15 is a defensible range for the aggregate. Apply that range to $9.6 trillion and you get roughly $770 billion to $1.44 trillion of delta-equivalent underlying exposure.

And even that overstates the pressure, because most of it is offsetting. Long calls sit against short calls. Dealers run books that are, by design, close to delta-neutral. Net dealer delta is usually a small number maintained near zero by continuous hedging.

The third number is the one that actually determines whether September 18 is a quiet Friday or a violent one. It is dollar gamma — the rate of change of delta with respect to the underlying, scaled by position size. Gamma is the curvature of the payoff. It is the thing market makers are forced to trade against, and it is the only number in this entire analysis that has a directional consequence.

The headline reports the first number, ignores the second, and never mentions the third. The notional is not the risk; it is merely the limit.

Gamma Is the Only Object in the Room

Here is the mechanism, and it is worth stating without decoration because most commentary gets it backwards.

A market maker who is long options is long gamma. As the underlying rises, their delta rises, so they must sell. As the underlying falls, their delta falls, so they must buy. The hedging flow is countercyclical. It leans against price movement. This suppresses realized volatility and produces the phenomenon traders call pinning — the tendency of a large open interest strike to act as a gravitational center into expiry, because the aggregate hedging demand is strongest wherever the gamma is concentrated.

A market maker who is short options is short gamma. The flow inverts. As the underlying rises, their delta falls short of the required level, so they must buy more. As it falls, they must sell more. The hedging flow is procyclical. It amplifies movement in both directions. This is the substrate of a gamma squeeze, and it is the reason that the final hours of a heavily-weighted expiry can produce price action that looks nothing like the preceding three weeks.

Everything that follows is downstream of which of these two regimes is dominant. And here is the uncomfortable part: the sign of dealer net gamma is not observable from public data. Open interest tells you how many contracts exist at each strike. It does not tell you whether the dealer is long or short them. Public gamma exposure estimates differ from each other by hundreds of billions of dollars precisely because they must impute the side of every trade, and the imputation rules are guesses dressed in decimal places.

So the $9.6 trillion figure does not tell you the direction of anything. It tells you the size of the board. It is compatible with a week of listless drift and a week of cascading squeezes, and no amount of additional precision on the notional number will resolve which one you get.

This is where my priors about headline metrics were formed, and they were formed badly before they were formed well. In 2017 I spent twelve-hour days auditing the Solidity source of the Golem Network's token distribution contract. I found three integer overflow vulnerabilities in the pledge logic — arithmetic that silently wrapped past its maximum and produced numbers that were confidently wrong. I wrote the proof, submitted the pull request, and was told it was too academic.

The lesson was not that the vulnerabilities were fake. They were real. The lesson was that a number can be arithmetically valid and economically meaningless at the same time, and that the market has a strong preference for the version of a number that requires no qualification. That is exactly what is happening to $9.6 trillion. It is a correctly computed figure wearing the wrong label, and the label is doing all the work.

Charm, Vanna, and the Compression of Time

Gamma is only the first derivative of the story. Two more matter, and they are the reason the final week behaves differently from the preceding one.

Charm is the change in delta with respect to the passage of time. Near expiry, charm stops being a gentle drift and becomes a cliff. An option that is near the money with two days remaining has a delta that can swing from 0.3 to 0.7 over a modest move in the underlying; with two hours remaining, the same option's delta can traverse the same range on a move an order of magnitude smaller. The hedging obligation that accompanies that delta change does not scale down with the remaining time. It scales up.

Vanna is the change in delta with respect to implied volatility. When the volatility surface shifts — and it always shifts into a large expiry — delta hedges must be adjusted even if the underlying has not moved a single tick. This is the mechanism by which a volatility event becomes a price event. A repricing of the surface forces flows in the underlying, and those flows are read by participants as directional information when they are in fact derived from a second-order Greek.

The practical consequence is that the final week of a quarterly expiry is a regime of its own. Realized volatility tends to compress going in, because positive-dealer-gamma conditions dominate when overwriting flow is heavy, and then release on the other side, because the gamma that was pinning the index expires worthless and the balance sheet that was absorbing movement stops absorbing it. Traders call the release a volatility crush when it deflates implieds. They call it something less printable when it goes the other way.

What matters for anyone reading a $9.6 trillion headline is that neither charm nor vanna appears anywhere in it. The headline describes a stock. The market prices a flow.

Zero-Day Options and the Terminal Case

There is a structural variable that has changed the arithmetic of expiry windows more than any other in the last several years, and it is not mentioned in the source material at all: the migration of volume into options with zero days to expiration.

A zero-day option is the limiting case of charm. It has no meaningful extrinsic value until it suddenly has all of it. It goes from worthless to binary in the space of a session. For a market maker, the hedging obligation on a zero-day book is not spread across weeks of decay. It is concentrated into hours, and within those hours into minutes.

In U.S. index options, zero-day contracts have on many sessions grown to account for more than half of total volume. That is a structural transformation of the market's intraday physics. It means that the realized volatility of the underlying is partly manufactured by the hedging of instruments that expire before the close. It means that the largest liquidity events of the quarter are now embedded inside an intraday rhythm that did not exist a decade ago.

The interaction with the headline figure is direct. If a material share of the $9.6 trillion in expiring notional is zero-day or one-day paper, then the mechanical amplification of intraday movement is far larger than a delta-adjusted aggregate would suggest. If instead the expiring book is dominated by longer-dated index positions and structured product hedges, the amplification is concentrated in the closing auction and the settlement calculation.

The source material does not disclose the term structure. Neither does the headline. This is a gap that no amount of commentary can fill, and I want to be explicit that I am not going to invent a number to close it. What I will say is that the term structure is the single highest-value piece of missing information in this entire story — higher value than the notional, higher value than the delta adjustment, higher value than any estimate of gamma.

The AMM Is a Short Straddle, and That Is Not a Metaphor

Here is where this stops being a story about equity options and becomes a story about decentralized finance, and it is a connection I have not seen made in the coverage of this number.

In 2020, during the first DeFi summer, I wrote a Python simulator to model liquidity provision under volatile conditions. The goal was unremarkable — I wanted to check the standard derivation of impermanent loss that every blog post was copying from every other blog post. The standard derivation was wrong. It assumed an arithmetic mean where the constant product formula requires a geometric one, and the error produced a systematic underestimate of the divergence between holding and providing.

The corrected derivation produced a result that took me a while to accept, because it recast the entire activity. The payoff profile of a constant-product liquidity provider is a short straddle. The LP sells volatility. The position is concave in price: it underperforms holding on every large move in either direction, and it outperforms in a narrow band around the entry price. The fee income is not a yield in the conventional sense. It is the premium collected for being short gamma.

Once you see that, the connection to the equity options expiry is not an analogy. It is an identity. The market maker on the other side of a $9.6 trillion expiry and the LP on the other side of a Uniswap pool are the same economic actor wearing different infrastructure. Both are warehousing the variance that someone else does not want to hold. Both are paid a premium for it. Both are exposed to the same underlying risk factor, which is realized variance.

This has three consequences that the news cycle misses entirely.

First, it means the crypto market does not need a sentiment channel to be affected by an equity options expiry. It needs only a shared risk factor, and realized variance is exactly that. When dealer hedging flows compress or amplify intraday movement in the S&P, the same compression or amplification shows up in the volatility of every asset that trades as a high-beta expression of global risk appetite — which, since 2022, is most of them.

Second, it means that the health of a DeFi lending market and the health of an options book are measurable with the same instruments. Dollar gamma, gamma concentration by strike, the ratio of open interest to daily volume, the share of flow that is mechanically forced rather than discretionary. These are not crypto metrics or TradFi metrics. They are the vocabulary of any balance sheet that is short convexity.

Third, and least comfortable, it means that the LP position is structurally under-compensated in exactly the way that short-volatility positions always are. The premium looks adequate for most of the time and is catastrophically inadequate in the tail. The distribution of outcomes is not a distribution of income; it is a distribution of income with a small probability of total loss attached, and the loss arrives precisely when the LP is least able to hedge it.

The hash is not the art; it is merely the key. The pool is not the yield; it is the premium. Those are the same sentence with different nouns.

TVL Is the Notional of DeFi

Once I understood the LP position as a short straddle, a second pattern became impossible to unsee, and it is the reason this story about a $9.6 trillion headline feels familiar rather than foreign.

Total value locked, the canonical metric of decentralized finance, is a gross number. It counts deposits at every layer of the stack. A user deposits ETH into a lending market, receives a receipt token, deposits that receipt token into a yield aggregator, receives a second receipt token, deposits that into a farm, and at each hop the underlying value is counted again. A thousand dollars of capital can produce three thousand dollars of TVL without anything changing except the number of places the same dollar is written down.

The parallel to notional is exact. Both metrics measure the size of the sandbox. Neither measures the size of the exposure. Both are computed at the layer where they look largest, because the party reporting has an interest in the number being large. Both are quoted without the adjustment that would make them meaningful, because the adjustment requires work and the raw figure requires none.

I ran into the same pattern in 2021, when I spent three weeks examining the IPFS pinning arrangements behind the major profile-picture projects. More than half of the "permanent" assets I traced depended on centralized gateways that were already failing under load, and the on-chain metadata pointed at files whose durability was a function of someone's AWS bill. The marketing said permanent. The infrastructure said until the invoice stops being paid. Roughly sixty percent of the collection was a pointer to a file that could disappear.

What I learned from that research is not that the projects were fraudulent. Most of them were sincere. What I learned is that the metric a system reports about itself is almost never the metric that determines whether the system survives. TVL does not survive a bank run. Notional does not hedge anything. Permanent storage does not persist. Each of them is a claim about a boundary, reported as though it were a claim about a payload.

Deterministic Engines Still Cascade

There is a line of argument that a scheduled expiry cannot be a systemic risk, because it is known in advance. The date is on every calendar. The open interest is public. Every desk has modeled it. If the event were dangerous, the danger would already be in the price.

I spent the worst part of 2022 and the first half of 2023 inside the MakerDAO liquidation engine, reverse-engineering its state machine, because I wanted to understand why a mechanism whose parameters were fully public and fully deterministic could still produce outcomes that surprised the people who designed it.

The answer was not that the design was wrong. The answer was that determinism of the mechanism does not imply determinism of the outcome. The liquidation engine executes a defined sequence: a vault breaches its collateralization ratio, an auction is triggered, bidders with the right keepers participate, the debt is covered, the surplus or deficit is booked. Every branch of that sequence is auditable. Every threshold is published.

What is not deterministic is the state of the world in which the sequence executes. When a correlated shock arrives, every vault breaches at once, every auction fires simultaneously, the keeper population that was adequate for one liquidation is overwhelmed by a thousand, and the auction clearing price collapses because there are more sellers than there is bid depth. The engine did exactly what it was specified to do. The specification assumed a world that stopped existing.

The same structure applies to a quarterly options expiry. The hedging rules of every market maker are known, at least in form. The obligation to arrive at the terminal date with no residual exposure is contractual. What is not known is the depth of the book on the other side of that obligation, and the depth of the book is a function of who else is being forced to trade in the same direction at the same moment.

The 2022 analysis I wrote on debt ceilings during liquidity crunches made a narrow point that generalizes: a ceiling is a supply constraint, and a supply constraint matters only relative to demand at the worst possible moment. If every participant is contractually obliged to transact on the same day, the constraint binds on that day and not before. The opacity is not in the mechanism. It is in the coincidence.

A Curve With a Vote Attached

While I am on the subject of mechanisms that are published and still fail, it is worth stating what the DeFi lending market does with interest rates, because the contrast with the options world is instructive and almost never drawn.

Aave and Compound do not discover interest rates. They compute them. The model is a piecewise-linear function of utilization — the ratio of borrowed assets to supplied assets — with a small number of governance-set parameters: a base rate, a slope up to an optimal utilization point, and a steeper slope beyond it. The kink at the optimal point exists to make borrowing expensive when liquidity is scarce.

That is the whole mechanism. There is no term structure, no forward curve, no credit spread, no market-clearing process in which borrowers and lenders converge on a price. There is a line on a chart, and the line is where it is because a governance vote put it there. When utilization drifts above the kink, the rate rises, but it rises on a schedule that was chosen when nobody knew what utilization would be.

Compare that to the volatility surface, which is the analogue in the options world. Implied volatility is not set by a committee. It emerges from continuous two-sided flow across strikes and expiries, and the surface it produces encodes the market's collective view of the distribution of future returns. It is flawed, it is noisy, it is sometimes wrong in ways that cost people their capital — but it is discovered rather than declared.

A curve with a vote attached is not a market. It is a policy with a decimal point.

I do not say this to indict DeFi lending, which does a difficult job under real constraints. I say it because the current sideways market has made the difference visible in a way that a trending market hides. When prices move fast, every lending market looks fine, because borrowers are either comfortably collateralized or already liquidated. When prices chop, utilization drifts toward the kink and stays there, and the rate that borrowers pay stops being a signal about scarcity and becomes a number that the protocol's governance set eighteen months ago and has not revisited.

In that environment, the spread between what lenders earn and what the risk actually costs widens without anyone announcing it. Which is precisely the failure mode of a short-volatility position: the premium looks stable right up until the moment the accounting catches up.

What to Actually Watch, and What Cannot Be Watched

If I were running a book into a quarterly expiry of this size, and I am not, these are the observations that would carry information, ordered by the ratio of signal to effort.

The first is the sign of dealer net gamma. It is the highest-value variable and the least observable. Every public estimate of it is an inference built on an assumption about who is long and who is short, and the assumptions differ enough that the estimates are not comparable across providers. In the absence of a genuine disclosure, the practical approach is to treat the sign as unknown and to size positions for both regimes. That is unsatisfying, and it is honest.

The second is the volatility term structure. A front-end that trades at a premium to the back is the market pricing a near-term event. A flat or inverted front-end into a large expiry is the market pricing an event it expects to be absorbed. The term structure is the closest thing available to a public statement of positioning, because it is a price rather than an inference.

The third is the share of zero-day volume on the day itself. If it exceeds half of total index option volume, the intraday mechanics are dominated by hedging of instruments that expire before the close, and the realized volatility of the session is partly manufactured rather than discovered. That is a measurable, observable, actionable number, and it is the one variable in this entire analysis that a retail participant can actually watch in real time.

The fourth is the post-expiry print. The release of pinning gamma shows up in the days that follow, not on the day itself. A jump in realized volatility in the T+1 to T+5 window, with no corresponding news, is the fingerprint of a hedging regime that switched off.

The fifth is the cross-asset correlation into the event. If crypto and equities move together on expiry day with an unusually tight beta, the transmission is running through the risk-appetite channel, and the crypto market is a leveraged expression of an equity volatility event. If they decouple, the transmission is idiosyncratic and the story is different.

What cannot be watched, and what the headline implies can be watched, is the $9.6 trillion itself. It is a constant. It will be true regardless of what happens. It is the least informative number in the entire narrative, and it is the only one that was published.

The Blind Spots

Now the part the coverage will not tell you, and the part I would be dishonest to leave out.

The second-order problem with a number like this is not that it is wrong. It is that it is directionally seductive in a specific way. A large number implies a large consequence, and the implication bypasses the analytical step where the consequence would have to be demonstrated. Any reader who absorbs the $9.6 trillion before they absorb the delta adjustment has already formed a conclusion that the delta adjustment will not dislodge, because the revision is effortful and the headline is not. This is the mechanism by which a technically accurate disclosure produces a technically inaccurate market belief. My 2017 experience with the Golem overflow taught me the shape of this failure: a number can be arithmetically valid and economically meaningless, and the market will almost always prefer the version that requires no qualification.

The reflexive-source problem runs deeper than bias. Citadel Securities does not merely have a perspective on the options market. Its balance sheet is a load-bearing component of the options market. If a meaningful share of the hedging obligation created by a $9.6 trillion expiry lands on the books of a small number of firms, then a statement about the rising influence of the options market is a statement about the rising influence of those firms. Nothing in the disclosure is false. The framing simply omits that the speaker is the subject.

The concentration problem is the one I would rank highest if I had to pick a single systemic concern, and it is structurally invisible in a notional figure. A market whose intraday volatility is generated by the hedging behavior of market makers is a market whose microstructure depends on a handful of balance sheets remaining solvent, liquid, and willing to warehouse risk through a stressed session. Notional measures how much paper is outstanding. It says nothing about how many firms stand behind it. If the answer is five, the fragility is not in the $9.6 trillion. It is in the denominator nobody reports.

The financing-layer problem is the one that gets the least attention and deserves the most. Options expiry is a well-modeled event. Margin, repo, and prime brokerage are not. Every large derivatives position has a funding leg, and the funding leg is where shocks actually propagate — through a margin call, a haircut revision, or a counterparty stepping back from a book it no longer wants to finance. A perfectly hedged option book can still fail if its financing evaporates, and the notional figure provides exactly zero information about whether it will.

The cross-jurisdictional layer is where this connects to a story the crypto press covers constantly and understands intermittently. The reason derivatives infrastructure has become the strategic prize in every major financial center is that derivatives are where the margin lives, and margin is where the jurisdiction captures the order flow. When Hong Kong tightens its virtual asset licensing regime, the operative question is not whether the rules protect investors — it is which venue's clearing and margin business migrates where. The prize is not the tokens. The prize is the layer that prices them. That is why the licensing regimes keep tightening and the exchange licenses keep being granted, and why an options expiry in the United States is a matter of competitive interest to a regulator in Asia.

And there is a cautionary parallel sitting in the middle of the crypto industry that nobody wants to draw. The Lightning Network has been the industry's flagship scaling solution for years, and it has spent most of that time failing at routing, bleeding channels to management complexity, and remaining structurally inaccessible to anyone who is not running a node. The lesson is not that the idea was bad. The lesson is that an infrastructure layer can be technically sound, widely discussed, and economically peripheral at the same time. The industry's actual economic activity migrated to perpetual futures and options — instruments with real margin, real liquidation engines, and real volume — while the payment channels stayed a research program with a website.

That migration is why a crypto outlet covers a U.S. equity options expiry. The center of gravity of this industry moved to derivatives years ago, and the coverage followed the money. The notional is not the risk. But it is where the readers are.

The Takeaway

Here is my forward-looking judgment, stated as a forecast rather than a summary, because a summary would be worthless after six thousand words of analysis.

The next structural shock in this market will not come from an expiry date. It will come from the interaction between two concentrations that no notional figure captures: the concentration of zero-day volume, and the concentration of the balance sheets that warehouse it. The expiry itself is scheduled, modeled, and priced. The fragility is in the fact that intraday realized volatility now depends on the continuous hedging behavior of a small number of firms, in instruments that expire before the close, at a moment when the cost of funding those positions is itself a variable that nobody publishes.

My second judgment is about crypto's position in this structure, and it is the one I would stake the most on. The industry will keep importing its volatility regime from U.S. equity option flows as long as its own derivatives markets lack a native term structure with depth comparable to the surface it depends on. Deribit's quarterly expiry is not a smaller version of the S&P event. It is a thinner book with more concentrated positioning and faster liquidation cascades, which means the same mechanics arrive earlier and hit harder. Importing volatility is not the same as having a volatility market, and the difference is only visible in the tail.

Both of these judgments have the same defect: they are directionally confident and quantitatively imprecise, which is the appropriate state of belief when the underlying data is a single number with no year attached.

So the question I keep returning to is not whether September 18 mattered. It is this: if the market's most important mechanical flows are generated by the hedging of instruments that expire in hours, and if the balance sheets generating those flows are not disclosed in any figure the market reports about itself, then what exactly is the market pricing — and on whose behalf?

The hash is not the art; it is merely the key. The expiry is not the risk; it is merely the date. And a market that reports the boundary while hiding the payload is not being opaque about its size. It is being opaque about its structure.