# Short Squeeze AI Stocks: Testing the Thesis Against Fintel Data and SEC Filings Short interest across the AI and quantum complex runs from about 1% to 22% of float, and almost none of it is squeezable. Shayne Heffernan checks $NVDA, $GOOGL, $AMZN, $PLTR, $SMCI, $SPCX, $IONQ, $RGTI and $QBTS against the settlement data. Canonical HTML: https://www.livetradingnews.com/short-squeeze-ai-stocks-testing-the-thesis-against-fintel-data-and-sec-filings Last modified: 2026-08-04 --- By Shayne Heffernan ยท 2026-08-04 Tags: short squeeze, short squeeze AI stocks, AI stocks, short interest, days to cover, $IONQ, $RGTI, $QBTS, $SMCI, $PLTR, $NVDA, $GOOGL, $AMZN, $SPCX, Regulation SHO, fails to deliver, SEC filings, SpaceX lockup, quantum stocks, Shayne Heffernan Signed: ML-DSA-65, anchored on Armature L1. Nothing in this article is investment advice. The US equity market is carrying two incompatible stories at once. On the surface, capital is being deployed into artificial intelligence and quantum computing at a scale with no precedent. Underneath it sits a large population of short sellers betting that those valuations are a bubble with a date on it. The obvious question is whether the second group is trapped. Heavily shorted stocks in a rising sector are the classic setup for a squeeze, and the AI complex looks like it should qualify. So I went and checked, ticker by ticker, against the actual settlement data rather than the narrative. The answer is not the one the short squeeze thesis wants. Short interest across AI and quantum is genuinely elevated as a percentage of float, in some cases four to twenty times the level of the mega caps. But almost every one of these names fails the test that actually matters, which is whether the shorts can get out. Days to cover sit at two to three days across the complex. Borrow is cheap. Shares are available. That is a crowded trade, not a cornered one. And the largest supply event in this group this week is not a squeeze at all. It runs in the opposite direction, and it belongs to a ticker most of the short squeeze commentary has been misreading. ![Short interest as a percentage of float against days to cover for the main US-listed AI and quantum names in August 2026, showing that the quantum stocks carry the highest percentages while every name in the complex can be covered inside three days of normal volume.](https://kxco-cloud.lon1.digitaloceanspaces.com/public/blog/ai-short-interest-scoreboard-2026-08.svg) ## Short volume is not short interest, and the difference decides everything Most short squeeze arguments collapse at the first step, because they treat four different data sets as one. When a dashboard shows a stock at 55% short volume, that is the share of one day's publicly reported trades that were marked short. It is not a measure of how many shares are held short. Market makers and high frequency desks sell short and buy back within the same session as a matter of routine plumbing, and every one of those sales prints as short volume. FINRA says so directly in its own guidance, [Short Interest, What It Is, What It Is Not](https://www.finra.org/investors/insights/short-interest), and its [daily short sale volume files](https://www.finra.org/finra-data/browse-catalog/short-sale-volume-data/daily-short-sale-volume-files) carry the same warning: the daily volume data excludes activity that is not publicly disseminated, is not consolidated with exchange data, and is not the same data set as short interest. Here is what each source can and cannot support. Data set Who publishes it Frequency What it actually tells you Short sale volume FINRA (TRF, ADF, ORF) plus each exchange Daily, by 6pm ET Share of publicly reported trades marked short that day. Mostly intraday plumbing. Says almost nothing about positioning Short interest Brokers report to FINRA Twice a month, mid month and month end, due 6pm ET on the second business day after the settlement date Actual shares held short. The real positioning number, but always two to three weeks stale Days to cover Derived, short interest divided by average daily volume Follows short interest How long it would take shorts to buy back at normal volume. The single best proxy for whether a squeeze is mechanically possible Fails to deliver SEC, from NSCC data Twice monthly file, published on a lag Settlement failures by security. Evidence of borrow scarcity when persistent, not when spiky Cost to borrow Securities lending desks, aggregated by data vendors Continuous What shorts are paying to stay short. Rising fees mean supply is tightening. Cheap borrow means it is not The combination that matters is short interest that is high as a share of float, days to cover in the high single digits or above, a borrow fee that is climbing, and persistent fails. One of those four on its own means very little. All four together is the setup that produced the squeezes people remember. ## What the SEC filings tell you, and what they do not Filings are useful for the long side of the ledger and for intent. They are close to useless for measuring short positioning, because the United States has no equivalent of a short position disclosure regime for equities. **Form 13F.** Filed quarterly by managers with at least $100m in qualifying assets. It shows long holdings in 13F securities, plus options positions. It does not show short stock positions, which is the single most misunderstood point in this whole area. A fund can be long $2bn of a name in its 13F and short more than that through swaps and options with no disclosure obligation at all. When you see a large 13F long in $GOOGL alongside heavy short volume in the same ticker, the most likely explanation is a hedged or relative value book, not a directional bet. **Forms SC 13D and SC 13G.** Filed on crossing 5% beneficial ownership, with 13D signalling activist intent and 13G signalling passive. Worth watching, because a concentrated activist position shrinks the effective float, and a shrinking float is one of the few things that genuinely tightens borrow. **Form 4.** Insider transactions. Useful for reading internal conviction, though systematic sales under Rule 10b5-1 plans carry almost no information, and at large caps a buyback programme absorbs far more supply than insiders release. **Regulation SHO.** This is the part worth understanding properly, because it is where the mechanical force in a squeeze actually comes from, and it is routinely described wrongly. A short seller must deliver by settlement, which since May 2024 is T+1. If delivery fails, [Rule 204](https://cdn.cboe.com/resources/regulation/circulars/regulatory/RG12-022.pdf) requires the clearing participant, meaning the clearing broker rather than the depository, to close the fail by purchasing or borrowing shares of like kind and quantity, generally by the start of regular trading hours on the settlement day after the fail arose. If it does not, the penalty is a pre borrow requirement: that participant, and every broker dealer clearing through it, cannot accept or effect a short sale in that security without first borrowing or arranging to borrow. Separately, securities with large and persistent fails land on the threshold securities list. That penalty is the real pressure mechanism. It does not force a buy at any price, and there is no clearing house that goes into the market on shorts' behalf. What it does is progressively remove the ability to add new shorts in a name where fails are not clearing, which starves the trade of the fresh supply it needs to defend a position. That is a slower and less cinematic process than a forced buy in, and it is the one that is actually in the rules. ## The scoreboard: how shorted are AI stocks, actually Here is the current picture across the names people ask about, using the most recent FINRA settlement data as aggregated by Fintel. Ticker Company Short interest As a share of float Days to cover Squeeze mechanics [$IONQ](https://fintel.io/ss/us/ionq) IonQ about 22.4% of float Highest in the complex About 2 Percentage yes, mechanics no. Borrow fee 0.46% $RGTI Rigetti Computing about 15% of float Very high Low Same pattern as IonQ $QBTS D-Wave Quantum about 14.25% of float Very high Low Same pattern as IonQ [$SMCI](https://fintel.io/ss/us/smci) Super Micro Computer 97.1m shares at the 15 July settlement, up from 81m in May 13.51% of shares outstanding at the May reading 3.0 Highest absolute in AI hardware, but covers in three days [$PLTR](https://fintel.io/ss/us/pltr) Palantir Technologies 76.4m shares, up 10.5% on the period 3.5% of float 2.0 Rising conviction, no scarcity [$GOOGL](https://fintel.io/ss/us/googl) Alphabet 79.66m shares 1.37% of float Under 1 Un-squeezable [$NVDA](https://fintel.io/ss/us/nvda) NVIDIA 284.7m shares 1.22% of float Under 1 Un-squeezable [$AMZN](https://fintel.io/ss/us/amzn) Amazon 106.6m shares 1.09% of float Under 1 Un-squeezable Two things fall out of that table immediately. The first is that percentage of float and days to cover point in different directions here, which is unusual. Normally a stock with 22% of its float short has a days to cover in the high single digits, because the position is large relative to how much the stock trades. In the quantum names it does not, because 2026 turned them into some of the most actively traded stocks on the tape. Enormous short positions sit on top of enormous volume, and the ratio stays low. The second is that the AI hardware complex is far less cornered than its reputation. $SMCI carries the largest short position in the group in absolute terms and it has been growing, from 81m shares in May to 97.1m in July. But average daily volume of roughly 45.7m shares means the whole position clears in about three days. Days to cover was 1.78 in May and 3.0 in July. That is a rising number, and it is worth watching, but it is nowhere near the double digit readings that define a genuine trap. ## The mega caps are un-squeezable by construction Nobody shorting $NVDA, $GOOGL or $AMZN at roughly 1% of float thinks these companies are going away. At that size the short book is hedging, index-relative positioning, and volatility expression. $NVDA is the market's liquidity instrument for the entire theme. If a fund wants to reduce AI beta without selling positions it wants to keep, it shorts Nvidia, because that is where the depth is. Nothing about 284.7m shares short against a float that size is a squeeze candidate. $GOOGL is the one mega cap with a genuine fundamental short case, and it is legal rather than operational. Judge Mehta's search remedies [took effect on 3 February 2026](https://www.litigationlogic.io/legal-news/google-search-antitrust-remedies-appeal-2026/), and both sides are now appealing. Google filed its notice in January 2026, aiming primarily at the data sharing requirements. The DOJ and a coalition of state attorneys general [filed notices of appeal in early February](https://searchengineland.com/doj-states-appeal-google-search-antitrust-remedies-ruling-468230), cross-appealing for the structural relief Mehta declined to order, including Chrome divestiture and termination of the Apple default deal. The D.C. Circuit has not set a briefing schedule, so this runs well into 2027. That is a real overhang, and it is also the reason the short is not a squeeze setup. A multi year appellate process does not produce the single catalyst that forces covering. If Google wins on the data sharing point, the re-rating is a slow grind that costs shorts money over quarters, not a gap that liquidates them in a session. $AMZN shorts are expressing a view on the capital expenditure cycle rather than on Amazon. The argument is that AWS is committing enormous sums to AI infrastructure whose monetisation is unproven, and that the depreciation lands before the revenue does. It is a defensible position. At 1.09% of float it is also irrelevant to any squeeze discussion. ## $PLTR: rising conviction, no scarcity behind it Palantir is the most argued-over stock in the market, and the short book reflects that. Short interest is 76.4m shares, up 10.5% on the prior period, at 3.5% of float. Days to cover is 2.0, up about 14% on the period and up roughly 101% over the past year. That last figure is the interesting one, and it is the one the bulls should watch rather than the headline percentage. Days to cover doubling year on year means the short position is growing faster than the liquidity available to exit it. That is the direction a squeeze setup develops from. At an absolute level of 2.0 days it is still nowhere near dangerous, but the trend is the right shape. The structural argument for $PLTR is not really about short interest. It is that the long base is unusually concentrated and unusually price insensitive, which means the marginal seller in a rally is thin on the ground. That produces sharp, short-lived gaps on contract news. It does not produce a multi day forced unwind, because at 3.5% of float there simply is not enough short exposure to fuel one. You can [follow the ticker here](https://www.livetradingnews.com/ticker/PLTR). ## $SMCI: the biggest short position in AI hardware, and the shortest exit If any AI hardware name is going to squeeze, it is [$SMCI](https://www.livetradingnews.com/ticker/SMCI). It carries the largest short position in the complex and the fundamentals are volatile enough to deliver the catalyst. The most recent print showed why both sides have a case. Q3 FY2026 revenue was $10.24bn, up 123% year on year but roughly 17% below the $12.33bn consensus, while non-GAAP EPS of $0.84 beat the $0.62 estimate by 35% as gross margin nearly doubled sequentially to 10.1%. Management raised the FY2026 revenue outlook to $38.9bn to $40.4bn and guided Q4 to $11bn to $12.5bn, both above expectations, with Nvidia GPU platforms above 80% of sales. The stock rose 18% to 19% on the print. That is the profile of a stock that can move violently on news while remaining easy to short. A 19% move is a squeeze in colloquial terms. It is not a short squeeze in the mechanical sense, because the shorts who wanted out got out inside three days at prices they chose. The honest position on fails to deliver here is that anyone claiming persistent, structural FTDs in this name needs to point at the SEC fails file and show the run of days, not assert it. Spiky fails are ordinary settlement noise. Persistent fails put a security on the threshold list, which is a published, checkable fact rather than an inference. I am not going to assert a borrow squeeze in $SMCI on the basis of a chart pattern, and the borrow data does not currently support one. ## $SPCX is SpaceX, and it is the opposite trade This needs stating plainly, because it is circulating widely in the wrong form. **$SPCX is Space Exploration Technologies Corp, the SpaceX listing. It is not a proxy, a derivative wrapper or a data-stream identifier for Super Micro Computer.** Super Micro is $SMCI. Anyone building a short squeeze thesis on $SPCX as an AI server play is analysing the wrong company. What $SPCX actually is happens to be far more interesting this week, and it is the reverse of a squeeze. SpaceX listed on 12 June 2026, priced at $135, and ran more than 67% to a peak above $225, briefly touching a valuation around $2.66 trillion. Less than 5% of total shares were made available at launch, which is exactly the thin float condition a squeeze needs. The problem for that thesis is what happens next, because the lock-up is not a single 180 day cliff. It is a staged release keyed to earnings windows and fixed dates. Date Event 12 June 2026 IPO priced at $135, under 5% of shares in public float 4 August 2026 First quarterly results as a public company, after the close 6 August 2026 About 20% of restricted holdings unlock, roughly 911.5m shares, close to $123bn at recent prices August to October 2026 Further tranches of about 7% every two to four weeks Q3 2026 earnings Earnings-triggered release of roughly 28% 8 December 2026 All remaining 180 day shares release A stock two days away from [$123bn of insider supply becoming saleable](https://www.investing.com/news/stock-market-news/spacex-ipo-lockup-expiry-123b-in-shares-set-to-unlock-in-early-august-2026-93CH-4796311) is the textbook opposite of a cornered short. Unlocks resolve the scarcity that makes squeezes possible, and they hand short sellers the borrow they were previously struggling to find. If you are looking for the dominant supply and demand event in this group in early August 2026, it is this, and it points down rather than up. Our [space coverage is here](https://www.livetradingnews.com/center/space). ## Quantum is where the percentages live: $IONQ, $RGTI, $QBTS The quantum names carry the highest short interest as a share of float anywhere in this complex. $IONQ is around 22.4%, $RGTI around 15%, $QBTS around 14.25%. On the percentage screen alone, quantum wins the most-shorted title outright, and it is not close. The percentage is where the squeeze case stops. $IONQ's borrow fee is 0.46% annualised, with about 1.6m shares available to borrow. That is not a hard-to-borrow stock. Shorting IonQ is close to free. Days to cover is around two, because volume in 2026 has been enormous. A short position of 22% of float that can be exited in two days at a 0.46% carry is a crowded position, and a crowded position is a very different animal from a trapped one. This matters because the "impossible to roll, borrow fees above 50%" version of the quantum short story is not what the data says. Those conditions do exist in small caps from time to time. They are not present in $IONQ, $RGTI or $QBTS right now. What is present is a large, cheap, liquid short base against stocks with minimal revenue and heavy cash burn, which is precisely the position a rational fundamental short should want to hold. Our [quantum computing coverage is here](https://www.livetradingnews.com/center/quantum-computing). The genuine risk to those shorts is not mechanical, it is headline risk. These are companies whose valuations respond to error correction milestones, government awards and partnership news, and the moves on that news are large. Losing 40% on a position is painful whether or not you were forced out of it. But that is volatility risk, correctly priced by the options market, not a structural trap. ## What a real short squeeze requires, and where this complex fails ![The five conditions a short squeeze requires, shown as sequential gates with a pass or fail verdict for the US-listed AI and quantum complex in August 2026, where high short interest passes but days to cover, borrow scarcity, persistent settlement fails and float shrinkage all fail.](https://kxco-cloud.lon1.digitaloceanspaces.com/public/blog/short-squeeze-conditions-2026-08.svg) A short squeeze is a supply problem, not a sentiment problem. Five conditions have to hold at once. Condition Why it matters AI and quantum complex, August 2026 High short interest relative to float Provides the fuel **Pass.** 14% to 22% in quantum, 13.5% in $SMCI Days to cover in high single digits or above Means shorts cannot exit at prices they choose **Fail.** Two to three days across the complex Rising cost to borrow Signals lending supply is tightening **Fail.** $IONQ at 0.46% with shares available Persistent fails to deliver, threshold listing Proves shares physically cannot be located **Not evidenced.** Assert it only from the SEC fails file Shrinking effective float Removes the shares shorts need to buy back **Fail.** $SPCX is adding float, not removing it One of five. That is a crowded short trade in a strong sector, which produces sharp rallies and painful drawdowns for shorts. It is not the mechanical trap the phrase "short squeeze" describes. ### If the conditions were met, this is how it would run The sequence is worth knowing, because conditions change and $PLTR's days to cover doubling in a year is exactly how the picture starts to shift. **Recognition.** Short interest rises while institutional accumulation also rises, and the float available to lend starts to tighten. This is visible in the twice monthly short interest file and the borrow desks well before it is visible in price. **Catalyst.** Something forces a re-rate. A contract award, a capex guide raise from a customer, an earnings beat that invalidates the thesis. **Margin pressure.** As the mark moves against the position, prime brokers raise requirements. Where days to cover is high, shorts cannot exit at their own pace and end up competing with each other for the same shares. **Gamma and settlement.** Options market makers who are short calls buy the underlying to stay hedged, which adds forced demand on top of the covering. If fails are persistent, Rule 204's pre borrow penalty starts blocking new short supply from entering to absorb it. **Price discovery.** For a period, price is set by the supply and demand of trapped sellers rather than by valuation. This is the phase that produces the moves people remember, and it ends when the position is cleared. Every step past the first depends on days to cover. That is the gate, and it is the gate this complex does not currently pass. ## So which AI stock is the most shorted? The verdict **By short interest as a share of float, the answer is $IONQ at roughly 22.4%,** followed by $RGTI at about 15%, $QBTS at about 14.25%, and $SMCI at 13.51% of shares outstanding at its May reading. $PLTR at 3.5% is a distant fifth. The mega caps at 1.09% to 1.37% are not in the conversation. **By squeeze risk, the ranking is different, and much flatter.** No name in this group currently clears the days to cover test. The closest thing to a developing setup is $SMCI, where short interest rose from 81m to 97.1m shares and days to cover went from 1.78 to 3.0, and $PLTR, where days to cover has doubled year on year off a low base. Both are trends worth tracking. Neither is a trap today. **By absolute dollars at risk, it is the mega caps,** because 284.7m shares of $NVDA is an enormous sum of money regardless of how small a fraction of the float it represents. Those positions will not squeeze. They will be lost or won slowly, on fundamentals. The useful conclusion is not that the AI trade is an anti-short trade. It is that the AI trade is currently a well-functioning market, where shorts can express a view and exit it, and where the upside case has to be argued on cash flows rather than on plumbing. Bulls do not need a squeeze. Betting on one here means betting on a mechanism the data says is not in place. ## Questions readers are asking about short squeeze AI stocks ### Which AI stock has the highest short interest in 2026? $IONQ, at roughly 22.4% of float, is the most shorted name in the AI and quantum complex. $RGTI is around 15% and $QBTS around 14.25%. In AI hardware, $SMCI is the leader at 13.51% of shares outstanding at its May 2026 reading, with the position having grown to 97.1m shares by the 15 July settlement. ### Is a short squeeze in AI stocks likely right now? On the current data, no. Every name in the group can be covered in about two to three days of normal volume, borrow costs are low, and shares are available to borrow. High short interest without borrow scarcity produces sharp rallies, not forced unwinds. ### Is $SPCX a Super Micro Computer proxy? No. $SPCX is Space Exploration Technologies Corp, the SpaceX listing, which came public on 12 June 2026. Super Micro Computer trades as $SMCI. The two are unrelated companies in unrelated industries, and confusing them will produce an analysis of the wrong balance sheet. ### What is happening with the $SPCX lock-up? SpaceX reported its first results as a public company after the close on 4 August 2026. Two trading days later, on 6 August, about 20% of restricted holdings unlock, roughly 911.5m shares worth close to $123bn at recent prices. Further tranches of about 7% follow every two to four weeks into October, with a roughly 28% release triggered by Q3 earnings and the remaining 180 day shares releasing on 8 December 2026. ### Is $PLTR a short squeeze candidate? Not on today's mechanics. Short interest is 3.5% of float with days to cover at 2.0. The number worth watching is the trend: days to cover has roughly doubled over the past year, which means the short position is growing faster than the liquidity available to exit it. That is how a setup develops, and it is not one yet. ### Why does short volume look so high if short interest is low? Because they measure different things. Short volume counts publicly reported trades marked short on a given day, including the intraday buying and selling that market makers and high frequency desks do as routine plumbing. Short interest counts shares actually held short, is reported twice a month, and is the only one of the two that describes positioning. FINRA publishes an explicit warning about the distinction. ### Do 13F filings show short positions? No. Form 13F covers long holdings in 13F securities and options positions. There is no general short position disclosure requirement for US equities, so a fund's public filings can show a large long while its actual net exposure is flat or negative through swaps and derivatives. ## Disclosure Live Trading News and KXCO are part of the same group. This article references the KXCO ontology, which is the relational mapping method KXCO applies to compute and AI supply chain data, published at [kxco.ai/ontology](https://kxco.ai/ontology) with the working graph at [kxco.ai/ontology-live](https://kxco.ai/ontology-live). That ontology maps entities, claims and chokepoints in the compute supply chain. It does not track short interest, fails to deliver or options positioning. The short interest analysis in this article was assembled by hand from the sources cited above, and the same relational discipline was applied to it: every figure is attached to a publisher and a date, and claims that could not be sourced were dropped rather than softened. More on the method at [kxco.ai](https://kxco.ai). Further AI equity coverage is in the [AI Stocks Center](https://www.livetradingnews.com/center/ai-stocks). ## Disclaimer This report is for informational and educational purposes only and reflects the analytical opinions of Shayne Heffernan regarding market mechanics and data structures. It does not constitute financial advice. Short interest, days to cover and borrow data are point in time figures that change with each reporting cycle, and the readings quoted here carry the dates given. Short selling and investing in volatile AI and quantum stocks carry significant risk of capital loss. Always conduct your own due diligence. --- This Markdown mirrors https://www.livetradingnews.com/short-squeeze-ai-stocks-testing-the-thesis-against-fintel-data-and-sec-filings. The HTML page is canonical. Site index for AI clients: https://www.livetradingnews.com/llms.txt