GraniteShares 2x Long NVDA Daily ETF
GraniteShares — Will Rhind (founder/CEO)
Single-stock leveraged ETF seeking +200% (2x) of NVIDIA's DAILY return via swaps. A short-term trading instrument — NOT a long-term hold: compounding/volatility drag means it does not deliver 2x NVDA's cumulative return over periods longer than a day. Expense ratio 1.05%.
- Levered — Leverage is material to outcomes, including on the downside.
Top holdings
~200% notional — Entire fund is leveraged swap + collateral exposure to a single name — NVIDIA — not a diversified basket
Recent moves
No discretionary holdings changes — exposure is mechanically reset daily to 2x NVDA via swaps; AUM has stayed large (~$5.5B) on persistent retail demand for leveraged NVIDIA exposure.
Our take
The most popular single-stock leveraged ETF and a pure high-octane bet on NVIDIA, not a thematic AI fund. Useful only as a tactical, actively-monitored trade — the daily-reset decay punishes buy-and-hold in choppy tape. Included as the iconic leveraged-NVDA proxy, with the strong caveat that it's a trading tool, not an investment.
The strategy's logic is internally coherent and it does what it says — our skepticism is not that NVDL is broken, it's that the number the reader is shown and the number the reader earns are different numbers, and the wrapper does nothing to close that gap.
Three things compound against the holder simultaneously and only one of them appears in the fee table. First, volatility decay: NVIDIA is a high-volatility underlying in its own right (NVDA 5-year beta 2.21, stockanalysis.com, Jul 20 2026), and decay scales with the SQUARE of volatility, so one of the most volatile large-caps in the market is a poor candidate for daily-reset leverage even though it is among the most popular. Second, the 1.05% expense ratio (stockanalysis.com, Jul 20 2026) is charged on 2x exposure. Third — and this is the cost fund marketing never surfaces — 2x exposure means roughly 1x notional is financed inside the swap at approximately the overnight rate, and SOFR was 3.59% as of Jul 17 2026 (sofrrate.com). That financing does not appear in the expense ratio; it is netted inside the swap return. So the honest all-in annual carry is closer to ~4.6% than the advertised 1.05%, BEFORE any decay. A reader who thinks they are paying 1.05% is off by roughly 4x. Separately, the marketing-friendly 'average annual return since inception of 139.63%' (stockanalysis.com, Jul 20 2026) should be read as an artifact, not a track record: it spans the 2023-24 AI melt-up AND a mid-life leverage change from 1.5x to 2x on Jan 22 2024. It describes a regime that no longer exists measured partly under a mandate that no longer applies.
Thesis
NVDL is not an NVIDIA fund with extra kick — it is a bet on NVIDIA's PATH, not its destination. The product does exactly one thing: reset to 2x NVIDIA's DAILY return via swaps, every single day. That daily reset means the holding period return is a function of realized volatility as much as direction, and the fund's own prospectus says so plainly: it 'should not be expected to provide 2 times the cumulative return of NVDA for periods greater than a day' (GraniteShares NVDL fund page, accessed Jul 21 2026). The single most useful fact a reader can have is the live scoreboard: over the 52 weeks to Jul 20 2026, NVDA returned +17.91% while NVDL returned +10.18% (both stockanalysis.com, Jul 20 2026; bestetf.net reports +11.40% for NVDL as of Jul 18 2026 — see our source-conflict note). A 2x long fund delivered roughly 0.57x the underlying on the primary source, and under 0.64x on the alternative — either way it underperformed the unleveraged stock while carrying double the risk. That gap, not the leverage multiple, is the actual product.
Daily-reset 2x (200%) leveraged exposure to a single equity, NVIDIA, obtained synthetically. The fund holds no meaningful economic position in NVDA shares directly; it holds swap exposure (plus cash collateral and a small equity sleeve) sized so that a 1% NVDA move produces a 2% NAV move that day. At each close the notional is rebalanced back to 2x current NAV. This rebalance is mechanical and non-discretionary — there is no manager view, no timing, no hedging. Structurally it forces the fund to buy exposure after up days and sell after down days, which is a systematic buy-high/sell-low at the daily frequency. That is the source of volatility decay, and it is a feature of the mandate, not an execution failure. Inception Dec 13 2022 — and critically, the fund LAUNCHED at 1.5x (150% of NVDA's daily change) and changed to 2x (200%) effective January 22, 2024. The track record is therefore a composite of two different mandates.
- 2026-07Updated to Jul 31 2026: NVDA's 1-year return is +13.01% while NVDL's is -1.26% (stockanalysis, Aug 1) or +0.34% on Yahoo adjclose. Capture has fallen from about 0.57x to roughly zero.
Assessment
- Radical honesty in the primary document — the prospectus states outright that the fund should not be expected to deliver 2x NVDA's cumulative return beyond one day. The structural risk is disclosed at the source; the misunderstanding is downstream of the issuer, not created by it.
- The mandate is executed cleanly and mechanically. There is no style drift, no manager discretion, no hidden active bet. What you get is precisely and only 2x-daily NVDA, which makes the exposure fully predictable at the one-day horizon it is actually designed for.
- Genuine scale and liquidity relative to the category — roughly $3.73B AUM (stockanalysis.com, Jul 20 2026) versus $607M for Direxion's competing NVDU (same source, same date). For a vehicle whose only legitimate use is short-horizon tactical trading, tight spreads and depth are the feature that actually matters, and NVDL has the most of it.
- The swap wrapper delivers leverage inside a '40 Act fund with no margin account, no margin calls, and loss capped at the invested amount. An investor cannot be forced to post collateral or lose more than they put in — a real structural improvement over doing this with margin or options.
- Daily reset is genuinely the right design for the stated purpose. It keeps leverage constant at 2x rather than letting it drift, which is what a one-day tactical trader actually wants.
- The headline result is the critique: NVDL +10.18% vs NVDA +17.91% over the 52 weeks to Jul 20 2026 (stockanalysis.com, both, same date). A naive 2x expectation was roughly +35.8%; the fund delivered roughly 25.6 percentage points less and underperformed the unleveraged stock outright. The shortfall is source-dependent — on bestetf.net's +11.40% (Jul 18 2026) it is ~24.4 points — but the sign and magnitude are the same either way. This is decay made visible, and it is not an anomaly; it is the mechanism working as designed in a choppy tape.
- The advertised 1.05% expense ratio materially understates the true cost of carry. Roughly 1x notional is financed near overnight rates inside the swap (SOFR 3.59%, Jul 17 2026), and that cost is invisible in the fee table because it is netted inside the swap return. All-in drag is plausibly ~4.6% annually before decay. We consider this the single most under-disclosed economic fact about the product — not hidden, but structurally unreportable in the number retail investors compare on.
- An identical-mandate competitor was both cheaper and better over the same window: Direxion's NVDU charges 0.92% and returned +12.77% over 1 year versus NVDL's 1.05% and +10.18% (stockanalysis.com, Jul 20 2026). Same objective, same underlying, same reset frequency — NVDL is the more expensive wrapper (1.05% vs NVDU 0.92%); on 1-year performance the two have since converged and NVDL is marginally ahead, so the fee is the durable gap, not the outcome. NVDL's advantage is liquidity, not cost or realized tracking.
- The since-inception annualized figure (+139.63%) is a composite across a leverage-mandate change — 1.5x at launch, 2x effective Jan 22 2024 — layered on a once-in-a-decade trend. It is not a like-for-like record of the strategy an investor buys today, and quoting it as one flatters the product substantially.
- Path dependency is unforgiving and asymmetric. Daily-reset leverage cannot mean-revert its way back: a drawdown requires a proportionally larger gain to recover, and 2x makes the hole deeper than the underlying's. The fund can be structurally unable to participate fully in a recovery it correctly predicted the direction of.
- Flows point to a transient rather than convicted holder base. Net flows have been negative over the trailing year on every source we could retrieve (bestetf.net: -$1.12B as of Jul 18 2026), but the pattern is violently two-way — large single-week inflows and outflows in both directions through 2026 — which reads as episodic speculation cycling through the vehicle rather than stable ownership. Note the magnitude of trailing-year outflow differs materially across providers; we treat the direction as informative and the level as unresolved.
- Exposure is synthetic and therefore carries counterparty risk to the swap dealers, plus concentration risk in a single name with no diversification whatsoever. Two independent failure modes stack on top of the market risk.
- 2026-08At Aug 1 2026 assets stand at ~$3.19B against NVDU's $527M, both smaller than a fortnight earlier, though NVDL remains much the larger. Yahoo carried $4.05B on Jul 29, so the level is unsettled.
Record
The trailing year is the most instructive window available and it cuts against the naive reading of the product. NVDA rose +17.91% over the 52 weeks to Jul 20 2026, yet NVDL returned +10.18% (stockanalysis.com, both figures, same date; bestetf.net shows +11.40% as of Jul 18 2026 — the conclusion is unchanged on either). Attribution: NVIDIA's direction was positive but its PATH was violent, and daily-reset leverage converts volatility into a direct return drag that scales with variance. Add the 1.05% fee on doubled exposure and roughly 1x notional financed near SOFR at 3.59%, and the roughly 25-point shortfall against a naive 2x expectation is fully explained by structure — no tracking failure or manager error is required. The peer cross-check reinforces this: NVDU, running the identical mandate, returned +12.77% over the same period, so decay hit the category, but NVDL captured less of the move than its cheaper rival did. The since-inception average annual return of +139.63% points the other way and is the figure marketing leads with; it is real but not repeatable, generated by a sustained low-chop uptrend in 2023-24 and measured partly under a 1.5x mandate that ended Jan 22 2024. The correct generalization is not 'decay always wins' — it is that NVDL amplifies trend and is punished by chop, and the trailing year was chop with an upward tilt.
- 2026-07To Jul 31 2026 the peer gap has closed: NVDU's 1-year return is -1.46% against NVDL's -1.26% (stockanalysis, Aug 1). NVDU is still the cheaper wrapper at 0.92% vs 1.05%, but no longer the better performer.
Risks & fit
- Volatility decay is the dominant risk and it is structural, not a scenario. In a sideways-but-violent NVIDIA tape, NVDL can lose money over months while the underlying is flat or up — the trailing year already demonstrated the milder version of this.
- Financing cost is rate-sensitive and uncapped. The ~1x borrowed notional reprices with overnight rates (SOFR 3.59%, Jul 17 2026); a higher-for-longer regime raises the hurdle NVIDIA must clear before a holder breaks even, with no offsetting benefit.
- Gap risk. A large overnight or pre-market move in NVIDIA (earnings, export-control headlines, a supply-chain shock) is amplified 2x at the open with no opportunity to adjust between resets. A ~50% single-day NVDA decline would be an effectively terminal event for fund NAV — a mechanical consequence of the 2x multiple, not a forecast.
- Single-name concentration with zero diversification — every idiosyncratic NVIDIA risk (customer concentration, competitive displacement, regulatory action on China sales) transmits at double intensity.
- Swap counterparty and collateral risk, since the exposure is synthetic rather than physically held.
- Issuer discretion over the mandate itself is a demonstrated risk, not a hypothetical: leverage was already changed once, from 1.5x to 2x effective Jan 22 2024. The terms of the exposure can be altered under the holder.
- Behavioral and holding-period risk — the product is engineered for a one-day horizon and is demonstrably held longer, meaning the realized experience of the typical holder diverges from the designed use case.
Our skepticism rests on the claim that realized volatility in NVIDIA is high enough, often enough, that decay plus carry structurally erodes the leverage benefit over multi-month horizons. The clean falsifier: a sustained period — call it 12 months — in which NVIDIA trends upward with LOW realized volatility, and NVDL's rolling 12-month return meets or exceeds 2x NVDA's cumulative return net of all costs. Daily-reset leverage compounds ABOVE the stated multiple in smooth trends; that is genuine math, not a loophole, and it is roughly what happened in 2023-24. If NVIDIA's realized volatility compressed durably toward market-average levels while it continued to appreciate, the decay critique would weaken sharply and the carry would be the only meaningful drag. Conversely our read is confirmed if NVDL continues posting 12-month returns below the unleveraged stock, as it did to Jul 20 2026. The measurable dividing line is NVIDIA's realized volatility, and it is checkable at any time.
This is exposure designed for a specific and narrow use: expressing a high-conviction directional view on NVIDIA over hours to days, by a trader who monitors the position actively, sizes it as risk capital, and understands that the 2x multiple applies to one day and not to their holding period. The category also serves hedging and pair-trade uses where a leveraged leg reduces the capital tied up. It is structurally mismatched to buy-and-hold, retirement, or set-and-forget accounts — not as a matter of opinion but because the daily reset makes multi-period returns path-dependent, and the issuer's own prospectus says the fund should not be expected to deliver 2x NVDA's cumulative return beyond a single day. Anyone whose actual goal is long-horizon NVIDIA exposure is buying a different risk than the one they intend. Within the 2x-NVDA category specifically, NVDL is the liquidity leader while NVDU is the cheaper wrapper — the trade-off is execution depth versus carry.