USDC vs USDT Funding: One Exchange, Two Bitcoin Rates at the Same Settlement

TLDR. USDC vs USDT funding is not a distinction most dashboards make, and it should be. A Bitcoin perpetual quoted in USDC and a Bitcoin perpetual quoted in USDT on the SAME exchange are two books with two premium indices and two quote-currency reference prices, so they charge two different rates at the same settlement stamp. Measured on 2026-09-21 over the 270 settlements from 2026-06-23 16:00 UTC to 2026-09-21 08:00 UTC: on Binance the two contracts printed an identical rate at only 18 of 270 settlements, the median gap between them was 0.00197 points per 8 hours, and at 12 settlements they charged in OPPOSITE directions. Bybit's pair disagreed on direction at 45 of 270 and Bitget's at 38. The gap is not a rounding artifact. Run the identical arithmetic across VENUES instead, holding the quote currency at USDT, and the three venue-to-venue medians are 0.00244, 0.00284 and 0.00370 points, so changing the unit a Bitcoin perpetual is priced in moves its funding rate by between a half and essentially all of what changing the exchange does. Part of it tracks the stablecoin itself, and only part: the gap correlates with how far the two quote currencies sat apart (r of -0.34 on 270 settlements) and not at all with how fast that distance was moving (r of -0.00). The thinner-book explanation, which the cross-venue ordering seems to support, does not survive a within-venue test.

Does the quote currency change a Bitcoin perpetual's funding rate?

Yes, at almost every settlement, and the size is worth knowing before you quote a venue's funding rate as a single number. Funding on a perpetual is a premium component plus an interest component, and the premium half is measured against that contract's own index price in that contract's own quote currency. Two contracts on the same exchange that differ only in quote currency therefore have separate order books, separate premium indices and separate reference prices, and nothing in the mechanism forces them to agree. On Binance the interest component is not the source of the difference: read on 2026-09-21, the venue publishes the same 8 hour funding interval and the same 0.01 per cent interest rate for BTCUSDT and for BTCUSDC alike.

Read the middle-half column before the median column. On Binance half of all settlements fell between 0.00088 and 0.00354 points, so the median is the centre of a wide distribution and not a constant you can budget for. Annualized at three settlements a day, the median Binance gap of 0.00197 points per interval is 2.16 points a year, the Bitget gap 2.35 points and the Bybit gap 2.66 points. For a live reading of the thing this gap sits inside, Athenum's own cross-venue funding grid at 2026-09-21 08:50 UTC put Bitcoin perpetual funding at an annualized 8.97 per cent on Binance, 9.00 on Bybit and 10.95 on both OKX and Hyperliquid. Four of the five venues in that grid printed a non-zero rate and the remaining one read zero and neutral at that timestamp, which is a reminder that any single-instant cross-venue reading is fragile, and the reason the comparison in the next section is run over the whole quarter instead.

One property of this window belongs in the open, because it changes how the identical settlements read. No settlement on any of the six series printed above 0.01 per cent, which is exactly the interest anchor both Binance contracts carry, while the downside was unconstrained and Bitget's USDT contract reached minus 0.0127 per cent. Every identical-rate settlement in the table sits on that anchor: all 18 on Binance, all 18 on Bybit and all 21 on Bitget. Those are not two books agreeing on a rate, they are two books pinned to the same constant, and nowhere in the sample did the two contracts agree at any other value.

We did not discover that a venue can carry two funding rates. Our own Athenum measurement of funding against price asserted it in passing on 2026-08-28, using Bybit's linear and inverse boards at one settlement, and concluded that a venue is not a board. This post is the size of that claim over 270 settlements, three venues and the pair that assertion named but never measured.

How often did the two contracts charge in opposite directions?

At 12 of 270 settlements on Binance, 38 on Bitget and 45 on Bybit, which means the sign of the payment, not merely its size, depended on which quote currency you traded. The clearest single case is Bybit at 2026-08-28 00:00 UTC, where the USDT-quoted contract charged longs 0.002538 per cent and the USDC-quoted contract PAID longs 0.008977 per cent, the same exchange, the same asset, the same minute, opposite directions. Bybit runs a third Bitcoin board, the coin-settled inverse BTCUSD, and at that same stamp it settled at minus 0.004754 per cent, which is the figure our August post published from the same second: three boards, three different rates, one clock. Bitget produced a wider one at 2026-08-19 16:00 UTC: the USDT contract paid longs 0.0118 per cent while the USDC contract charged them 0.0038 per cent.

Over the same 270 settlements the two contracts took opposite sides 12 times on Binance, 38 times on Bitget and 45 times on Bybit. The venues line up with how lopsided the two books are, from 3.7 times on Binance to 102.5 times on Bybit, but three venues is three points and that ordering is not a measured relationship.

The ordering is suggestive and it is not evidence. Ranked by how much larger the USDT book traded in the 24 hours to the capture, Binance sits at 3.7 times, Bitget at 42.3 and Bybit at 102.5, and the disagreement counts rise in the same order. Three venues is three points. It is the kind of pattern that deserves a within-venue test rather than a sentence, and that test is further down, where it fails.

Is the quote currency worth as much as the exchange?

Almost, and this is the comparison that decides whether any of the above matters. Run the identical arithmetic on the identical 270 stamps, but change what varies: hold the quote currency fixed at USDT and compare one VENUE against another. That holds the estimator, the window and the settlement clock constant, so the two gaps are directly comparable.

Every quote-currency gap is smaller than every venue gap, and not by much. The three venue medians run 0.00244 to 0.00370 points per 8 hours and the three quote-currency medians run 0.00197 to 0.00243, so changing the unit a contract is priced in moves the rate by between a half and essentially all of what changing the exchange does. At the top of that range the two are indistinguishable: swapping BTCUSDT for BTCPERP on Bybit moved the rate a median 0.00243 points, and swapping Bybit's BTCUSDT for Binance's moved it 0.00244. The direction counts say the same thing, 12 to 45 sign flips within a venue against 29 to 55 across venues. Cross-exchange funding dispersion is a staple of derivatives commentary; the dispersion sitting inside a single exchange is not, and it is the same size.

Do the two contracts track the same dollar?

No, and this is the part of the mechanism that is easiest to check for yourself. Each contract prices Bitcoin in its own quote currency, so the USDC contract's index is BTC in USDC and the USDT contract's index is BTC in USDT. Across 95 daily closes from 2026-06-19 to 2026-09-21, Binance's BTCUSDC index sat a median 5.89 basis points BELOW its BTCUSDT index, ranging from 13.55 basis points below to 3.54 above, and it was above on only 9 of those 95 days. A Bitcoin costing fewer USDC than USDT is the same statement as USDT being the cheaper unit, which is exactly what our Athenum 720 hour study of the Coinbase premium and the Tether discount found on the spot books, at a median of 7.58 basis points.

The two contracts reference two different dollars. Across 95 daily closes to 2026-09-21 the BTCUSDC index sat a median 5.89 basis points below BTCUSDT, and the gap narrowed sharply after mid-August and crossed above zero on 9 of the 95 days, ending at plus 3.04 basis points on 2026-09-21 with the USDC index briefly the dearer of the two.

A basis of that size does not mechanically produce a funding gap, which is what makes the next section necessary rather than decorative. Funding's premium component is a RATIO of two quantities denominated in the same currency, so re-pricing both the perpetual and its index in a slightly cheaper unit divides out. Only the part of the stablecoin gap that fails to pass through to both legs equally can reach the funding rate.

How much of the quote-currency gap does the stablecoin explain?

Some of it, and less than half of it. Pairing each of the 270 Binance settlements with the index gap in the 8 hours ending at that stamp gives a correlation of -0.34 between the signed funding gap and the LEVEL of the stablecoin gap, which is an r squared of 0.118. Against the 8 hour CHANGE in that gap the correlation is -0.00. The level matters and the drift does not, which is the opposite of the pass-through story you would write down first.

Split the 270 settlements into thirds by how far apart the two quote currencies sat, and the median funding gap moves monotonically from plus 0.00091 to minus 0.00146 points per 8 hours. Every one of the three middle-half bars reaches zero, so this is a tendency across 90 settlements at a time and not a rule you can apply to the next settlement.

Sorted into thirds by that level, the median funding gap runs from plus 0.00091 points per 8 hours when the two currencies sat furthest apart, through plus 0.00006 in the middle third, to minus 0.00146 when they sat closest, with 90 settlements in each bucket. The direction is consistent and the separation is weak: all three middle-half bars straddle zero, so a single settlement tells you nothing. The anchor does not drive that association either. Dropping the 58 Binance settlements with at least one leg pinned at 0.01 per cent moves the correlation from minus 0.34 to minus 0.30 and leaves the ordering intact, at plus 0.00080, plus 0.00014 and minus 0.00150 points across 70, 72 and 70 settlements. The honest summary is that the stablecoin gap is one input among several, and the remaining seven eighths of the variance is not explained here.

Is the smaller book the explanation?

Not on the evidence here, and this is the test that breaks the tidy story. If the gap were driven by the USDC book being thin, then within a single venue the gap should widen on the days when the USDC contract carries the smallest share of that venue's Bitcoin perpetual turnover. It does not.

Days sorted into thirds by the USDC share of the venue's Bitcoin perpetual turnover. Binance is flat at 0.00224, 0.00196 and 0.00232 points per 8 hours, and Bybit runs the WRONG way for a thin-book story, widening from 0.00189 to 0.00295 as the USDC share grows. The middle-half whiskers overlap across every bucket.

On Binance the three buckets read 0.00224, 0.00196 and 0.00232 points per 8 hours, which is flat. On Bybit the gap WIDENS as the USDC share grows, from 0.00189 through 0.00252 to 0.00295, the opposite of what a thin-book story predicts. Why it runs that way is not tested here: nothing in this capture measures volatility, so the reading stops at the ordering. So the cross-venue ordering in the second chart survives as a description of three venues and dies as an explanation. The mechanism this post can defend is narrower: two books, two premium indices, two quote-currency references, with the currency gap explaining about an eighth of the variance.

What does the difference cost a position held for 90 days?

Less than the per-settlement numbers suggest, and not in a consistent direction. Summed over the same 270 settlements, a long paid 1.634 per cent of notional on Binance's USDT contract against 1.516 on its USDC contract, 1.145 against 1.111 on Bybit, and 1.292 against 1.479 on Bitget. The USDC leg was cheaper on two venues and dearer on the third, which tells you the gap behaves like noise around a common signal rather than a standing subsidy for one quote currency. Over a quarter the difference ranged from 0.034 to 0.188 points of notional depending on the venue, so it matters to a carry desk and is invisible to a swing trader. It is also small enough that the practical lesson is about MEASUREMENT rather than about a trade: if you are comparing funding across venues, comparing a USDC-quoted contract on one against a USDT-quoted contract on another mixes a venue difference with a quote-currency difference, and the second is the same size as the first.

Our Athenum explainer on funding intervals names two reasons the same perpetual pays differently across exchanges, positioning and the length of the interval. On this evidence there is a third, and it operates without leaving the building. If you are sizing the cost of carrying a position, the Athenum funding rate calculator turns a rate into a holding cost, and the Athenum liquidation price calculator is the companion for the leverage side.

How can you reproduce this?

Every number above comes from public endpoints that need no key, and the whole capture takes under a minute.

1. Pull the funding history for both contracts on one venue over the same window: BTCUSDT and BTCUSDC on Binance, BTCUSDT and BTCPERP on Bybit and on Bitget. Keep the settlement timestamp with each rate.

2. Match on the timestamp, not on position in the list. Rounding to the nearest hour is worth doing, because a venue can return a stamp a millisecond off the hour: 120 of Binance's 270 settlement stamps in this window are off the hour, against 0 of 270 on Bybit and on Bitget, so an exact key match silently drops the larger half of the venue you have most data for.

3. Count three things before computing any average: how many settlements are identical, how many differ, and how many carry opposite signs. The third number is the one that changes how a reader should think.

4. Report the middle half alongside the median. A median gap of 0.00197 points across a middle half of 0.00088 to 0.00354 is a different claim from the same median on a tight distribution.

5. Test the thin-book explanation inside one venue before repeating it. Sort days by the small contract's share of turnover and compare like with like, rather than reading an ordering off three venues.

What this does not show

Seven limits, stated rather than buried. The window is one quarter, from 2026-06-23 to 2026-09-21, and funding is a regime variable. The asset is Bitcoin only, and a less liquid asset's USDC book is thinner again. Three venues are in the table and they are not the only ones that could be: KuCoin lists XBTUSDTM and XBTUSDCM side by side on the same 8 hour clock, read on 2026-09-21, and it is simply not in this sample. OKX is absent for a different reason, namely that it publishes no BTC-USDC-SWAP, so on OKX the pair does not exist to compare. The index comparison is Binance alone because that is the venue we pulled both index series for; Bybit publishes the same series for BTCPERP and BTCUSDT, and this post does not use it. The correlation of -0.34 describes a linear association across 270 settlements and carries no causal claim. The within-venue turnover test covers Binance and Bybit only, so Bitget supplies one of the three cross-venue points and sits outside the test that kills the thin-book reading. And a contract's quote currency travels with its settlement currency and its collateral rules, so nothing here separates the effect of the price unit from the effect of what you post as margin.

What survives is narrow and checkable. On three venues, over one quarter, a Bitcoin perpetual's funding rate depended on the quote currency at all but 18 to 21 settlements out of 270, the direction of the payment differed at 12 to 45 of them, that gap is between a half and essentially all of the gap between two different exchanges measured the same way, and roughly an eighth of it tracks the distance between the two stablecoins.

You can re-check every figure above on Athenum, which streams live derivatives data from Binance, Bybit, OKX, Bitget, Hyperliquid and Deribit into one normalized view and keeps its 34 calculators free: no account, no email, no usage limits. To see the cross-venue funding grid this post reads from, start a free 7 day Pro+ trial.