Kumo Research · Stock perps
What does it cost to hold a stock perp? A year of funding rates on Tesla, Nvidia, Apple and other on-chain stocks
Perpetual futures have no expiry date, so they use a funding rate: a small payment between longs and shorts that keeps the perp price near the real price. For anyone holding a stock perp for weeks or months, funding is the main cost. We analysed a year of hourly funding on Hyperliquid's stock, commodity and currency perps.
By Alex Wu · Published · 6 min read
Key findings
- Holding a long stock perp cost between 1.8% (Apple) and 9.6% (Alphabet) a year on average from December 2025 to November 2026; the average across eight large stocks was 6.8% a year.
- Funding is mostly a weekday cost. Across the eight stocks, longs paid an average of 8.8% a year (annualised) on weekdays but only 1.7% on weekends.
- About half the time, funding simply sat at the built-in base rate of 0.000625% per hour – roughly 5.5% a year. Longs paid in 74–91% of all hours.
- Not every market charges longs. On the WTI oil perp, shorts paid longs an average of 16% a year; on the yen market, shorts paid about 5.6% a year.
- Stock perps were often more expensive to hold than Bitcoin and Ether perps in the first half of 2026, and cheaper in the second half, when crypto funding rose to 8–11% a year.
Why funding matters
When you buy a share, you pay once and can hold it forever at no extra cost. A perpetual future works differently. Because it never expires, the exchange needs another way to keep its price close to the real asset. That mechanism is the funding rate: a payment between the two sides of the market, settled every hour on Hyperliquid. When the perp trades above its reference price – usually because more traders want to be long – longs pay shorts. When it trades below, shorts pay longs.
For a quick trade, funding hardly matters. For anyone holding a stock perp for weeks or months as a substitute for owning the stock, it is the main ongoing cost, and it is rarely shown clearly. We collected every hourly funding payment for Hyperliquid's trade.xyz markets since December 2025 – eight large US stocks, the Nasdaq-100, gold, silver, WTI oil, the euro and the yen – plus Bitcoin and Ether for comparison.
All figures below are annualised: an average hourly rate multiplied by the 8,760 hours in a year, so they can be compared with interest rates.
The cost of holding a long stock perp
What longs paid to hold each perp
Average funding rate on Hyperliquid, annualised, December 2025 – November 2026 (negative = shorts paid longs)
- Alphabet9.6%
- Nvidia8.8%
- Amazon8.7%
- Meta7.2%
- Microsoft6.9%
- Tesla6.0%
- Coinbase5.1%
- Apple1.8%
- Nasdaq-100 (XYZ100)2.7%
- Silver10.0%
- Gold8.4%
- Euro (EUR)3.0%
- Japanese yen (JPY)-5.6%
- WTI oil-16.1%
- Ether6.2%
- Bitcoin5.5%
Source: Hyperliquid funding history (hourly), trade.xyz markets and main perps · Chart: Kumo Research, kumocast.io
From December 2025 to November 2026, longs paid on average:
- Alphabet 9.6%, Nvidia 8.8% and Amazon 8.7% a year – the most expensive.
- Meta 7.2%, Microsoft 6.9%, Tesla 6.0% and Coinbase 5.1%.
- Apple 1.8% – by far the cheapest.
The average across the eight stocks was 6.8% a year. In other words, holding $10,000 of exposure through a long stock perp cost roughly $680 over the year in funding, before trading fees. A perp holder also receives no dividends; on the other hand, no capital is tied up beyond the margin.
The Nasdaq-100 perp (XYZ100) was cheaper than most single stocks at 2.7% a year, and gold and silver cost 8.4% and 10.0%.
| Market | Average (annualised) | Last 90 days | Hours longs paid | Hours at the base rate | Highest hourly rate | Lowest hourly rate |
|---|---|---|---|---|---|---|
| Alphabet | +9.6% | +7.9% | 91% | 44% | +0.027% | -0.037% |
| Nvidia | +8.8% | +5.5% | 89% | 48% | +0.036% | -0.067% |
| Amazon | +8.7% | +5.7% | 91% | 49% | +0.070% | -0.045% |
| Meta | +7.2% | +7.4% | 89% | 50% | +0.030% | -0.033% |
| Microsoft | +6.9% | +3.0% | 86% | 48% | +0.028% | -0.043% |
| Tesla | +6.0% | +6.4% | 84% | 50% | +0.062% | -0.049% |
| Coinbase | +5.1% | +8.9% | 84% | 44% | +0.083% | -0.102% |
| Apple | +1.8% | +2.1% | 74% | 48% | +0.050% | -0.032% |
| Nasdaq-100 (XYZ100) | +2.7% | +4.7% | 79% | 57% | +0.033% | -0.078% |
| Silver | +10.0% | +7.8% | 91% | 51% | +0.161% | -0.083% |
| Gold | +8.4% | +6.9% | 91% | 55% | +0.050% | -0.041% |
| Euro (EUR) | +3.0% | +4.4% | 54% | 42% | +0.012% | -0.012% |
| Japanese yen (JPY) | -5.6% | -4.8% | 13% | 35% | +0.015% | -0.021% |
| WTI oil | -16.1% | -24.1% | 60% | 28% | +0.094% | -0.100% |
| Ether | +6.2% | +8.7% | 83% | 52% | +0.008% | -0.008% |
| Bitcoin | +5.5% | +8.7% | 80% | 46% | +0.009% | -0.007% |
Funding is paid every hour on Hyperliquid. "Base rate" is the most common hourly value: 0.000625% per hour (about 5.5% a year) for trade.xyz markets, 0.00125% (about 11% a year) for main crypto perps. Gold, silver, oil and FX markets start between late December 2025 and early January 2026.
The base rate does most of the work
The table shows that in roughly half of all hours, funding on the trade.xyz markets was exactly 0.000625% per hour – about 5.5% a year. This is the base rate built into the funding formula, which applies whenever the perp is trading close to its reference price. It works much like an interest charge for borrowing the exposure. Hyperliquid's main crypto perps such as Bitcoin and Ether use a base rate twice as high (about 11% a year).
So most of the cost of a stock perp is not caused by crowded positioning – it is the default rate. The extra on top, or the discount below it, comes from the hours when the perp trades at a premium or a discount. Longs paid funding in 74% (Apple) to 91% (Alphabet, Amazon) of all hours.
Weekdays vs weekends
Funding is mostly a weekday cost
Average annualised funding on weekdays vs weekends, %
- Weekdays
- Weekends
Source: Hyperliquid funding history, December 2025 – November 2026 · Chart: Kumo Research, kumocast.io
Funding is heavily concentrated on weekdays, when the US market is open and traders are most active. Across the eight stocks, longs paid an average of 8.8% a year (annualised) on weekdays but only 1.7% on weekends. On Coinbase, weekend funding was even slightly negative (−2.5%) – shorts paid longs. This matches what we found in our tracking study: weekends are quiet, and the perps rarely trade at a premium then.
When longs get paid: oil and the yen
Funding is not always a cost for longs. On the WTI oil perp, shorts paid longs an average of 16% a year (24% in the last 90 days), even though longs paid in 60% of hours – the hours when shorts paid were much larger. On the yen market, funding was negative in most hours, averaging −5.6% a year. In these markets more traders wanted to be short than long, and the funding mechanism paid the longs for taking the other side. The euro market charged a modest 3.0%.
Compared with Bitcoin and Ether
Month by month: stock perps vs Bitcoin and Ether
Average annualised funding, %
- Average of 8 stock perps
- Bitcoin
- Ether
Source: Hyperliquid funding history · Chart: Kumo Research, kumocast.io
Over the full period, holding a long Bitcoin perp cost 5.5% a year and an Ether perp 6.2% – slightly less than the stock average. The comparison changed during the year. In February to April 2026, when crypto markets were weak, BTC and ETH funding fell to near zero or below, and it stayed under stock-perp funding through June, while stock perps kept charging 3–10%. From July, as crypto funding rose back to 8–11% a year, the average stock perp became the cheaper of the two.
Individual months can be expensive. Nvidia's funding averaged 21.6% (annualised) in February 2026 and Microsoft's 17.2% in June 2026, and the most extreme single hours reached about 0.08% for Coinbase and 0.16% for silver.
OKX for comparison
OKX's USDT-margined stock swaps settle funding every eight hours. Over August to November 2026 their annualised averages were lower – between about 0.5% (Microsoft) and 8.2% (Alphabet), mostly 3–4% – with many funding periods at exactly zero. Different venues use different formulas, so the cost of the same exposure can differ by several percentage points a year.
What this means
- Holding a long stock perp has cost around 2–10% a year in funding, 6.8% on average for eight large US stocks.
- Most of that is the built-in base rate (about 5.5% a year on trade.xyz markets); premiums add to it on busy weekdays.
- Weekends are cheap, and some markets – such as oil and the yen – have paid longs.
- Funding varies by venue and by month, so it is worth checking before holding a position for a long time.
For live levels and volatility ranges on these markets, see our Tesla, Nvidia, gold and oil pages.
Method, data and limits
Data: hourly funding history from Hyperliquid's API for trade.xyz markets (xyz:TSLA, NVDA, AAPL, MSFT, AMZN, META, GOOGL, COIN, XYZ100, GOLD, SILVER, CL, EUR, JPY) and main perps (BTC, ETH), 1 December 2025 – 2 November 2026 (later start for commodities and FX). OKX funding history for USDT stock swaps, 27 July – 2 November 2026.
Measures: annualised funding = mean hourly rate × 8,760 (OKX: mean 8-hour rate × 1,095). Positive = longs pay shorts. "Base rate" = the most common hourly value. Weekends are Saturday and Sunday in New York time.
Limits: funding formulas and base rates can be changed by the venue; costs exclude trading fees, spreads and liquidation risk; returns of the underlying are not included. Not investment advice.
Download the data (CSV, CC BY 4.0)
Sources
How to cite
Kumo Research (2026). "What does it cost to hold a stock perp? A year of funding rates on Tesla, Nvidia, Apple and other on-chain stocks". Kumocast. https://www.kumocast.io/research/stock-perps-funding-rates
This study describes past market data. It is not investment advice and past behaviour does not guarantee future results.
Live data for these markets
More from Kumo Research
- Stock perpsDo on-chain stock perps track the real stock? Hyperliquid and OKX vs Nasdaq prices, hour by hour6 min read →
- Stock perpsDo stock perps predict Monday's open? What 24/7 trading on Tesla, Nvidia and other stocks says about the next opening price5 min read →
- Stock perpsEarnings night on-chain: how Tesla, Nvidia, Microsoft and other stock perps react before Wall Street opens6 min read →