If your Bitcoin position is underwater on any of your earlier buys, you may be sitting on legal tax savings the IRS will let you collect — without giving up your Bitcoin position. Here is how the math actually works in 2026.
Tax-loss harvesting is one of the few legal moves that turns short-term price pain into long-term real-money tax savings. With stocks, the IRS’s “wash sale” rule throws sand in the gears — you can’t sell at a loss and immediately buy back the same security. With Bitcoin, that rule (as of 2026) still doesn’t apply. The combination is unusually good for taxpayers.
This article walks through what tax-loss harvesting actually is, why Bitcoin is treated differently than stocks, the specific 4-step procedure, and the mistakes that turn this from a clean tax win into a tax-record disaster. Standard disclaimer: we are not your CPA. The principles below are general; for amounts that would meaningfully change your tax return, talk to a qualified tax professional before you trade.
You bought 1 BTC at $95,000. Today it’s worth $80,000. You have an unrealized loss of $15,000. The loss isn’t real for tax purposes — you haven’t sold — until you trigger it.
If you sell, you realize the $15,000 loss. That loss can be used to offset capital gains elsewhere on your tax return:
Tax-loss harvesting is the deliberate triggering of losses you would have ridden out anyway, in order to capture the tax benefit. For someone in a 24% tax bracket, a $15,000 capital loss harvested correctly is worth roughly $3,600 in tax reduction (assuming it offsets short-term gains or, more likely, meaningful long-term gains).
For stocks, mutual funds, and most securities, the IRS has a “wash sale” rule: if you sell at a loss and buy “substantially identical” shares within 30 days before or after the sale, the loss is disallowed for that tax year. The intent is to prevent the “tax loss without economic loss” trick.
The wash-sale rule is in IRC Section 1091, which applies specifically to stock or securities. The IRS classifies Bitcoin as property, not a security. As of the 2026 tax year, no statute or regulation extends Section 1091 to digital assets — meaning you can sell Bitcoin at a loss and buy back the same Bitcoin five minutes later, and the loss is fully deductible.
Congress has periodically considered closing this gap. The Build Back Better bill of 2021 contained a provision that would have applied wash-sale to digital assets; it didn’t pass. Several Senate proposals since have included similar language; none have become law as of May 2026. So for now, the loophole is alive.
You don’t need to sell everything — you only want to sell the specific buys (“lots”) that are currently at a loss. If you have been dollar-cost-averaging into Bitcoin for two years, your account history probably contains lots bought at $90k, $100k, $110k, $80k, etc. You only want to harvest the ones currently underwater.
This requires specific identification — choosing which Bitcoin you’re selling at the time of sale. By default, exchanges report under FIFO (first-in, first-out), which is rarely optimal. To use specific identification:
Place a sell order for the loss lots. The exchange records the sale at the current market price. The difference between your cost basis (what you paid for that lot) and the sale price is your realized loss. Document everything: timestamp, lot ID, cost basis, sale price, fees.
If you self-custody, you may need to first move the Bitcoin onto an exchange to sell. Moving between your own wallets is not a taxable event — only the eventual sale is.
This is the move the wash-sale rule would block for stocks but doesn’t for Bitcoin. As soon as your sell order fills, you place a buy order for the same dollar amount of Bitcoin. Net effect: you still own roughly the same amount of Bitcoin (slightly less, due to fees and any spread), but your cost basis has been reset to the current price, and you have a realized capital loss to use on your taxes.
One small detail: there will be a tiny gap in price between your sell and buy — spread, slippage, fees. Realistically, that’s 0.05–0.5% of the amount transacted on a major US exchange. So harvesting a $15,000 paper loss might cost you $20–$80 in transaction friction.
The IRS wants to see:
Tax software like CoinTracker, Koinly, or TaxBit imports your exchange and on-chain history and generates Form 8949 directly. For anything beyond a few transactions a year, this is well worth $50–$200 a year in software cost.
The benefit of tax-loss harvesting depends on:
A simple rule of thumb: harvesting becomes clearly worthwhile once your unrealized losses exceed about $1,000–$2,000 in a year. Below that, the documentation cost and the small price-spread friction approach the savings. Above $5,000 of losses, harvesting is almost always worth the 30 minutes of work.
The IRS allows you to use FIFO, LIFO, HIFO (highest-in-first-out), or specific identification — but you must be consistent within a single tax year, and you must document your choice. Switching between methods to optimize different sales is not allowed and will trigger problems if you’re audited. Pick a method (specific-ID is usually best) and stick with it for the year.
If you don’t use specific identification, the exchange will sell your oldest lots first — which for many people are the lots with the lowest cost basis (the biggest gains). That can convert a loss-harvesting move into a gain-recognition disaster. Always confirm specifically which lots are being sold.
If you have to move Bitcoin between wallets or exchanges to do the harvest, make sure the moves themselves don’t accidentally trigger taxable events — a wallet-to-wallet transfer is not taxable, but an exchange-to-exchange transfer that goes through an automatic conversion (some exchanges do this) is.
The new lot you bought to replace the harvested one has a new cost basis (today’s price). Your tax software needs to know this. Manual records, or tax software with auto-import, must reflect the new lot accurately, or you’ll under-report future gains and create a bigger problem later.
If you have $300 of paper losses, harvesting them might save you $50 in taxes and cost you $4 in fees and 30 minutes of work. Worth it for some, not worth it for others. Below about $1,000 of losses we generally suggest skipping.
One sophisticated pattern: use a down period to harvest losses while also rebalancing your overall allocation. If your target is “5% of net worth in Bitcoin” and the current allocation is 7% (because Bitcoin dropped less than your other holdings, say), you can sell down to 5% — harvesting losses on the lots that are underwater — and use the proceeds to buy whatever asset is now underweight. The IRS doesn’t care that the cash didn’t come back as Bitcoin.
For more on how Bitcoin fits a broader portfolio, see our Bitcoin vs stocks portfolio allocation piece and how much Bitcoin should I own.
Yes, but the wash-sale rule applies to spot Bitcoin ETFs the same way it applies to any other ETF. If you sell IBIT or FBTC at a loss and buy back within 30 days, the loss is disallowed. There are two paths to keep the wash-sale clock from tripping:
An interesting twist: holding spot ETF in a brokerage and real Bitcoin in self-custody opens up a possible legal arbitrage — sell the ETF at a loss, replace with real Bitcoin (which the IRS does not consider a security at all). The loss should be valid because the assets are different in tax classification. Ask your CPA before relying on this; the IRS hasn’t formally blessed it.
For a deeper comparison, see our ETF vs real Bitcoin guide.
This is one of the cleanest tax moves available to Bitcoin holders in 2026. The math compounds: a $5,000 harvested loss in a 24% bracket is $1,200 of real, in-pocket money. Do this consistently across down periods and over a full Bitcoin cycle and the savings can fund a meaningful chunk of your future Bitcoin purchases. Just don’t mess up the records — the audit cost of a sloppy harvest can wipe out the savings five times over.
For broader context on how the IRS actually views Bitcoin, our 2026 Bitcoin taxes guide walks through the full tax framework.