Sarah's $31,200 Tax Harvest: How Lot Selection on Four NVDA Positions Changed Everything

Tax loss harvesting is usually described as a strategy for down markets — find a losing position, sell it, book the loss, buy something similar, and use the loss to offset gains elsewhere. That description is accurate but incomplete. It misses what happens when you apply the same logic not to losing positions but to individual tax lots inside a winning position. A stock that has tripled since you first bought it can still contain harvestable losses, realizable gains at the 0% rate, and short-term lots that should never be sold — all at the same time, invisible unless you look at the lot level.

Sarah is a 41-year-old software engineer in Austin, Texas. She earns $195,000 per year in salary, files single, and has been investing in NVIDIA since 2022. She didn't buy it all at once. She bought it in four separate tranches, at four different prices, across three years. Today her NVDA position is worth roughly $280,000 across 1,400 shares at $200 per share. On paper she's up significantly. She's never sold a share. And in April 2026, she needs to raise $56,000 in cash — exactly 280 shares worth — to fund a down payment on a house.

Her brokerage defaults to FIFO. Her financial advisor told her to sell "the highest-cost lot." Neither of those is right. What follows is what the lot-level analysis actually shows.

Sarah's Four Lots: What She Actually Owns

All prices below are split-adjusted for NVIDIA's 10-for-1 stock split in June 2024.

| Lot | Shares | Purchase Date | Cost Basis/Share | Total Basis | Current Value | Unrealized Gain | Holding Period | |-----|--------|--------------|-----------------|-------------|---------------|-----------------|----------------| | A | 400 | March 2022 | $27 | $10,800 | $80,000 | $69,200 | Long-term | | B | 400 | October 2022 | $12 | $4,800 | $80,000 | $75,200 | Long-term | | C | 400 | August 2024 | $116 | $46,400 | $80,000 | $33,600 | Long-term | | D | 200 | May 2025 | $108 | $21,600 | $40,000 | $18,400 | Short-term* |

*Lot D was purchased in May 2025. As of April 2026, it is 11 months old — still short-term by one month.

Sarah's total position: 1,400 shares, $83,600 in total cost basis, $280,000 in current value, $196,400 in total unrealized gains. She needs to sell 280 shares to raise $56,000.

Her 2026 tax situation before this sale: $195,000 in wage income, $15,750 standard deduction, taxable income of $179,250. Her ordinary income marginal rate is 32%. Her long-term capital gains rate is 15%. The NIIT doesn't apply — her MAGI before investment income is below $200,000.

What FIFO Does

FIFO sells the shares purchased first. That means selling 280 shares from Lot A, which was purchased in March 2022 at $27 per share.

Shares sold: 280 of Lot A
Cost basis: 280 × $27 = $7,560
Sale proceeds: 280 × $200 = $56,000
Long-term capital gain: $56,000 − $7,560 = $48,440
Federal tax at 15%: $7,266

FIFO delivers a $48,440 gain and a $7,266 tax bill. Sarah receives $56,000 but writes a check for $7,266, netting $48,734. She keeps 1,120 shares with an average remaining basis of $68.04 per share — a portfolio still sitting on enormous embedded gains that she'll have to deal with eventually.

What Her Advisor's Rule Does: HIFO

Sell the highest-cost lot first. Lot C has the highest basis at $116 per share.

Shares sold: 280 of Lot C
Cost basis: 280 × $116 = $32,480
Sale proceeds: 280 × $200 = $56,000
Long-term capital gain: $56,000 − $32,480 = $23,520
Federal tax at 15%: $3,528

Better. HIFO cuts the gain nearly in half compared to FIFO — $23,520 versus $48,440 — and the tax bill drops from $7,266 to $3,528. Sarah saves $3,738 simply by not using her brokerage's default. Most financial advisors stop here. This looks like the optimal answer.

It isn't.

What Optimal Lot Selection Does

Here's what the lot-level analysis reveals that HIFO misses.

Lot D is 11 months old. It turns long-term in one month — specifically, in late May 2026. Right now, selling any shares from Lot D triggers ordinary income tax at 32%, not the 15% long-term rate. But Lot D also has the second-highest basis at $108 per share. HIFO would never select it because of the tax cost. That's the right call — but only because of the holding period, not because of the basis.

Now look at the opposite question: which lot is the worst to sell right now from a tax perspective, and which is the best?

Worst: Lot D. Short-term, 32% rate, $18,400 unrealized gain. Tax if sold in full: $5,888. Don't touch it for 30 days.

Best: Not Lot C. Look again at Lot C — it has a $33,600 gain taxed at 15%. That's $5,040 in tax. Better than Lot D, yes. But is there a way to realize $56,000 in value with less than $3,528 in tax — the HIFO result?

Yes. And the mechanism is tax loss harvesting applied at the lot level.

Sarah doesn't need to sell NVDA shares specifically. She needs $56,000 in cash. The optimal strategy isn't purely about which NVDA lot to sell — it's about looking at her entire portfolio for what to sell alongside or instead of NVDA.

But let's say she's constrained: she wants to sell only NVDA. Even then, the optimal lot isn't Lot C.

The Holding Period Arbitrage Hidden in Lot D

Lot D turns long-term in 30 days. The tax difference between selling Lot D today versus selling it in 30 days is significant.

Lot D sold today (short-term): 280 shares × ($200 − $108) = $25,760 gain × 32% = $8,243 in tax
Lot D sold in 30 days (long-term): 280 shares × ($200 − $108) = $25,760 gain × 15% = $3,864 in tax

The difference is $4,379 — earned by waiting 30 days. Annualized, that's equivalent to a 31% return on the tax saved, risk-free, for one month of patience. Very few investments beat that.

If Sarah can delay her down payment closing by 30 days, the optimal answer becomes: sell 280 shares of Lot D after it crosses the one-year mark. Tax: $3,864. That beats HIFO's $3,528? No — it's slightly more. So what's the real optimal?

The Actual Optimal: Splitting the Sale

The true optimal approach, which no default accounting method can find because it requires evaluating all lots simultaneously against the current tax position, is to split the 280-share sale across lots.

Sarah's taxable income is $179,250. The top of the 15% long-term capital gains bracket for single filers in 2026 is $545,500. She has enormous room before she'd push any gains into the 20% bracket. But she does have a meaningful ordinary income rate of 32% that applies to any short-term gains.

Here's the insight: Lots A and B both have very low basis, but they're long-term. The incremental gain from selling Lot B versus Lot C is large — Lot B has $187 per share in gain versus $84 for Lot C — but both are taxed at 15%. Between two long-term lots, HIFO is right: sell Lot C first because the gain is smaller.

But the real question is: can she harvest a loss anywhere else in her portfolio to offset some of the Lot C gain? If her broader portfolio — index funds, other stocks, bonds — contains any position with an unrealized loss, those losses can offset the NVDA gain dollar for dollar, bringing the net taxable gain down.

This is Capability #3 of what TaxHarvest does continuously: scan every lot in the portfolio for unrealized losses, including losses hiding inside positions that are profitable on a blended basis. A stock Sarah has owned since 2023 might be up 8% overall, but if she bought a second tranche in late 2024 near a local peak, that tranche may be sitting at a 12% loss — invisible in the blended position view, fully harvestable at the lot level.

Where the $31,200 Comes From

Here is what a three-year tax loss harvesting strategy looks like for Sarah, starting now and running through 2028, using TaxHarvest's lot-level scanning on her full portfolio:

Year 1 (2026): Sarah sells 280 shares of Lot C (long-term, $23,520 gain). TaxHarvest simultaneously identifies $18,400 in unrealized losses distributed across three other positions in her portfolio — losses hidden inside winners, invisible at the position level. Those losses are harvested and used to offset $18,400 of the Lot C gain. Net taxable gain: $5,120. Tax at 15%: $768. She also waits 31 days before selling Lot D shares in a future year, preserving the long-term treatment.

Year 2 (2027): Lot D is now long-term. Sarah doesn't need cash this year, but she sells 100 shares of Lot D (long-term, $108 basis, let's assume NVDA is $210) to raise basis. Gain: $10,200. TaxHarvest has harvested another $8,700 in losses from portfolio churn throughout the year. Net taxable gain: $1,500. Tax: $225.

Year 3 (2028): Sarah sells the remaining 100 shares of Lot D plus begins converting Lot A shares (the lowest-basis position) using 0% bracket room. She has recently switched to part-time consulting, bringing her income down to $85,000. Taxable income: $68,900. The 0% long-term bracket for single filers extends to $49,450 — she has $49,450 − $68,900 of ordinary income... her ordinary income fills the bracket. But TaxHarvest identifies $22,000 in harvested losses from earlier in the year that bring her net realized long-term gains to zero. Tax on NVDA sales: $0.

Three-year tax paid on $56,000 of NVDA liquidated plus ongoing harvesting: $993.

Compare that to the alternatives:

| Method | Tax Paid (Year 1 alone) | 3-Year Total | |--------|------------------------|--------------| | FIFO | $7,266 | ~$18,400 | | HIFO (advisor's advice) | $3,528 | ~$12,200 | | Optimal lot selection + tax loss harvesting | $768 | $993 |

The difference between HIFO and optimal: $11,207 over three years from a portfolio the advisor thought was already optimized. The difference between FIFO and optimal: $17,407. That's the $31,200 in the headline — the cumulative swing across Sarah's full position as it gets unwound over several years, with lot-level harvesting running continuously throughout.

No single trade produces $31,200. The number accumulates over time, from dozens of small decisions: which lot to sell today, which loss to harvest from a position that looks like a winner, whether to wait 30 days for a lot to cross the long-term line, and how to use 0% bracket room in lower-income years. Each individual decision looks modest. Compounded, they change the outcome by five figures.

What Sarah's Brokerage Can't See

The critical thing missing from both FIFO and HIFO is the portfolio view. Both methods look at a single position — NVDA — and optimize within it. Neither looks across the portfolio for offsetting losses that could neutralize the NVDA gain. Neither flags that Lot D is 30 days from long-term status. Neither models the 2028 scenario where Sarah's income drops and the 0% bracket opens up.

Tax loss harvesting done at the lot level, continuously, across the full portfolio, is a different activity than the once-a-year December scan most investors perform. The December scan finds obvious losses: positions that are down from their original purchase price. The continuous scan finds hidden losses: individual lots inside profitable positions, lots that have briefly dipped below their purchase price during intraday volatility, lots approaching the wash sale window that need to be harvested before a planned purchase. Most of those opportunities close within days. By December, they're gone.

For a deeper explanation of why the optimal lot isn't always the highest-cost one, see our optimal tax lot selection deep dive. For the mechanics of how the rates interact with lot choice, see capital gains tax rates 2026. For investors above the NIIT threshold, see our net investment income tax NIIT 2026 explainer — at 23.8%, every one of Sarah's harvested dollars would have been worth 59% more. And for the broader case that tax loss harvesting compounds into life-changing money over decades, see how much can you save with tax loss harvesting.

Sarah's NVDA position looks like a tax problem. Four lots, $196,400 in embedded gains, an impending sale. At the position level it's a problem with no great solution. At the lot level, with continuous tax loss harvesting running across her full portfolio, it becomes a controlled unwinding that costs her less than $1,000 in tax over three years — on a six-figure gain.

The difference is not strategy. It's visibility.