Tech Employee RSU Withholding Gap & ISO AMT Tax Simulator

RSU vs ISO Taxes: The Engineer's Guide to Not Getting Destroyed by AMT

Your startup offer includes 10,000 ISOs at a $2 strike price. Four years later the company IPOs at $50. You exercise, hold for the favorable tax treatment, and feel brilliant — until April, when your accountant says you owe $140,000 in Alternative Minimum Tax on money you have not actually received. This is the ISO AMT trap, and it has bankrupted careful, intelligent people. It happens because ISOs are taxed twice in two parallel universes: regular tax (nothing at exercise) and AMT (tax on the "bargain element" immediately). If you hold the shares past December 31st, you owe AMT on phantom income.

Let us make it concrete. You exercise 10,000 ISOs at $2 when the fair market value is $50. Bargain element: $48 × 10,000 = $480,000. For AMT purposes, that $480,000 is income this year — at roughly 26–28% AMT rates, that is ~$125,000–$135,000 in tax. But you have not sold anything. If the stock then drops to $20 (it happens — lock-up expirations are brutal), you hold shares worth $200,000 and owe $130,000 in tax on them. People have literally gone bankrupt this way: the dot-com crash created a generation of engineers who owed AMT on millions in stock that became worthless. The IRS does not accept "but the stock crashed" as payment.

RSUs are simpler but sneakier. When RSUs vest, the full value is ordinary income that day — $50,000 vesting at 37% federal plus state means ~$20,000+ in tax, usually handled by "sell to cover" (your employer sells some shares automatically). The gotcha nobody mentions: if you do not sell the rest immediately and the stock drops, you paid tax on value you no longer have, and there is no refund. The standard advice from every equity-comp advisor: sell RSUs at vest and diversify. Holding vested RSUs is not "believing in the company" — it is doubling down your human capital and financial capital on the same bet.

The strategies that actually work, in order of importance. First: model AMT before exercising ISOs — this calculator exists for exactly that; never exercise blind. Second: consider exercising ISOs early (when the 409A valuation is low) to minimize the bargain element — the classic startup move is exercising at grant when FMV ≈ strike, creating near-zero AMT exposure. Third: for a qualifying disposition (2 years from grant + 1 year from exercise), sell just enough in the exercise year to cover the AMT bill — the "sell to cover AMT" approach sacrifices some long-term capital gains treatment to avoid the bankruptcy scenario. Fourth: AMT paid generates AMT credits usable in future years, so the money is not entirely lost — but credits only help if you have future tax liability to offset, cold comfort when you are cash-poor.

A few more traps from the field. The 90-day post-termination exercise window: leave your startup and you typically have 90 days to exercise vested options or lose them — forcing a huge AMT decision under time pressure. Negotiate for extended windows (2–10 years) before joining; it is the single most valuable contract term nobody negotiates. NSOs (non-qualified options): taxed as ordinary income at exercise on the spread, no AMT games, simpler but less favorable — many late-stage startups grant these. And Section 83(b): for early-exercise or restricted stock, filing within 30 days of grant (not a day later — the IRS is merciless) lets you pay tax on the current low value instead of the vested high value. Miss the 30-day window and the election is gone forever.

When to get professional help: if your bargain element exceeds ~$100,000, if you are exercising across tax years, or if you are leaving a company with significant unexercised options — pay a CPA who specializes in equity comp ($500–$1,500 for a planning session). It is the highest-ROI professional fee in personal finance. This calculator gives you the numbers to bring to that meeting: bargain element, estimated AMT, qualifying vs disqualifying outcomes, and the sell-to-cover math. Walk in with data, not anxiety.