The second option

25,000 options. One year to vest. A meeting on a houseboat in Sausalito. That is how Eric Gullichsen joined Nvidia's technical advisory board in 1993, and on Sunday he posted what happened next: in April 1996 the CFO wrote to say 15,625 of his shares had vested and he was required to exercise them. He did, and forgot about it. In 2024, with Nvidia on every screen at a day-trader friend's desk, he went home and reopened the folder.

The agreement said quarterly installments over one year. The letter had applied four. Ten quarters of a four-year schedule is exactly 62.5%, which is exactly 15,625 of 25,000. The missing 9,375 shares, after a cumulative 480x in splits, are 4.5 million today. Nvidia didn't dispute the document. It said the claim was decades time-barred, and his own lawyers concluded they were unlikely to get past a motion to dismiss. His verdict:

That's the wrong lesson. The clock didn't excuse Nvidia's error. It refused Gullichsen a second option: a free call on thirty years of other people's work.

Look at what the claim was worth when it arose. In 1996 Nvidia was private and bruised. The NV1 had shipped into a DirectX that drew triangles only, not the quads its texture mapping was built on, and the company had laid off a large share of its staff. Another 9,375 options in that company were a lottery ticket. The shortfall became a billion dollars because Jensen Huang and several thousand engineers spent three decades turning a stumbling graphics startup into the company on every trading-floor screen. Had they failed, the folder would have stayed closed. Nobody reconciles the vesting schedule on a stock that went to zero.

The law noticed this asymmetry a century before Nvidia existed. In 1889 the Supreme Court took up Galigher v. Jones, a fight over shares a broker had sold out from under his client, and had to decide which price the client should recover. New York had once allowed the highest price up to trial, which could be years later, and found the hardship so great it narrowed the rule. The Court adopted the narrowed version:

The point of that rule is the point here. A wronged owner recovers what he'd have had by acting once he knew, not the upside of waiting to see which way the market went. Anything more and the defendant is writing him an option for free. A statute of limitations is the blunter tool built on the same instinct. And Gullichsen's notice wasn't hidden. It sat in his own folder, in an agreement he'd signed, contradicted by a letter he'd accepted.

The clock didn't excuse Nvidia's error. It refused Gullichsen a second option: a free call on thirty years of other people's work.

The strongest objection is that Nvidia wrote the wrong number. The CFO and outside counsel at Cooley both put a four-year schedule on a one-year grant, the company held every advantage in information, and an advisor starting internet ventures in Tonga trusted the people who kept the cap table. All true, and "so sue us" at a settlement meeting is not a good look for anyone. The law does have a remedy for the wrongdoer who hides the facts: clocks that start when the victim could have found them. It is little help when the facts were two pieces of paper apart, both in the claimant's possession, for twenty-eight years.

For anyone holding a grant — founders, early hires, advisors paid in paper, most of the people a venture firm spends its week with — the practical rule is short. The company's letter is a claim about your agreement. It is not the agreement. Reconcile them the week the letter arrives, while the gap is worth a lottery ticket and nobody has a reason to fight about it.

Gullichsen held two options in 1996. He exercised the one Nvidia mailed him. The other expired in a folder.