Currently free to try
Skip to content
NINEHOUNDTHE EASTERN OBSERVATORY

Famous lives / NINEHOUND JOURNAL

Jensen Huang: the difficult request that kept NVIDIA alive

Finishing the contract could leave the company stranded. Abandoning it could remove the cash it needed. Huang had to approach the customer with both problems.

NineHound · 2026-11-124 min read中文原文
01

When a contract stops being good news

A supplier normally visits a customer with an answer. Jensen Huang once had to bring a much less comfortable message: our direction is wrong, you should find another partner, and we still need your help. That is a difficult combination to put across when the customer has every reason to expect delivery.

In his 2023 National Taiwan University commencement address, Huang recalled an early Sega console contract that NVIDIA could no longer sensibly complete along its chosen technical path. Continuing threatened the company’s future; stopping threatened the income needed to keep going. What had been a commercial opportunity had become a trap.

The interest of the episode lies in that collision of obligations. A founder can revise a presentation in minutes. A customer’s expectations, money already spent and the work still required do not disappear with the old slide. Changing direction has people attached to it.

02

A request with no guaranteed answer

According to Huang’s account, he explained the mistake, recommended that Sega find another partner and asked for financial help. Sega agreed. The breathing room helped NVIDIA bring RIVA 128 to market and survive. This is his retrospective account, rather than a transcript of the private negotiation.

Two admissions sit together in the request: I have not done what you needed, and I still need something from you. Stating only the second would leave the customer guessing about the first. Stating both exposes the supplier to refusal, but gives the other party a clearer basis on which to make a decision.

There is no need to invent a dramatic pause or a speech across a boardroom table. The predicament already supplies the tension. A customer is not obliged to fund a supplier’s mistake, and preserving the supplier’s dignity does not settle its bills. The offer of help was Sega’s decision too. Later success should not erase that other participant from the story.

03

The same person later stayed the course

Huang’s account of CUDA complicates any simple lesson about quitting early. NVIDIA committed to the computing platform in 2007 and continued through years of expense and criticism. Its own account presents that investment as a long commitment made under pressure.

One course was abandoned; another was maintained. Calling both decisions visionary merely gives them a flattering common name. A more useful distinction asks whether a project’s underlying conditions have broken down, or whether a plausible direction still needs time and work to develop.

Both situations can initially look like expenditure, scepticism and disappointing returns. Discomfort alone cannot distinguish them. Evidence about users, the problem being solved and the next achievable test can help. That is our interpretation of the contrast, not a claim to know every consideration inside NVIDIA at the time.

04

An Eastern reading of changing course

The Chinese idea of biantong, adapting through change, can sound effortless when presented as wisdom. Here it begins with embarrassment: the person who approved a direction must acknowledge its failure. The greater that person’s authority, the more uncomfortable the admission may be.

The goal and the chosen method become separable. A company may still pursue useful technology while leaving a particular contract or approach behind. When a leader treats a method as an inseparable part of personal identity, revising it starts to feel like denying the entire person. That feeling can keep an exhausted project alive long after its justification has weakened.

CUDA supplies the necessary complication. Outside criticism does not automatically establish that a direction should be abandoned. Adaptation requires tolerating an admission of error; sustained work may require tolerating other people’s dissatisfaction. The hard question is what one is protecting: a credible possibility or an image of always being right.

05

What the admission did, and did not, accomplish

Humility did not manufacture a product, supply a team or guarantee a market. It reopened possibilities that still required other people and further work. That is a more demanding lesson than a claim that a good attitude ensures success.

For someone facing a troubled project, the practical starting point is to describe the trouble before defending the self: what has changed for the customer, where the approach fails and what resources remain. A request for help then has substance. A decision to continue can also have a reason beyond the refusal to lose.

Sources and further reading

The Sega episode follows Huang’s own 2023 commencement recollection. Cultural interpretation is editorial, not a birth-chart explanation of business results.

Try it for yourself.

Explore questions and change in the I Ching ↗

Keep exploring

All articles ↗