Nvidia without Jensen Huang? What Bun teaches about bus factor
The Wall Street Journal published a story with a headline I read twice: Nvidia is reportedly preparing to remove Jensen Huang as CEO. He is the founder, in charge since 1993, of the company that became the hardware foundation for almost all the AI you use to write code.
Before going further, an honest warning. The WSJ story is behind a paywall and I couldn't confirm the details with another source. I don't know the reason, I don't know the timeline, I don't know who on the board would be pushing this. So treat it as one newspaper's report, not as a done deal. I'm not going to make up backstage drama.
What I can do is talk about what this headline exposes, and that holds even if the story changes tomorrow: how much we depend on a single person. At Nvidia, in open source and, yes, in Bun.
What we know about Nvidia and Jensen Huang
Not much, and it's better to admit it.
What is fact regardless of the headline:
- Jensen Huang co-founded Nvidia in 1993 and has never left the CEO role.
- The company passed 5 trillion dollars in market value in 2025.
- CUDA, launched in 2006, is the reason almost every AI framework runs on Nvidia cards first.
What is a report: that there is an internal move to change the leadership. Until I read an official confirmation, it stays in that drawer.
If you're thinking about changing a technical decision because of a newspaper headline, take a breath. Don't change it.
Why this matters to people who write code
In the short term, for your code, nothing changes. Your bun install stays the same. Your deploy stays the same.
But the headline stings for a real reason. A big share of the coding assistants we use runs on Nvidia GPUs. The price of inference, the queue for hardware and the pace of the models all go through one company's decisions. And that company has always been read as an extension of one person.
Engineering has a name for this: bus factor. How many people need to disappear for the project to stop. When the answer is "one", you have a problem, even if that one person is brilliant.
Open source has already shown both sides of this:
- Python, 2018. Guido van Rossum stepped down as "benevolent dictator". The language set up an elected council of five people and moved on. Planned succession works.
- xz, 2024. A maintainer, alone and burned out, accepted help from a stranger. The help was a backdoor, CVE-2024-3094. It was only caught because an engineer found an extra half second on an SSH login strange.
The difference between the two cases is governance. Both had plenty of talent.
What Bun has to do with this
Everything. Bun is one of the most "one person" projects I use every day.
It came out of Jarred Sumner's head. It reached version 1.0 in September 2023. In December 2025 it was acquired by Anthropic, which uses the runtime in Claude Code. It is still open source, under the MIT license.
The acquisition solved one risk and created another. Startup money runs out on a deadline, and that stopped being a concern. In exchange, the roadmap now answers to the priorities of a single company. Compare that with Node, which lives under a foundation, with a technical committee and several companies paying people to maintain it.
That's no reason to drop Bun. I didn't. It's a reason to know where you're standing.
And Node has been through this. In 2014, part of the maintainers got tired of Joyent's management and created the io.js fork. A year later the two merged again, this time with a foundation. The code survived because it was open and because there were enough people to carry it.
Depending on a single person is technical debt that nobody puts in the backlog.
How to measure the bus factor of your stack
You can do it in five minutes, with no new tool. Clone the repository of a critical dependency and run:
git shortlog -sn --no-merges --since="1 year ago" HEAD | head -5
You'll see who actually committed in the last year. If the first name has ten times more commits than the second, you've found your single point of failure.
Then, three quick questions:
- Who can publish a new version? One person or a team?
- Who pays the bill? A foundation, one company, or nobody?
- Does the license let someone fork it if things go wrong?
You don't need to do this for the 900 dependencies in your node_modules. Do it for the five that would take your product down: runtime, framework, ORM, authentication and whatever touches money.
How to make a Bun project less of a hostage
This is where the headline turns into real work. And the work is small.
First, find out how tied in you already are:
grep -rnE "Bun\.|from ['\"]bun" src
Every line that shows up is a place that only runs on Bun. The goal isn't to get to zero. It's to know the number.
Second, wherever a standard API does the same job, use the standard one. Bun implements the node: modules, so you lose nothing:
// only runs on Bun
const config = await Bun.file('config.json').text()
// runs on Bun and on Node
import { readFile } from 'node:fs/promises'
const config = await readFile('config.json', 'utf8')
Third, keep the server out of the runtime. With Hono, the application is the same on both sides:
// app.ts
import { Hono } from 'hono'
const app = new Hono()
app.get('/health', (c) => c.json({ ok: true }))
export default app
On Bun, that export default already starts the server. On Node, one file changes:
// node.ts
import { serve } from '@hono/node-server'
import app from './app'
serve({ fetch: app.fetch, port: 3000 })
Fourth, test the emergency exit. A weekly CI job that boots the app on Node and hits /health is enough. An escape plan that was never tested is not a plan.
Now the obvious objection: "so should I wrap everything in an abstraction layer?". No. That's trading a hypothetical risk for real complexity, every day. Use Bun.password, bun:sqlite and whatever else makes sense. Just keep those calls in a few files, so that migrating is an afternoon of work and not a quarter.
My take
I don't know if Jensen Huang leaves, stays, or if the story gets denied next week. And I'm suspicious of anyone who says they know.
What I do know is that the market's reaction to a headline like this shows the size of the bet on one person. If even the man in the leather jacket can become a succession topic, so can the maintainer of your favorite lib.
My strong opinion: we pick tools by looking at benchmarks and ignore governance, and that's a mistake. A benchmark only shows how the tool is doing today. What becomes of it in three years depends on who runs the project.
I'm sticking with Bun. It's fast, the DX is great and the license protects me. I just don't let it be the only exit door. If you want to see how this shows up in what I build, take a look at my projects.
LinkedIn summary
I read the headline twice: Nvidia is reportedly preparing to replace Jensen Huang, CEO since 1993. I couldn't confirm the details, so I'm treating it as a newspaper report, not as fact. But it exposes something that holds even if it gets denied tomorrow: how much we depend on a single person. Engineering has a name for this: bus factor. Python survived Guido stepping down because it planned the succession. xz almost became a global backdoor because it had one maintainer, alone and burned out. I use Bun every day and I'm sticking with it. I just don't let it be the only exit door: standard APIs where possible, Bun-specific calls in a few files, and a CI job that boots the app on Node. We pick tools by looking at benchmarks and ignore governance. A benchmark shows how the tool is doing today. Whoever runs the project decides how it will be doing in three years. In the article I show how to measure the bus factor of your stack in five minutes, with one git command. And you, have you checked who maintains the five dependencies that would take your product down? #BusFactor #OpenSource #Bun #SoftwareEngineering #Nvidia