After Jensen, what happens to Nemotron?
I don't particularly want to write about the scandal surrounding Jensen Huang's departure from NVIDIA. There are legitimate questions for the board and investigators, but I don't think the scandal was the technology story of the week.
For me, that was the week we lost email and then stopped airplanes from taking off. Microsoft's widespread service outage began on November 12; the FAA ordered a nationwide ground stop the following day after air traffic control communications failed.78 Those were the stories that reached into everybody else's plans, whether or not they follow the AI industry.
Huang's departure matters here for a different reason. He was a prominent advocate for open-weight models. What happens to Nemotron without him?
The systems we already depend on
The early explanations for the outages are depressingly familiar. CBS News reported that investigators' leading theory for the FAA failure was a software update to the private telecommunications network connecting air traffic control facilities: an update, a reboot, a crash.12 Microsoft's service updates described servers becoming unavailable, pushing work onto the remaining infrastructure. That is a cascading capacity problem, not necessarily a sudden increase in demand.8
Neither account establishes AI as the direct cause. An update is not evidence of AI-written code, and overloaded servers are not evidence that chatbots consumed the missing capacity. Microsoft's account attributed its problem to network changes, not AI agents.2
Still, I find it difficult to believe that the industry's rush toward AI has no bearing on the conditions in which we operate these systems. How is it changing engineering priorities, spending, deployment pressure and the complexity we're prepared to accept? Those are questions I want investigated, not conclusions I can extract from an outage notice.
I'd rather spend the week asking those questions than following a billionaire's personal life.
An open model needs more than a champion
On November 12, NVIDIA announced Huang's immediate resignation as president, CEO and board member. CFO Colette Kress became interim CEO, and Harvey C. Jones became the company's first board chair. The announcement promised continuity in operations, customer commitments and the product roadmap.1
That is reassuring, but it isn't a specific commitment to Nemotron's future releases.
Huang's support for open weights was explicit. In September's announcement of NVIDIA's agreement to acquire Hugging Face, he argued that open weights distribute AI capabilities across companies, institutions and communities. He also promised that Hugging Face would remain open to competing computing platforms.4 The deal remains pending, another commitment worth watching through this transition.3
Nemotron is tangible work behind that position. NVIDIA publishes model weights, datasets and training recipes; its developer documentation includes local deployment options.5 Releasing those resources takes money and serious engineering. Developers receive something they can build with, rather than just an invitation to rent an API.
The question is whether the next release will give developers as much independence as the last one.
What I expect, and what I'd watch
My expectation is continuity before retreat. Open models make commercial sense for a company selling the hardware used to train and run them. TechCrunch's reporting makes that same distinction: Huang's advocacy mattered, but NVIDIA's interest in a broad developer ecosystem didn't disappear with him.3
A less dramatic change seems more plausible to me than an abrupt shutdown: slower releases, less published training information, or more attention to the hosted product than the downloadable model. That is a risk to watch, not something Kress has announced.
I'd apply the book's five questions to the next checkpoint: can we obtain it, what does its license permit, can we deploy it independently, how much of its development can we inspect, and could another team meaningfully reproduce it? Access, Permission, Deployment, Transparency and Reproducibility tell us more than another executive endorsement of “open source.”
The current Lightning release supplies downloadable weights and deployment instructions under OpenMDW-1.1.9 That license grants broad rights, with notice requirements and other terms.11 Give that permission credit. But permission to use today's release is a different thing from a promise to fund tomorrow's.
NVIDIA's November 17 earnings call is an early opportunity to ask for specifics.1 Will Nemotron's releases, data and recipes continue? Will independent deployment remain a priority?
After a week of broken dependencies, I want more than reassurance that the person at the top believes in openness. I want users to retain practical choices when leadership changes or services fail.
For Nemotron after Huang, that's the question: but is it open enough for others to carry it forward?
Commentary based on NVIDIA announcements, model documentation and reporting checked November 14, 2026.
Sources
- 1: NVIDIA: CEO transition announcement
- 2: GeekWire: Microsoft and FAA outages
- 3: TechCrunch: NVIDIA’s support for open models
- 4: NVIDIA: agreement to acquire Hugging Face
- 5: NVIDIA: Nemotron documentation
- 7: CBS News: nationwide ground stop
- 8: Microsoft service-health updates reproduced by ENow
- 9: Nemotron 3.5 Lightning model card
- 11: OpenMDW-1.1 license
- 12: CBS News: reported telecommunications-update failure
