At the New York Times DealBook Summit on November 29, 2023, Nvidia chief executive Jensen Huang told Andrew Ross Sorkin that AI capable of passing tests that reflect basic human intelligence could arrive within about five years, depending on how artificial general intelligence is defined. He also said the United States was a decade or two away from full independence in its chip supply chain, and described Nvidia's culture as one that still behaves as if the company is close to failure. The full interview is embedded below, followed by a summary of the main themes and why they matter for the chip industry.
Jensen Huang, the CEO of Nvidia, the leading maker of chips for artificial intelligence, discusses rising competition in the industry and why he enjoys living in a "state where we're about to perish." The interview was conducted by Andrew Ross Sorkin of The New York Times at the annual DealBook Summit, recorded live in front of an audience at Jazz at Lincoln Center in New York. Read more about highlights from the day at The New York Times DealBook Summit coverage.
Key takeaways
- On AGI: Huang said that if AGI means software that can pass a wide range of tests at a level fairly competitive with a normal human, it could be about five years away. He stressed that the answer depends on the definition.
- On supply chains: he said full independence of the US semiconductor supply chain was a decade to two decades away, while calling the effort worthwhile.
- On chip design: he agreed that AI is already part of how Nvidia designs its chips, and said the company's current chips would not be possible without it.
- On competition: he acknowledged that rivals and customers were building their own AI chips, and described a company culture that assumes it can fail at any time.
These points are summarized from the video and from press coverage published on November 29 and 30, 2023. They reflect what was said at the time and have not been updated for later events.
What was discussed
Artificial general intelligence and its definition
Sorkin asked Huang whether AGI would arrive within the next ten years. Huang's answer turned on definitions: if the bar is a system that does well on a broad set of tests of reasoning and knowledge, the kind of tests used to assess people, he expected that within about five years (CNBC report via NBC New York). Commentators noted that this is a narrower bar than many people mean by AGI, so the five-year figure should be read with the definition attached (see, for example, Fox Business).
Supply chain independence
Asked about efforts to move chip manufacturing away from China and Taiwan, Huang said, as Bloomberg reported: "We are somewhere between a decade and two decades away from supply chain independence." He described the push as worth pursuing but pointed out how deep and interlinked the global semiconductor supply chain is: materials, equipment, packaging, assembly and test are spread across many countries (see also the Taipei Times coverage).
Competition and culture
The video's title line refers to Huang's view that Nvidia should act as though it is always close to the edge. He has said in many interviews that the company nearly failed in its early years, and that the sense of urgency from that period is part of its culture. At the summit he discussed how hyperscalers and other chip companies were developing their own AI accelerators, and why he believed Nvidia's full-stack approach of chips, systems, networking and software would keep it competitive.
AI designing chips
Huang agreed that AI tools are already used in chip design and said Nvidia's chips, including the H100, could not be built today without them. For engineers, this is a useful marker: machine learning was already embedded in EDA flows by late 2023.
Deep dive: the details and how to read them
This part looks at the statements more closely and at what they mean for chip engineers. Skip it if the overview is all you need.
Why the definition of AGI changes the timeline
| Definition | What it requires | How Huang's answer applies |
|---|---|---|
| Passing human tests | Scores fairly competitive with a normal person on broad exams of knowledge and reasoning | The definition behind his "about five years" |
| General human-level ability | Learning and doing most tasks a person can, in the open world, without task-specific training | He did not give a five-year date for this; it is the broader sense most people mean |
For hardware planners, the practical point is the same either way: Huang expected demand for training and inference compute to keep growing for years, which is the assumption behind data center accelerator, memory and packaging investment.
What "supply chain independence" would involve
A leading-edge AI accelerator passes through many specialized steps before it ships, and each tends to be concentrated in a few places:
- Chip design, with EDA tools and licensed IP.
- Wafer fabrication on a leading-edge process, which needs lithography and other equipment from a small number of suppliers.
- High-bandwidth memory, made by a few memory companies.
- Advanced packaging, which joins the logic die and memory stacks on an interposer.
- Assembly, test, boards and systems.
Moving every one of these steps into one country means building fabs, packaging lines and supplier networks, and training the people to run them. That is why Huang's estimate was counted in decades rather than years.
Where AI was already used in chip design
By late 2023, machine learning was used in EDA flows for tasks such as placement, routing optimization, design-space exploration, verification coverage closure and computational lithography. These tools do not design a chip on their own; they shorten iterations and improve results on steps that previously relied on heuristics and engineer time.
Why it matters for the semiconductor industry
| Theme from the interview | What it means for chip makers |
|---|---|
| AI demand and the AGI timeline | Continued investment in data center accelerators, high-bandwidth memory and advanced packaging capacity |
| Supply chain independence a decade or two away | Multi-region manufacturing plans take many years, so near-term supply still depends on Asia |
| AI in chip design | EDA flows increasingly rely on machine learning, and design teams need people who can use those tools well |
| Custom chips from customers | Competition for AI silicon widens beyond traditional chip companies to cloud providers |
FAQ
When and where was this interview recorded?
On November 29, 2023, at the New York Times DealBook Summit at Jazz at Lincoln Center in New York, with Andrew Ross Sorkin as interviewer.
Did Jensen Huang say AGI would arrive in five years?
He said it could, if AGI is defined as AI that is fairly competitive with a normal human on a broad set of tests. He made clear that a different definition would give a different answer.
What did he say about US chip independence from China?
That full independence of the supply chain was somewhere between a decade and two decades away, although the effort was worth making.


Your comments will be moderated before it appears here.