ComputePublished Top storyAll news from that day
Ongoing case: What the AI buildout costs and needs 11 storiesNvidia chief Jensen Huang says a gigawatt of data centre costs about $50bn
Interviewed by Yahoo Finance's Deirdre Bosa at GTC Berlin, in a video published 21 October 2026, Huang also said Nvidia's chips for AI factories last 'eight to maybe more' years.
Check our sources · 14 facts from 2 sources
Video: Yahoo Finance on YouTube
Key points
- Huang said "a gigawatt of data center is like 50 billion dollars" (01:46).
- Huang said going from landing at a data centre to operating "could be 4 to 6 months", and might shorten to "maybe 3 to 4 months" (02:20).
- Huang said a four-year chip life cycle "really is probably eight to maybe more" (06:42); asked whether Blackwell lasts 7-8 years, he answered "Probably." (08:32)
What happened
Jensen Huang, Nvidia's founder and chief executive, was interviewed by Deirdre Bosa of Yahoo Finance in Berlin, straight after his keynote at Nvidia's GTC Berlin event. The video, "Nvidia's CEO on the AI debt scare, China, and jobs | Full interview", was published on Yahoo Finance's YouTube channel on 21 October 2026.
Bosa described Nvidia's DSX pod as an AI factory in a shipping container and asked whether it would make data centres easier to finance. Huang said "a gigawatt of data center is like 50 billion dollars" (01:46), and that from the point computers land at a data centre to operating "could be 4 to 6 months" (02:01).
He said "you might be able to shorten that 4 to 6 months to maybe 3 to 4 months" (02:20) by doing some of the manufacturing in the supply chain.
On how long the chips stay useful, Huang said "what people thought used to be a four-year life cycle really is probably eight to maybe more" (06:42). Bosa asked whether he meant seven to eight years for Blackwell, and he answered "Probably." (08:32).
For Vera Rubin, the next generation, he said "It'll be, you know, 7-8 years" (08:41). He said "you might be able to underwrite a loan against the asset itself" (07:06), and that the risk in underwriting Nvidia's computing "is the lowest of any asset out there" (07:58).
On open models, Huang said "I think 600 billion dollars worth of venture financing went into AI native startups" (13:51) in the first half of this year, and "80% of those AI native startups are using open models" (13:58). The figures are Huang's own; he named no source for them in the interview.
Underlined sentences link to the facts we checked. Dotted words open a short explainer.
Fact 6 of 14 · from Yahoo Finance (YouTube), 21 Oct 2026
those computer computers land at a data center. From that point to operationalizing could be 4 to 6 months.
Fact 7 of 14 · from Yahoo Finance (YouTube), 21 Oct 2026
what if we were to do some of the manufacturing of that in the supply chain. The advantage is that you might be able to shorten that 4 to 6 months to maybe 3 to 4 months
Fact 8 of 14 · from Yahoo Finance (YouTube), 21 Oct 2026
what people thought used to be a four-year life cycle really is probably eight to maybe more.
Fact 13 of 14 · from Yahoo Finance (YouTube), 21 Oct 2026
the first 6 months of this year, several hundred billion dollars, I think 600 billion dollars worth of venture financing went into AI native startups.
What it means for you
Our viewThese are the views of Nvidia's chief executive, whose company sells the hardware, so treat the $50 billion per gigawatt and the chip lifetime as a supplier's estimate.
The lifetime matters for anyone costing AI capacity, because a longer useful life spreads the same cost over more years, and Huang argued that lenders could underwrite a loan against the asset itself (07:06).
The 7-8 year figure for Blackwell was the interviewer's, which he accepted with "Probably." If you budget for AI compute or sign multi-year capacity deals, ask your supplier which useful-life and depreciation assumptions its price rests on.
It adds no new facts.
Your reaction
We count reactions per story and day, never who reacted. The counts help us choose what goes in the monthly issue. If you are signed in, your own page shows yours too.
We count reactions per story and day, never who reacted. They help choose what goes in the monthly issue.
Check our sources
Every sentence above is checked against these 2 sources.
1 Jensen Huang | NVIDIA Newsroom
Open the source archived copy-
Jensen Huang's title on NVIDIA's own newsroom biography is Founder & CEO.
cite
2 Nvidia’s CEO on the AI debt scare, China, and jobs | Full interview
Open the source archived copy-
The video 'Nvidia’s CEO on the AI debt scare, China, and jobs | Full interview' is on the Yahoo Finance YouTube channel and was uploaded on 2026-10-21.
citeNvidia’s CEO on the AI debt scare, China, and jobs | Full interview; channel Yahoo Finance; uploadDate 2026-10-21T04:34:03-07:00
-
In the video, Deirdre Bosa introduces herself and says she is live in Berlin, joined by Nvidia CEO Jensen Huang fresh off the stage at GTC Berlin (00:00).
citeHi everyone. I'm Deirdre Bosa live in Berlin and I am pleased to be joined by Nvidia CEO Jensen Huang fresh off the stage here at GTC Berlin.
-
Bosa asked about the DSX pod, which she called an AI factory in a shipping container, and whether giving lenders more certainty about a data centre's cost and start of earnings would make it easier to finance (01:14).
citeDSX pod. An AI factory in a shipping container. ... So if you can give lenders essentially more certainty about what a data center costs and when it starts earning money when it comes online do you think that makes it easier to finance?
-
Huang said 'a gigawatt of data center is like 50 billion dollars' (01:46).
citeSo a gigawatt of data center is like 50 billion dollars.
-
Huang said that from the point the computers land at a data center, 'From that point to operationalizing could be 4 to 6 months' (02:01).
citethose computer computers land at a data center. From that point to operationalizing could be 4 to 6 months.
-
Huang said 'you might be able to shorten that 4 to 6 months to maybe 3 to 4 months' by doing some of the manufacturing in the supply chain (02:20).
citewhat if we were to do some of the manufacturing of that in the supply chain. The advantage is that you might be able to shorten that 4 to 6 months to maybe 3 to 4 months
-
Huang said 'what people thought used to be a four-year life cycle really is probably eight to maybe more' (06:42).
citewhat people thought used to be a four-year life cycle really is probably eight to maybe more.
-
Bosa asked whether Huang had said 7-8 years for the durability of Blackwell, and Huang answered 'Probably.' (08:27-08:32).
citeDid you say 7-8 years a moment ago? Is that what you think the durability is for Blackwell? >> Probably.
-
Asked about Vera Rubin, the next generation, Huang said 'It'll be, you know, 7-8 years' (08:41).
citeVera Rubin is probably going to be the same. It'll be, you know, 7-8 years
-
Huang said 'you might be able to underwrite a loan against the asset itself' (07:06).
citeyou might be able to underwrite a loan against the asset itself.
-
Huang said 'the risk involved in underwriting Nvidia's asset and our Nvidia computing is the lowest of any asset out there' (07:58).
citethe risk involved in underwriting Nvidia's asset and our Nvidia computing is the lowest of any asset out there.
-
Huang said 'I think 600 billion dollars worth of venture financing went into AI native startups' in the first six months of this year (13:51).
citethe first 6 months of this year, several hundred billion dollars, I think 600 billion dollars worth of venture financing went into AI native startups.
-
Huang said '80% of those AI native startups are using open models' (13:58); he named no source for the figure in the interview.
cite80% of those AI native startups are using open models.
Every sentence above is checked against these 2 sources. Each fact has its own address you can cite.
Source 1 · NVIDIA
Jensen Huang | NVIDIA Newsroom
Source 2 · Yahoo Finance (YouTube) · 21 Oct 2026
Nvidia’s CEO on the AI debt scare, China, and jobs | Full interview
The video 'Nvidia’s CEO on the AI debt scare, China, and jobs | Full interview' is on the Yahoo Finance YouTube channel and was uploaded on 2026-10-21.
Nvidia’s CEO on the AI debt scare, China, and jobs | Full interview; channel Yahoo Finance; uploadDate 2026-10-21T04:34:03-07:00
In the video, Deirdre Bosa introduces herself and says she is live in Berlin, joined by Nvidia CEO Jensen Huang fresh off the stage at GTC Berlin (00:00).
Hi everyone. I'm Deirdre Bosa live in Berlin and I am pleased to be joined by Nvidia CEO Jensen Huang fresh off the stage here at GTC Berlin.
Bosa asked about the DSX pod, which she called an AI factory in a shipping container, and whether giving lenders more certainty about a data centre's cost and start of earnings would make it easier to finance (01:14).
DSX pod. An AI factory in a shipping container. ... So if you can give lenders essentially more certainty about what a data center costs and when it starts earning money when it comes online do you think that makes it easier to finance?
Show all 13 factsShow fewer facts
Huang said 'a gigawatt of data center is like 50 billion dollars' (01:46).
So a gigawatt of data center is like 50 billion dollars.
Huang said that from the point the computers land at a data center, 'From that point to operationalizing could be 4 to 6 months' (02:01).
those computer computers land at a data center. From that point to operationalizing could be 4 to 6 months.
Huang said 'you might be able to shorten that 4 to 6 months to maybe 3 to 4 months' by doing some of the manufacturing in the supply chain (02:20).
what if we were to do some of the manufacturing of that in the supply chain. The advantage is that you might be able to shorten that 4 to 6 months to maybe 3 to 4 months
Huang said 'what people thought used to be a four-year life cycle really is probably eight to maybe more' (06:42).
what people thought used to be a four-year life cycle really is probably eight to maybe more.
Bosa asked whether Huang had said 7-8 years for the durability of Blackwell, and Huang answered 'Probably.' (08:27-08:32).
Did you say 7-8 years a moment ago? Is that what you think the durability is for Blackwell? >> Probably.
Asked about Vera Rubin, the next generation, Huang said 'It'll be, you know, 7-8 years' (08:41).
Vera Rubin is probably going to be the same. It'll be, you know, 7-8 years
Huang said 'you might be able to underwrite a loan against the asset itself' (07:06).
you might be able to underwrite a loan against the asset itself.
Huang said 'the risk involved in underwriting Nvidia's asset and our Nvidia computing is the lowest of any asset out there' (07:58).
the risk involved in underwriting Nvidia's asset and our Nvidia computing is the lowest of any asset out there.
Huang said 'I think 600 billion dollars worth of venture financing went into AI native startups' in the first six months of this year (13:51).
the first 6 months of this year, several hundred billion dollars, I think 600 billion dollars worth of venture financing went into AI native startups.
Huang said '80% of those AI native startups are using open models' (13:58); he named no source for the figure in the interview.
80% of those AI native startups are using open models.
Topics
The morning email
On the mornings we publish: the three top stories and up to four short ones. Free.
We email you a link to confirm. An issue may include one sponsor, always labelled Sponsored · Advertisement. Our emails count opens and clicks, not who made them. Unsubscribe in one click. What we keep
