WHEN: Today, Monday, September 22, 2025
WHERE: CNBC’s “Halftime Report”
Nvidia CEO Jensen Huang, OpenAI CEO Sam Altman and OpenAI President Greg Brockman joined CNBC’s Jon Fortt in the present day to announce a strategic partnership to deploy 10 gigawatts of Nvidia techniques. Following is the unofficial transcript of the interview that occurred on “Halftime Report” (M-F, 12PM-1PM ET) in the present day, September 22. Following are hyperlinks to the video on CNBC.com: https://www.cnbc.com/video/2025/09/22/nvidia-plans-to-invest-up-to-100-billion-in-openai-as-part-of-data-center-buildout.html, https://www.cnbc.com/video/2025/09/22/nvidia-ceo-on-the-100-billion-investment-in-openai-this-partnership-is-monumental-in-size.html and https://www.cnbc.com/video/2025/09/22/nvidias-jensen-huang-says-immigration-is-really-important-to-our-nations-future.html.
All references should be sourced to CNBC.
JON FORTT: I’m right here at Nvidia headquarters in Santa Clara with the CEO of the world’s most useful firm and the CEO and president of the world’s most useful non-public firm, Jensen Huang of Nvidia, Sam Altman, and Greg Brockman of OpenAI. So let’s dive proper into the information. Jensen, Nvidia is making a $100 billion funding in OpenAI and working collectively to construct out, I believe you are saying 10 gigawatts of capability over a number of years. The funding goes to come back with the gigawatts, one by one. You’re telling me, you guys, as rapidly as you guys can get it accomplished. Jensen, initially, set the stage. Why?
JENSEN HUANG: This is the largest A.I. infrastructure mission in historical past. This is the most important computing mission in historical past. Well, the rationale for that’s as a result of computing demand goes via the roof for OpenAI. You know, ChatGPT is the only most revolutionary A.I. mission in historical past. It’s getting used in all places, each trade, each nation. Every individual virtually that I do know makes use of ChatGPT. The computing demand goes via the roof. And so this partnership is about constructing an A.I. infrastructure that allows A.I. to go from the labs into the world. This is concerning the A.I. industrial revolution arriving. It’s a really massive deal.
FORTT: Yeah. 100 billion {dollars}, some huge cash. Sam, Greg, you guys are used to dealing with some huge cash in massive initiatives. So, Sam, I believe it was simply eight months and a day in the past the preliminary Stargate announcement speaking concerning the total transfer that OpenAI is making in constructing out this capability. Where does this match?
SAM ALTMAN: So, as Jensen mentioned, constructing this infrastructure is important to all the things we wish to do. Without doing this, we can not ship the companies folks need. We cannot hold making higher fashions. And now that we actually see what’s on the near-term horizon of how good the fashions are getting, the brand new use instances which can be being enabled, what folks wish to do, that is just like the gas that we have to drive enchancment, to drive higher fashions, to drive income, all the things. So that is serving to us get to a world, alongside with our companions at Stargate, Microsoft, Oracle, the place we will construct out rising quantities of infrastructure to ship on what the world is demanding out of those companies. There’s like no accomplice however Nvidia that might do that at this type of scale, this type of pace. It’s actually, like, fairly, fairly unimaginable. But this can develop on the Stargate ambitions and allow us to push additional and additional. We have discovered each step alongside the best way that we didn’t fairly set our sights sufficiently big, given the market demand. So this can assist us push in the direction of that subsequent stage. The compute constraints that the entire trade has been – and our firm, particularly, have been horrible. We’re so restricted proper now within the companies we will provide. There’s a lot extra demand than what we will do. And as we glance ahead one other yr or two years, when you’ve got – to illustrate it takes 10 gigawatts of compute or 5 gigawatts of compute, you could possibly select certainly one of two issues. You may select to remedy most cancers by doing a bunch of – having A.I. do a bunch of analysis, or you could possibly select to supply free training to everyone on Earth. No one desires to make that selection. And so more and more, as we see this, the reply is simply way more capability in order that we will serve the large wants and alternative with this.
FORTT: Greg, a technique of measuring that demand is the variety of customers that OpenAI has. ChatGPT says someplace between 700 and 800 million. How rapidly are you rising? What’s the know-how that you simply guys must carry to bear from Nvidia included as a way to fulfill that?
GREG BROCKMAN: Yes, so 700 million, rising extraordinarily rapidly. I believe that ChatGPT has grown from nothing to that scale and continues to develop quicker than any product in historical past. And, actually, the rationale we have been in a position to make this breakthrough and serve it to the extent that we’ve is by leveraging Nvidia’s platform. Like, we’ve labored collectively very intently since Jensen truly hand-delivered a server to us again in 2016. We have been simply doing a little math earlier. I believe that this deal is admittedly for a billion instances extra computational energy than that preliminary server. And so we’re in a position to truly create new breakthroughs, new fashions to have the ability to truly clear up issues, like create cures for ailments, and to truly be capable to empower each particular person in enterprise as a result of we can attain the following stage of scale.
FORTT: Jensen, step again for me. I imply, a couple of days – you are everywhere, actually. I imply, you are within the U.Okay. I noticed the white bow tie and all that, the tux.
ALTMAN: You appeared very good, Jensen.
FORTT: Here you look very good. But you are additionally doing numerous investments. The Intel funding introduced final week fairly a bit smaller than this one, however appears vital additionally, as a result of it is weaving Nvidia know-how within the PCN knowledge middle stage in a means that maybe it wasn’t earlier than. Where do these sorts of investments slot in? How do you consider the worth of the ecosystem to Nvidia?
HUANG: The Intel partnership is about recognizing that accelerated computing and A.I.’s day has arrived. Remember, common function computing was invented virtually 60 years in the past. And for the final 60 years, we’ve been following that fundamental blueprint, that fundamental structure to construct the ecosystem, the computing use of the world. And so, impulsively, accelerated computing’s time has come. And we’re fusing, if you’ll, the Intel structure with the Nvidia structure to carry them into the world of accelerated computing and A.I. So that is what that partnership’s about. This partnership, I imply, that is monumental in dimension. There’s by no means been an engineering mission, a technical mission, of this complexity and this scale ever. And it actually simply says that A.I. was within the early adopter part within the labs, and, lastly, it is breaking out into nearly each single trade, each single use case we will think about. It may be very quickly the place each single phrase, each single interplay, each single picture, video that we expertise on – via computer systems will by some means have been reasoned via or referenced by or generated by A.I. It’s going to be touched by A.I. by some means. So all of our computing experiences all through the day in all places in each trade will probably be powered by A.I. This first 10 – that is the primary 10 gigawatts. Surely, it appears like an infinite enterprise, however there is not any query that A.I. is transformational for each trade. But the necessary factor is, the A.I. infrastructure will probably be in all places and will energy computing experiences for everybody on daily basis. And it will be simply in all places.
FORTT: Sam, it looks as if the 2 most useful firms on the earth, Nvidia at primary over $4 trillion – about $4.25 trillion final time I checked in the present day – Microsoft just below $4 trillion, are going to be main traders in OpenAI. How ought to we perceive the governance, the affect that these firms have over OpenAI, the place these traces are and even, you realize, funding size-wise, the place that is going to web out?
ALTMAN: We’re thrilled to have them each as companions. They’re passive traders. Our nonprofit and board are in management, however the ecosystem is admittedly necessary to us. Nvidia and Microsoft are two of our most crucial companions and have been from the very entrance, from the very starting. And having them so aligned with our success is, I believe, nice for us, hopefully nice for them too.
FORTT: Greg, the place does Nvidia match alongside with the entire different infrastructure suppliers? Oracle – you guys type of touched Oracle and the inventory went up, although the title wasn’t connected to it. When they reported earnings and their steering, lots of people have been trying over at OpenAI. Where do these totally different know-how infrastructure gamers match when it comes to significance within the quantity of capability that you simply want?
BROCKMAN: Well, look, the mission that we’re attempting to do is one thing that, as Jensen was saying, is greater than any infrastructure build-out that is ever been accomplished. It’s a lot bigger than the Apollo program, for instance. And so we’re going as massive as we will with the largest companions on the earth. Nvidia is a core strategic accomplice for our build-out. There’s simply nobody who may construct as quick or as massive as they’re going to have the ability to speed up us to have the ability to do. We are working collectively with Oracle to do numerous the infrastructure builds with SoftBank and Stargate, to have the ability to truly do a bunch of that work as properly. We’re beginning to develop into simply attempting to determine each single means that we will truly get this compute to the world. But there’s actually been no accomplice like Jensen, like Nvidia. And it is simply been a really, very particular partnership for greater than a decade now. And we’re shifting right into a subsequent part collectively.
HUANG: Now, one of many issues that is actually necessary to say is that — and Sam and Greg hinted at it — that is additive. This mission, 10 gigawatts of A.I. infrastructure, is additive to all the things that has been introduced and contracted. Remember, they’ve contracted large quantities of capability via Azure, via OCI, via CoreWeave. And all of that’s powered by Nvidia. And we’re actually delighted working with all of those companions. And that is going to proceed to develop. This is additive, incremental on prime of that, which simply type of places it in perspective, the dimensions of A.I. computing that is wanted for the world.
FORTT: Well, I’m clearly not asking you –
HUANG: The demand is simply exponential.
FORTT: Yeah. To say something totally different from what’s already been included within the financials that you’ve got given or the steering that you’ve got given. But that is an announcement that is breaking proper right here on CNBC. Has this been factored into the numbers that you’ve got given Wall Street up so far?
HUANG: This is additive to all the things that we’ve spoken about to this point. It’s fairly unimaginable.
FORTT: We’re additionally speaking about A.I. infrastructure for the world. One of the headlines final week needed to do with China and how they’re dealing or not dealing with Nvidia. When you have a look at the place this capability is getting constructed out and the worldwide use that it will have, how do you consider the best way totally different areas, the best way totally different international locations are going to faucet into that and how that impacts Nvidia’s competitiveness?
HUANG: Well, President Trump was clear about this. And you could have heard me say this earlier than, that we wish the world to be constructed on the American tech stack. And the American tech stack, the A.I. computing stack consists of chips, infrastructure, fashions and functions. We have each, each cause to consider that the United States ought to lead in each single a kind of layers. And in doing so, we wish the world to be constructed on prime of American chips. We need American infrastructure, American fashions, and for the world to combine with American functions. And we must diffuse this functionality as quick as attainable as a result of the world is racing to carry A.I. out to the world. And so I believe the necessity to construct A.I. infrastructure all around the globe, that is only the start. Notice, many of the infrastructure we’re – infrastructure conversations we’ve been having are largely situated within the United States. But we’ll see A.I. infrastructure inbuilt Europe and within the southern elements of the world. You’re going to be, proper, Southeast Asia. All over the world, we’ll be constructing A.I. infrastructur
FORTT: Sam, there was a DeepSeek second a number of months in the past. The market type of freaked out about what this implies for infrastructure, for capability, for expense. What was the legacy of that? Given the second that we’re in proper now, this $100 billion funding Nvidia is making in you, clearly, you see worth in these chips, otherwise you would not want that type of cash for that type of tools. What does that imply? How has the best way the dialog round A.I. fashions has shifted affected the best way that you simply guys are trying on the build-out?
ALTMAN: Two ideas about this. First of all, I believe many of the world nonetheless thinks of A.I. as what ChatGPT can do. You know, it is a greater model of Web search or it helps me with some small duties right here and there. A.I. has moved an unimaginable distance. A.I. is now outperforming people on the most tough mental competitions we’ve. For the primary time with GPT-5, you are beginning to see scientists saying A.I. is making novel discoveries, small ones, however actual ones. And so there’s – I believe there’s this large mismatch on the earth of what most individuals assume A.I. can do and what the frontier of A.I. can do. And so when DeepSeek got here out, I believe folks had this temporary freak-out. I believe perhaps your inventory went means down in at some point, and folks have been identical to, oh, that is like the tip. People actually wish to predict the tip of the compute scaling by some means. And then it turned out that individuals want numerous A.I., and they want numerous the frontier A.I. And so it’s completely true that the associated fee per unit of intelligence will hold falling and falling and falling, and we predict that is nice. We will be capable to provide companies like ChatGPT, extra of it, cheaper, the entire thing. But on the opposite aspect, the frontier of A.I., the utmost mental functionality, goes up and up, and that allows extra and extra use, and numerous it. So there’s this large overhang that the world, I believe, doesn’t but grasp of the place the fashions already are in the present day, and while you throw numerous inference pc, then what they will do for you, that’s completely totally different than ChatGPT or generate a picture or no matter. And so I believe that was actually missed within the DeepSeek second, and why we wish to do that. It’s that there is – the fashions are at this level like truly fairly succesful for issues far past what most individuals use them for in ChatGPT, and the world is simply catching up with that. Second factor, we throw round these numbers, 10 gigawatts, $100 billion, and there’s a couple of syllables right here and there. We’re – it is like a Monday morning. We’re all type of like low vitality. We acquired numerous stuff we acquired to go do and pressured concerning the day forward. But the –
HUANG: And we’ve already had a reasonably lengthy day.
(LAUGHTER)
ALTMAN: We have had a protracted day. But the magnitude –
FORTT: This deal, sure, I imply, I flew out right here not understanding precisely who I used to be speaking to, when this got here collectively. So, sure, thanks, however yeah.
ALTMAN: For certain. Thanks for taking a flier on that.
(LAUGHTER)
ALTMAN: But the magnitude of the dimensions of this mission, $100 billion is a small dent in it. And the numbers are additionally like, they’re lacking the story of what this quantity of infrastructure is able to doing. Like, 10 gigawatts of compute, once more, simple to throw round numbers like that. But the quantity of labor it takes to construct that out, the dimensions and scale of those like multisquare-mile, gigantic issues and the complexity at each stage of the availability chain, and then what that quantity of brainpower, which doesn’t exist in the present day, can do already in the present day, what it’ll do because the fashions get higher, like, that is the actual deal. This is the factor folks have been ready for. They speak about A.I. or when it is going to do that and when it is going to do this. Like, the stuff that may come out of this tremendous mind will probably be outstanding in a means I believe we do not actually know the way to consider but.
FORTT: 100 billion is a small dent. Now you sound like President Trump.
(LAUGHTER)
BROCKMAN: One approach to contextualize the dimensions of what we’re speaking about and the compute shortage of the world that we’re heading in the direction of, you realize, ChatGPT in the present day, you discuss to it, it provides you solutions. But clearly you need an agent that is going to go do give you the results you want proactively when you’re asleep, be capable to manage your calendar or go attempt to work on initiatives for you. And so that you really need each individual to have the ability to have their very own devoted GPU, proper? So you are speaking on order of 10 billion GPUs we’ll want. This deal we’re speaking about, it is for tens of millions of GPUs. Like, we’re nonetheless three orders of magnitude off of the place we should be. So we’re doing our greatest to offer compute availability, however we’re heading to this world the place the financial system is powered by compute, and it will be a compute scarce one.
FORTT: So, Sam –
HUANG: You know, Jon, keep in mind, this new mission that we’re speaking about, 10 gigawatts or roughly 4 million, 5 million GPUs, that is roughly in a single mission what we shipped all yr this yr, OK, and twice as a lot as final yr, twice as a lot because the yr earlier than that. And in order that type of places it in perspective. This is a huge mission. And but –
FORTT: And we do not but know what number of years to amortize that over, type of, but it surely’s –
HUANG: Well, we simply want – OpenAI – I imply, OpenAI’s the quickest rising software program firm in historical past.
ALTMAN: We will see how briskly we will construct.
FORTT: So, concerning the infrastructure, we have been speaking about knowledge facilities, however I additionally wish to discuss to you concerning the edge. And I suppose Sam or Greg, this could possibly be both of you. Apple had their massive launch of promoting iPhones final week. Lots of people thought, oh, they’re behind on A.I. They’re attempting to push this concept that these telephones, as a result of the chips they’ve in them are A.I.-ready. And, in fact, they’re speaking to you guys. Are these a part of the infrastructure that you simply want? What ought to the world perceive concerning the function of the Edge and gadgets like this?
ALTMAN: Totally. This is the place – there’s this necessary concept, which is many of the world thinks about A.I. as ChatGPT. And but A.I. functionality has moved far past. So, for the type of chat with ChatGPT about most stuff, it is possible for you to to do numerous that on your cellphone. And we predict that is nice. We launched an open-source mannequin known as GPT-OSS lately that may run on a laptop computer or a cellphone. And it is unimaginable to me what you are able to do on a tool. It’s gone means additional than I believed. And we might like to see numerous that transfer to the Edge. Curing most cancers, like an A.I. that thinks actually onerous and cures most cancers, that may want the large knowledge middle. But for, like, the type of customary ChatGPT queries, we hope we will push numerous that to the Edge.
FORTT: So what is the significance of those platforms, whether or not it is from Apple, whether or not it is from the likes of a Qualcomm, in having the ability to present a regular platform for you guys, for different companions, for different utility suppliers to construct A.I.-driven functions on prime of, that are going to drive a few of this demand that we’ve been speaking about right here?
BROCKMAN: I believe you want all of it, proper? I believe that, regardless of how a lot will get constructed, we’re nonetheless going to be on this world of compute shortage. And each time there is a new compute platform, new availability of this type of computational energy, that it is attainable to design new algorithms, new strategies, new fashions, new merchandise which can be in a position to ship advantages to everybody. And so it is best to count on that you will note this complete ecosystem of fashions from us and from everybody else that is ready to reap the benefits of what’s on the market. It’s a bit bit like when the App Store first got here out, out of the blue, folks began creating apps and filling that platform as a result of there was this new availability, this new means of constructing. And I believe we’re heading in the direction of a platform shift that is going to be way more vital than something in historical past.
FORTT: Another potential, both roadblock or alternative is expertise, proper? You have gotten to have the correct folks to construct these items, to design these items, et cetera. Immigration is part of that. Jensen, there have been some strikes, some bulletins over the previous few days about that. How does that place the U.S., perhaps the H-1B situation and the additional cost on prime of that, how does that place the U.S. when it comes to competitiveness? What are your ideas on the place we have to go?
HUANG: We need all of the brightest minds that come to the United States. And, keep in mind, immigration is the muse of the American dream. And we characterize the American dream. And so, I believe immigration is admittedly necessary to our firm and it is actually necessary to our nation’s future. And I’m glad to see President Trump making the strikes he is making.
FORTT: What type of coverage wants do we’ve out of your perspective at OpenAI?
ALTMAN: We must get the neatest folks within the nation. And streamlining that course of and additionally type of aligning monetary incentives appears good to me.
FORTT: How far are we from yet one more massive announcement, proper, concerning the capital wants and the way you hope to meet that?
ALTMAN: You ought to count on loads from us within the coming months. We are actually – there are three issues that OpenAI has to do properly. We have gotten to do nice A.I. analysis. We need to construct these merchandise that individuals actually wish to use. And we’ve to determine how to do that unprecedented infrastructure problem and build-out, which there’s chips, there’s energy, there’s buildings, there’s a lot of items that go into that. And that latter class goes to be my massive focus within the coming month, and we’ve very formidable objectives. So, we may have loads to say.
FORTT: Jensen, the following massive catalyst that you simply see on the horizon? I imply, A.I. is what’s driving a lot of the market and the creativeness proper now of economies internationally.
HUANG: Look, we noticed within the final 5, 10 years A.I. transfer from generative to reasoning to now pondering fashions. Went from textual content to multimodal. It may be very clear that there will probably be a persistent A.I. linked to each machine to be all the things that is on the market. It could possibly be a persistent A.I. linked to the automotive, linked to sensible glasses, linked to the cellphone as you talked about, linked to human or robots, and robots of all totally different sizes and shapes. And all of that’s simply starting. None of it exists in the present day, and but all the things’s across the nook. That’s how thrilling that is.
ALTMAN: Yes.
HUANG: We’re actually going to attach intelligence to each utility, to each use case, to each machine. And we’re simply initially of that. That’s the rationale why we’d like such gigantic infrastructure. And that is the primary 10 gigawatts. I guarantee you of that.
FORTT: Well, I’m trying on the CNBC app, and there’s Nvidia’s inventory proper now. We cannot see OpenAI’s inventory but.
ALTMAN: Oh, sure.
FORTT: But name me once we can. Scott and the Halftime crew, again to you on this breaking information from Nvidia headquarters right here in Santa Clara.