Intel CEO Pat Gelsinger speaks at the World Economic Forum in Davos, Switzerland — 1/17/2024

you know I put semiconductors into four buckets memory chips you know dams flash memories into specialty chips you know these would be like sensors for cameras or RF chips for analog radios mature chips you know these would be things like for microcontrollers right and older cars and then Leading Edge right and today you know the Leading Edge Market is hot largely driven by AI right because you know imagine if you have an old node chip maybe you do 10 tops if you have a modern no chip you do 200 tops right you know it's just dramatically different you know the performance capabilities so those are the four major categories of semiconductors as I like to uh view it and obviously when you're making you know Factory commitments that take five six years to come into manufacturing you know tell me what the demand signal is next quarter you can't even answer that question much less what the demand signal is 5 to eight years in the future right and these are long lead Time Investments so you end up with crazy cyclicality in the industry because you know markets go up and down but you build factories that are five years into the future and the only way to run a semiconductor Factory you know a huge Capital asset if you run it empty it cost X if you run it full it cost X so you always run it full right and that creates some of these you know crazy economics around semiconductors and today we're in a reasonable position of Supply demand balance obviously we've had an oversupply memory we're probably going to go into an UND supply of memory because nobody's been investing in it you know we have lots of investment going into Leading Edge you know I expect there's going to be a lot of pressure on mature because uh China in particular has been building a lot of capacity uh in that area but uh right now as we think about the decade in front of us AI is this insatiable demand for compute right insatiable right we're going to be building the biggest machines that have ever been built and we're going to be driving a AI to the edge that everything right that we touch will be fueled by AI capabilities and it's going to need more capabilities as a results so we're going to see this being you know I think we have 10 great years in front of us for advanced logic requirements driving you know the industry and many of these factories that we're building uh right now you mentioned cutting at uh chips and uh uh may ask you we you I come back to this 80 20 5050 uh how does it apply to um so Cutting Edge chips and but but maybe first I want to ask you why did uh West or us particularly and Europe lose out and uh all those Cutting Edge chip factories I think 95% are now located in Taiwan if I'm not mistaken yeah Taiwan in particular a couple of other you know k as well you know but fundamentally as I like to as I like to describe it never was a vote taken in Brussels or DC to get rid of this industry but there were votes taken in the Asians uh uh countries Korea Taiwan China to get this industry you know they put in place long-term industrial policy tax investment Etc to attract this industry and 30 years later they attracted the industry right and all of a sudden in covid we all woke up and we realized oh my gosh what happened 30 years of industrial policy and we lost something so critical to our future the world needs balanced resilient Supply chains and that's what created the environment for the US and the EU chips act now when you look at I mean already 2016 if I'm not mistaken um China put out uh several development plans but some people uh even today China has not yet uh come up to the level of Taiwan um and uh why is it so difficult could you tell us to produce those Cutting Edge yeah as I like to say you know if you want to be a cuttingedge semiconductor supplier you just need to invest 20 30 billion of capital per year you need to invest you know 8 to 10 billion of R&D per year and you need to do that for 30 years and how much do you invest you know in technology you know we do 18 billion of R&D per year as a company per year you know uh we you know we'll do over 25 billion of capital last year that's going up next year it we'll go up the year after that you know this is extraordinarily longterm R&D and capital intensive it's a consolidating industry there's only three companies in the world that are capable of doing these Cutting Edge chips tsmc Samsung uh and uh Intel you have key Technologies like the euv technology from uh asml key research like iMac in Belgium you know and uh you know frankly Europe has many of the core Technologies and as I like to say between us Europe and Japan are the source of all of these core Technologies it's expensive it's R&D it's extraordinary Innovation and uh sometime clus I'm going to get you uh into a bunny suit in one of our Fabs because you these Fabs you know they're almost $30 billion to build one of these complexes these are the largest construction projects on Earth happening today to build the smallest things that have ever been built on Earth this is extraordinary right and what's being done and they're done then at scale over and over again we're producing these extraordinary chips you know today our most advanced chip is about a 100 billion transistors you know on a single package you know and as we continue the progression of Moore's Law And as I like to say until the periodic table is exhausted Moore's Law is alive and well and we're going to hit a trillion transistors by the end of the decade right in a single package device you know this is just extraordinary but if I understand directly it's not so much the development which is costly it's what what you call the foundaries and uh can you can you explain why sir are quite a number of Manufacturers but so few foundaries if I'm not mistaken so five foundaries or five to six I don't know yeah you know being able to you know lots of people can design chips and this idea of you know fabulous semiconductors you know it's R&D Centric right but they don't it's not capital and Manufacturing Centric there's very few people that manufacture you know semiconductors and only as I said three that can do Leading Edge you know today and you know this is very expensive very orang d uh intensive extraordinarily intellectual property heavy you know we have you know 100,000 plus patents are associated with the Leading Edge semiconductor manufacturer thousands yeah you associated with it you know it's just you know quite stunning right what's been built up over the 50y year history of the semiconductor industry and every aspect of your life needs more of them going forward so if you built for example now a factory I think in mber uh in in Germany um is it a Foundry or what or for Cutting Edge or is it uh just a normal chip factory no this is Cutting Edge that's Cutting Edge yeah and this will be you know we expect when it comes online you know our most advanced process technology you know that we're just uh soon to bring into manufacturing is what we call 18a sub 2 nanometers you know this will be beyond that so this will be on the order of 1 and a half nanometer devices that will build in mag bur so this will be not only the most advanced Manufacturing in Germany it'll be the most advanced Manufacturing in the world will occur you know in the magur site you know we're quite excited about uh getting that underway and to I know that uh Chancellor Schultz is very excited indeed indeed uh Deputy Chancellor Hab he he referred to you and probably saw you indeed but the most advanced manufacturing today in uh Europe you know is what we do in Ireland at 4 nanometers you know outside of us you know it's probably in the 10 to 15 nanometer range this is a huge leap forward for all of Europe so if I understand you correctly you may design chips and you may um Outsource the production to someone else to The Foundry right and we're going to be a we design chips and we're a Foundry for chips as well uh going forward but the intellectual property is with you right for The Foundry but the intellectual property for the design is with the design company yeah which you are both now this essentially I'm rebuilding Intel we're taking one broken company and we're creating two distinct companies inside of Intel a manufacturing and a Foundry Company and a product and fabulous company inside of one yeah thank you now you refer to artificial intelligence and uh there's one chip factor I don't want to mention the name who is very much uh talked about because uh uh his or its chips are particularly related to a AI yes could you explain and um are you also in this area or what what is the um what is special about AI yeah and you know this idea of you know high performance accelerated Computing right and I give credit to my friend Jensen right you know uh that hey they have pursued that domain pretty consistently for 20 years you know high performance throughput Computing and then he got lucky with AI really lucky but he had steadfastly pursued that architectural pursuit of high performance Computing largely in graphics you know for a period of time and then AI happened and remember AI is an overnight 50-year success yeah right the foundations of AI you know were in the late 60s right and then for over 40 years nothing happened right when I was the architect of the 46 right in the 80s I was going to make it a great AI chip what happened in the 80s and AI nothing right and then all of a sudden the compute got big enough the data got big enough the algorithms got good enough and we've seen this explosion of AI right and right now we're seeing these very large systems for training right and we think about you know GPT and open Ai and so on but the next several years of AI won't be around big model training you know for hundreds of billions and trillions of parameters the next several years of AI in my estimation is how do we use those models right how do we deploy them how do we inference them and how do we move them out of the Big Data Centers into every device that we use we've just launched what we call the aipc right how do we make you know instead of having a 100 billion parameters in the cloud how can I put 10 billion parameters of your data on your PC that you're operating locally and I call it the three laws of Edge and AI you know one is the laws of economics if it's on your device locally much more cost effective 10 100 times cheaper you know second is the laws of physics if I have to round trip to the cloud the speed of light is still the speed of light right versus doing it locally and the third is the laws of the land right the regulatory requirements am I going to take my realtime factory data to somebody else's Cloud environment for local inferencing of my manufacturing line absolutely not 80% of the data is still on premise and privately held so because of that we see this idea of going from training which still has exciting science in front of it to much more inferencing deployment and Edge uh you know applications and that's why I'm you know quite excited because there's going to be both we're going to continue to see the science of AI evolve you know but very much the deployment and utilization of it was where we believe you know the primary action will be for the next couple of years so the research comes mainly from inside the companies not from University and from the scientific world today because at the origin it was mainly the scientific World which drove the semiconductor development yeah and of course you are big benefici you are in your first uh stage of career you were very much at the one of those drivers but today can you tell us uh how much um the research is driven by industry and how much by so scientific world yeah you know uh and we've seen this shift of more and more of the research and development to Industry over a sustained period of time you know one piece of the chips Act was uh you know the establishment of nstc in the US you know we had hollowed out that Leading Edge uh research and one of the things I believe you know needs to you know for you know Western uh you know Democratic oriented countries we must rebuild that long-term research you know long-term research is 30 plus years AI took over 40 years to hit maturity the transistor took over 30 years to maturity you know we must keep that long-term investment and that's P primarily you know from government's public sectors dark but like things we have lost that because businesses like mine we're sort of like 15 years or less in the R&D and we've seen that shift over time and I think that's very unfortunate and now as we think about the science of AI I you know in particular you know today uh you know you remember Daniel Conan's book Thinking Fast thinking slow you know all of our AI systems today are thinking fast yeah we haven't brought reasoning into AI it's a huge area of research and most of the foundations of that you know were seeds planted 20 30 years ago about how we can prove correctness of systems today our systems hallucinate tomorrow if we're going to use them broadly they have to be right so how can they be fast and right right just like you know we've seen Thinking Fast and thinking slow huge areas of research in front of us huge areas you know as I like to say until the periodic tables exhausted we're not done yeah well we have about two-thirds of the periodic table to go still lots of new science new materials to be discovered you know uh into the future this is uh you know what I think for a technologist the next couple of decades I just wish I was still in my 20s right the next couple of decades are going to be absolutely spectacular the exponential face oh my gosh it's going to be so good clous you know right you know how old are you you and I need to find a way to get younger cuz it's going to be a thrilling few decades we keep young we keep young you see but how does it come that I come back to this 8020 semiconductors are essential element for National Defense and National Security why uh did uh the US particularly but also Europe have it slipping out yeah I think people were somewhat asleep at the wheel from the policy perspective there were you know a few people you know raising the uh concerns but remember we were in an era you know where I you know it was all about lowest cost Supply chains you know hey we don't care if we manufacture you know we're going to create the intellectual property we'll we'll invent the algorithms we'll write the software right you know we're sort of happy to bring other countries in the world into that manufacturing that's hard that's low endend we really don't want to work right you know we want to invent right and I think that was the overall General views right literally of the western world for a couple of decades then all of a sudden we realized in critical Industries right and I think the automotive industry is facing this right now you know we could lose the entire automotive industry because we've lost the supply chain right and the semiconductor industry you know literally you know we were on the hair's edge of never being able to recover the manufacturing of this industry you know I've uh you know said to some that you know had I started the rebuilding of Intel a year later I don't think I could have accomplished it right it was that much on the edge that we would have outsourced too many generations of Technologies and never been able to rebuild it in the future because particularly for semiconductors R&D and Manufacturing are comingled right you know a factory without R&D in the semiconductor industry you know essentially becomes a boat anchor within 6 months right you know it is that you know the because you're constantly doing research in the factory in the manufacturing you know process itself R&D and Manufacturing are inextricable in this industry what is the dependence of the industry on materials it's actually is this a limiting factor or you know it's actually surprisingly not that terribly constrainted right in that sense you know we don't use that many materials remember God's gift to mankind was silicon right sand the most prevalent material on earth right is our core material so he gave us lots of it right you know in the refining of it's in you know Japan and Asia right and Europe you know so the core materials are actually not that severely limited other portions of the supply chain but silicon itself isn't that bad but we are taking steps to make sure we have built resilience and sustainability you know the environmental aspects of the chemicals and that we use is a super important you know factor to us you know Intel is decades ahead of anybody else in terms of the sustainability you know we've committed to uh you know by 2030 you know full uh you know 100% renewable in our factories in our type one and type two you know full sustainability of our water use 2040 Net Zero you know so it's a huge and important topic you know for us over time and any place we find you know critical material limitations is how do we go you know expand that because remember most of the critical minerals issues are not minerals issues they're refining issues right you know where the minerals you know they're usually fairly available it's where they refined are the key points of constraint but could energy be a limiting factor because some people are concerned about the high consumption of electricity yeah and today you know when you think about the IT industry as a whole different studies but probably somewhere in the 3 to 5% range of energy consumption you know goes for it infrastructure today different studies would indicate you know in particularly with the surge of AI you know that that could triple or quadruple right over the next decade so that could become quite constraining you know to build these big massive data centers which by the way puts pressure on us how do we keep reducing the power requirements of our chips as well as you know things like water cooling right so that we can increase the efficiency of them as well another area of significant research I come back to China um when do you think China will um have caught up with it really with the frontiers of uh chip Technologies yeah you the uh export policies that have been put in place you know recently we've seen the the Dutch uh in particular you know in place the US policies Japanese policies Etc it sort of puts a floor in the tend to 7 nanometer range right you know for it and you know we you know are racing to go below 2 nanometers and then one and a half and you know we see no end to that in sight you know and I see the policies in place you know and it's not like China's not going to keep innovating but this is a highly interconnected industry you know the staging of trump right the mirrors of Zeiss you know the equipment assembly of uh asml the chemicals and resist in Japan the mass making of Intel all of those together you know I think this is a 10year gap and I think it's it's a sustainable 10-year Gap with the export policies that have been put in place today you know and I do believe that that portends well for the policies that have been established in the world you know right for export and uh competitiveness right in this uh environment and you know we're leaning hard into making sure that's the case now so so many Investments going on like in murg and so on I think there plans in France plans in Ireland additional ones aren't you afraid about uh because every country wants to have now its own uh uh asset uh aren't you afraid about over capacity well if you go back to the beginning of the discussion right you have this hey we're making eight you know five and eight-year Investments and we have cyclicality quarters I mean you can't even tell me what the demand is next quarter you know much less you know 5 years from now that said you know I think fundament compute and AI in particular are going to be like the gas law they will expand to consume every amount of compute capacity that is economically available you know so I believe that becomes such a driver of demand that I think Advanced logic capabilities you know I have no concern in this at all now of course when I explain to my board my crazy Capital requests I of course have to say that not only say that but say it with conviction now we have I think five minutes to go let me see there's any question here in the room there must be yeah please yeah johnman executive chairman of The International Institute for strategic studies to basic first is that I am told once you get one there is perhaps a limit a is that true uh and the second uh question I have is I am equally told that when Morris Chang establishes a Foundry in Arizona or elsewhere he does not have the confidence that that Foundry can reach the same level of expertise as his own in Taiwan uh for reasons of human capital so can you explain when you're producing something uh 80,000 times bigger than a a human hair why is the human skill so important yeah so you know on the first topic the way I like to describe it is Moore's law is like driving down the road on a foggy night right you maybe see 100 meters in front of you when you've gone 20 or 30 meters down the road you see about 100 meters in front of you and Wars law we see about a decade in front of us right and I've been in the industry now for 42 years right you know almost embarrassing to say it's been that long but uh you know with it right you know I've heard about the death of Moors law you know for about three decades right and I sit here today telling you I know how to build a trillion transistor chip by the end of the decade Moors law is alive and well and we're going to keep making it true because it's not a law of physics right as you suggest when you get to one nanometer you know you it's a law of innovation of commitment of economic pouring into the criticality of this technology and we as a company Intel are going to keep making it true second uh on uh you know uh you know tsmc is an extraordinary company you know they have refined a manufacturing capability and a customer orientation that is really unmatched in the industry however they do R&D and Manufacturing in one place Taiwan right Intel has operated with the distributed manufacturing footprint for over four decades we know how to operate in different segments of the world we know how to do this we're about you know I see my my best friend Senator Portman here from Ohio we're about to start the silic and Heartland uh in the center of the U of the United States bringing manufacturing back to the heartland of America you know with it we know to operate distributed manufacturing they don't yet right and this idea of tethering of R&D and Manufacturing right he's right they only have one place of R&D you know I have a distributed R&D model I have a center of R&D uh in Oregon and Arizona that are extraordinary in the world we're absolutely convinced we know how to do this let the race continue they're a great company but we're committed to being back to unquestioned technology leadership in the world and I think by 2025 we will have proven that to the world with manufacturing of sub2 nanometer chips that are simply the best in the world you know based on Western technology and we're pretty proud to be in that race how many people do you have in research and development uh you know we're uh on the order of 50,000 Engineers total right in the TD Spa you know in the core technology space in the order of 20,000 you know so yeah this is you know impressive yeah this is at scale R&D right let me we have time for one last question any anybody yeah sen report there well indry the difference that ships act makes and as a co-sponsor of that bill and having heard from a lot of my Republican colleagues about industrial policy and how that's not our Direction um some of them thought we were overreaching and doing too much uh the reality is there will still be considerable dependency uh even when you combine the EU and the US and you touched on that a little bit but if you could talk a little more about that I think I think going about 37% of chips 30 years ago um to about 12% today and even after 52 billion dollar and about 39 billion on the manufacturing side my understanding is will only be at what number in terms of the percentage of we'll be crawling our way back into the 30s right you know with chips one you know I do believe there'll need to be a chips 2 uh at some point to continue building those policy decisions and remember these were bad policy decisions in the US and tax policy right in capital investment policy long-term R&D Investments all of those aggregately caused us to go from 80 to 20 right to rebuild that we're not going to do that in one act as good as it was thank you Senator right you know in a few years this will take decades to rebuild this industry but I really believe that chips one Chips 2 fundamental industrial policy you know improvements in R&D improvements in manufacturing you know Capital uh policies you know rethinking of some of the financial requirements long-term you know human capital development I firmly believe that we can rebuild this industry and it's the right thing for our economies it's the right thing for our Industries and it's the right thing for our national security and we're deeply committed to help drive that you know simply put it's the right thing for the heritage of Intel it's the right thing for the technology industry and it is the right thing for the world we're going to make it happen thank you Pat we we have come to to an end of this disc and I think what we have seen is a leader with with something which we so much need today when we talk about leadership it's conceptual uh thinking um and you are in one of the most complex areas U Hightech uh at the at the edge and um I I think we all leave this with a better knowledge about the indust industry and impressed by your by your leadership thank you very much thank [Applause] you

Channel: CNBC Television

Views: 31,969

Keywords: CNBC, business news, finance stock, stock market, news channel, news station, breaking news, us news, world news, cable, cable news, finance news, money, money tips, markets, nyse, investing, traders, trading, stock trading, dow, s&p 500, stock charts, live markets, stocks to buy, stocks to trade, stocks and bonds, stocks and crypto, stocks explained, markets live, markets news, markets cnbc, markets explained, investing in stocks, investing 101, investing basics

Id: hw-kesm-ktI

Length: 29min 50sec (1790 seconds)

Published: Wed Jan 17 2024