Please create an account to participate in the Slashdot moderation system

 


Forgot your password?
Close
typodupeerror
×
AI Technology

TSMC Execs Dismiss OpenAI Chief's $7 Trillion Chip Plan as 'Podcasting Bro' Vision (msn.com) 91

Posted by msmash from the not-mincing-words dept.
Taiwan Semiconductor Manufacturing Co (TSMC) executives have dismissed OpenAI CEO Sam Altman's ambitious chip-making proposal as unrealistic, according to The New York Times. Altman, seeking to boost AI computing power, pitched a $7 trillion plan to build 36 semiconductor plants over several years during a visit to TSMC's Taiwan headquarters. TSMC leaders reportedly found Altman's proposal so far-fetched that they privately referred to him as a "podcasting bro," reflecting skepticism about his grasp of the semiconductor industry's complexities. The world's largest contract chipmaker, already grappling with multi-billion dollar expansion projects, viewed Altman's scheme as overly risky given the massive capital requirements and market uncertainties.

TSMC Execs Dismiss OpenAI Chief's $7 Trillion Chip Plan as 'Podcasting Bro' Vision More | Reply

TSMC Execs Dismiss OpenAI Chief's $7 Trillion Chip Plan as 'Podcasting Bro' Vision

Comments Filter:

  • The problem with tech bros is (Score:5, Insightful)

    by Rosco P. Coltrane ( 209368 ) on Friday September 27, 2024 @11:27AM (#64821695)

    However hare-brained their "visions" are, they're still dangerous and nefarious. Particularly Sam Altman.

    • Re: (Score:2)

      by shanen ( 462549 )

      Not a bad FP, but I'm wondering why you think Altman is particularly "hare-brained", "dangerous", or "nefarious".

      Well, I see a joke around the "hare-brained" part... How many Watts for a rabbit brain? Considering that the rabbit is in many ways more capable in the real world than large AI systems, it seems like the bunny wins. I've already mentioned the book from Jeff Hawkins that explains (among other matters) why the current AI summer is almost surely heading for another AI winter. (For PoC reference I ke

      • Rabbits aren't hares and hares aren't rabbits. Your comment is offensive to both >:-(

      • Considering that the rabbit is in many ways more capable in the real world than large AI systems,

        Well, the reverse is also true. Ask your bunny to help you programming (or any of the other thousands of different things people daily successfully use AI for). All you'll get is a confused look and some bunny excrement.

        it seems like the bunny wins.

        So no.

      • Re: (Score:2)

        by Rei ( 128717 )

        Well, you could ask his sister what she thinks of him...

  • good for them! (Score:3)

    by LazarusQLong ( 5486838 ) on Friday September 27, 2024 @11:33AM (#64821705)
    that 'vision' sounds entirely too optimistic, and doesn't appear to take into account anything in the real world... But that is just how it appears to me. I mean, what is the hard data that proves enough ROI to make that massive investment justifiable?
    • I mean, a vision doesn't necessarily need to have hard data behind it, but this sounds more like he was trying to make a pitch to this company to invest billions, hence the need for actual ROI projections based upon some level of reality.

      • Re:good for them! (Score:4, Insightful)

        by UnknowingFool ( 672806 ) on Friday September 27, 2024 @12:58PM (#64821959)

        a pitch to this company to invest billions, hence the need for actual ROI projections based upon some level of reality.

        Pitching the idea to build 36 chip fabs with $7T in the next several years is like pitching an idea to increase the birth rates in some countries by flying in a lot of women to those countries. Only on the very surface level does it make any sort of sense. The part Altman got right is the money required. The practicality and details are ignored.

        Let's start with building 36 factories has so many logistical issues to overcome. Zoning, land ownership, facilities resourcing (these plants use a lot of water and electricity). These are also specific type of factories that requires expertise to build. The local house builder, Bob, is not likely to know how to build these factories.

        In the case of chips fabs, the main limiting factor is even IF the site could be built with all the working facilities, there is only 1 company right now that make the lithography equipment, ASML. If you are not already on the waiting list to get a EUV machine in the next several years, the wait time could be 5 years at this point. Lastly, with all the equipment up and running, it is not guaranteed that fabs will instantly be making product. It takes time for ramp-up for a single fab. Right now TSMC is struggling with its Arizona fab due to cultural differences like US working hours and harsh treatment of workers.

        The divorce from reality is that Altman is pitching this to TSMC who has been doing this for decades. I imagine it would like pitching Apple on how to design smartphones.

        • yep, 100% his pitch was all dream and no reality. No wonder they didn't buy into it!

        • None of that is particularly hard to do. Factories are built every day, the problem is the investment. 7T doesnâ(TM)t fund 36 factories for their lifetime (unlikely even for the first 5 year) and by the time they are built, OpenAI will have been bought out or gone bankrupt. Even if you get to production, you now have 36 more factories on the market that have to sell, Altman isnâ(TM)t a capitalist so he doesnâ(TM)t understand that the market already is drowning in chips that are âoegood e

        • like pitching an idea to increase the birth rates in some countries by flying in a lot of women to those countries.

          Obviously a nonsense plan. Far simpler to fly in a few virile men instead.

        • a pitch to this company to invest billions, hence the need for actual ROI projections based upon some level of reality.

          Pitching the idea to build 36 chip fabs with $7T in the next several years is like pitching an idea to increase the birth rates in some countries by flying in a lot of women to those countries.

          Altman has essentially zero experience with building or operating fabs. He thinks that he just needs enough money and the fabs will build themselves ... like sort of how Intel is so successful.

          Who in their right mind is going to give him $7 billion to jump in and build and operate fabs. Altman is not crazy. He sounds like a great con man. The people giving him money ... those are the crazy people.

    • doesn't appear to take into account anything in the real world...

      Welcome to tech bro land. Two drink minimum.

  • Anyone can say big numbers. (Score:2, Funny)

    by Anonymous Coward

    I think a quadrillion dollars should be invested in chipmaking. Altman doesn't have real vision. Every man woman and child should be making chips.

    • Re: (Score:2)

      by cstacy ( 534252 )

      I think a quadrillion dollars should be invested in chipmaking. Altman doesn't have real vision. Every man woman and child should be making chips.

      Well AFAIK they all do.
      But we don't call them "chips" where I live.
      (And people do make trillions shoveling them. Like Sam Altman!)

    • Re: (Score:2)

      by zlives ( 2009072 )

      a few quadrillion dollars does sound about right, i think that will also give me my flying car and cold fusion

  • Those who know... (Score:5, Insightful)

    by Targon ( 17348 ) on Friday September 27, 2024 @11:45AM (#64821737)

    Intel put a LOT of money into trying to get their chip making back on track, and so far have failed miserably. It's not just about spending money, it's about having the right talent, even when you have the right equipment. The people at TSMC understand this all too well.

    • Re:Those who know... (Score:4, Insightful)

      by sg_oneill ( 159032 ) on Friday September 27, 2024 @12:14PM (#64821809)

      Right. If it was just a case of throwing money at it, the Chinese would have had the whole market sewn up long ago and you better believe Apple wouldn't be using TSMC fabs.

      But even if it could be done. Nobody is going to be giving Sam Altman 7 trillion to make chatgpt chips. AI might be important, but its not THAT important. Not yet, anyway.

      • Re: (Score:3)

        by m00sh ( 2538182 )

        Right. If it was just a case of throwing money at it, the Chinese would have had the whole market sewn up long ago and you better believe Apple wouldn't be using TSMC fabs.

        But even if it could be done. Nobody is going to be giving Sam Altman 7 trillion to make chatgpt chips. AI might be important, but its not THAT important. Not yet, anyway.

        China doesn't have access to the equipment now. They were on track and were strategically blocked.

        They don't have access to ASML equipment. They can't make their chips on TSMC either. However, they are welcome to buy Apple products made from these technologies but not the nVidia chips that could do AI.

        • Didn't read the thread huh? These guys won't have access to ASML hardware within the stated timeframe either

          • Any chip fab that has not ordered EUV machines from ASML will have to wait years for a machine. That is years for one machine. A chip fab is likely to use dozens of them. 36 chip fabs will need nearly a thousand EUV machines? Good luck

            • Re: (Score:2)

              by Targon ( 17348 )

              Even with that ASML equipment(which Intel did buy), Intel is still having problems getting the new fab processes working. This is where you can have professional tools, but if you don't know how to use them, you STILL won't be getting things done anywhere near as well as a professional. TSMC has professionals, Intel employees are wannabes.

              • The exact reason why Intel struggled so long on 10nm might come out some day. I heard rumors that it was because Intel bet on some sort of cobalt interconnect technology that was very finicky instead of copper. Supposedly cobalt theoretically had some improvement over copper. Ultimately the lower yields was not worth the improvement.

            • Re: (Score:2)

              by zlives ( 2009072 )

              clearly you are not understanding AI
              all you have to do is
              dump 7 trillion dollars in altman's bank account and profit.

            • Re: (Score:2)

              by AmiMoJo ( 196126 )

              Canon is making high end fab machines now too.

              I wouldn't underestimate China. Every time we do, they somehow get ahead anyway. I wouldn't be at all surprised if they had cutting edge fabs by the end of this decade.

              • Re: (Score:2)

                by PCM2 ( 4486 )

                I wouldn't underestimate China. Every time we do, they somehow get ahead anyway. I wouldn't be at all surprised if they had cutting edge fabs by the end of this decade.

                Sure, China has spies and they don't give a shit about intellectual property. But even so, your estimate seems optimistic.

                • Re: (Score:2)

                  by AmiMoJo ( 196126 )

                  They have a lot of engineers too, and pay the Taiwanese ones willing to come and show them how to build a fab a lot of money.

                  Think about Apollo. The US went from having just done a sub-orbital flight, behind the Soviet Union which did a full orbital flight, to landing on the moon in 9 years. When the government sets a goal and throws money at the problem, it can be solved remarkably quickly.

                  • Think about Apollo. The US went from having just done a sub-orbital flight, behind the Soviet Union which did a full orbital flight, to landing on the moon in 9 years. When the government sets a goal and throws money at the problem, it can be solved remarkably quickly.

                    You seem to ignore that US was behind the USSR by a month in 1961 when it came to orbital flights. Project Mercury started 3 years earlier in 1958. The US did not start from zero in 1961 to the moon in 1969 with no background. Also you seem to ignore that the US did not have crippling sanctions placed under them while trying to get to the moon. That is what has do. Build from zero lithography to EUV while under sanctions.

                    • Re: (Score:2)

                      by AmiMoJo ( 196126 )

                      China already has a working 7nm process. People thought it would take them years to develop that, and that the sanctions would make it almost impossible.

                    • China already has a working 7nm process. People thought it would take them years to develop that, and that the sanctions would make it almost impossible.

                      And where did China get that 7nm process? It is a modified DUV process from ASML. They did not develop any of that but are taking credit for it. That's like me buying a Tesla and saying I invented a new EV car line.

              • Canon is making high end fab machines now too.

                Canon has starting shipping nanoimprint lithography machines (NIL); however, even Canon says their machines do not compete against ASML's EUV [tomshardware.com]: "Canon says its strategy with its nanoimprint lithography machines is not to replace DUV and EUV tools from fabs but rather to co-exist with existing tools. If not for logic chips, NIL could be used to make 3D NAND. . ."

                I wouldn't underestimate China. Every time we do, they somehow get ahead anyway. I wouldn't be at all surprised if they had cutting edge fabs by the end of this decade.

                Japan and the US companies, while under no sanctions lost out to ASML when it comes to EUV. But you think China could do better. Sure . . whatever.

  • Process? (Score:2, Offtopic)

    by bill_mcgonigle ( 4333 ) *

    This is a bad time to start building AI fabs.

    Some bench prototypes use 1000x less power using different techniques.

    AI problems, at least some classes, are robust against small error rates.

    Traditional calculations need traditional chips with perfect error-corrected outputs.

    Engineering trade-offs are obvious.

    But OpenAI would love a $7T investment. In SV the biggest failures can be rewarded.

    Altman and TSMC guy are optimizing for different optima.

    • Re: (Score:1)

      by ndverdo ( 799508 )

      it's now almost 2 years after the release of ChatGPT. There are many vendors pitching x10 and more from ASIC to FPGA to Wafer. Hard to understand that in 2 years matrix-vector optimized for transformer training and inference is not all over the place in lieu of the absurd pricing for CUDA devices.

      Submitting a more than 100k context right now - just first step initially processing it is probably more than 0.5 kW.

  • Money is not capital (Score:4, Insightful)

    by RobinH ( 124750 ) on Friday September 27, 2024 @12:59PM (#64821963) Homepage
    This is what happens when you either think money can buy anything, or that money is capital. We use money to measure capital, but capital is actually the physical assets and human capital (available labour) that we can use to accomplish a project. When you're talking about the chip industry, which operates right on the edge of what's possible, the capital available to expand it by leaps and bounds simply doesn't exist, no matter how much money you throw at it. If you try, you'll just cause massive inflation in the industry. Yes, you'll end up spending 7 trillion dollars, but you'll only get marginally more for it.

    • Re: (Score:2)

      by m00sh ( 2538182 )

      This is what happens when you either think money can buy anything, or that money is capital. We use money to measure capital, but capital is actually the physical assets and human capital (available labour) that we can use to accomplish a project. When you're talking about the chip industry, which operates right on the edge of what's possible, the capital available to expand it by leaps and bounds simply doesn't exist, no matter how much money you throw at it. If you try, you'll just cause massive inflation in the industry. Yes, you'll end up spending 7 trillion dollars, but you'll only get marginally more for it.

      Well, if you seize/steal the physical assets and force the human capital to work for free, you don't need any money right?

      • Well, if you seize/steal the physical assets and force the human capital to work for free, you don't need any money right?

        Altman's plan was to build 36 new plants. How do you seize something that does not exist yet?

        • Re: (Score:2)

          by m00sh ( 2538182 )

          Well, if you seize/steal the physical assets and force the human capital to work for free, you don't need any money right?

          Altman's plan was to build 36 new plants. How do you seize something that does not exist yet?

          Seize Intels.

          They don't know how to use the equipment they have anyways and are just producing junk.

          • Seize Intels.

            And which Intel's plants are those? Intel has about 20 plants all over the world. [wikipedia.org] The vast majority are in the US. I do not see someone being able to "seize" them. They have a chip packaging plant in Malaysia that will open this year. There are not 36 of them to seize.

            They don't know how to use the equipment they have anyways and are just producing junk.

            Let me see if I understand you correctly. Altman's plan it to build 36 new plants for manufacture chips. Your solution is to seize plants in the US that are experiencing problems producing chips. How is your solution even under the best of ci

    • Re: (Score:2)

      by ceoyoyo ( 59147 )

      Capital is the things you use to make things, essentially tools.

      Labour is a separate thing, one reason for which is that we think owning capital is good but owning people is bad. "Human capital" is a finance invention of the same quality as "money is capital."

      I point this out because other than that you're right, and also believing in "human capital" is how you spend billons of dollars buying a company for "the team" and then everyone quits. Or a little more apropos to this story, converting your company to

      • Altman's plan is close to the idea of "If it takes one woman 9 months to make a baby, we should have 9 women do it and it should take 1 month". Altman narrowly only focuses on the fact that building chip fabs are very expensive to fund and that is one factor why there are not more: the capital required. However even with more capital, that does not mean TSMC can build 36 of them in a few years. With more capital, TSMC could build one or two more fabs than they planned. But Altman also does not realize that

      • Re: (Score:2)

        by RobinH ( 124750 )
        You're correct in that if you own a company, you can't really count your labour as capital. But if you're running a country you kind of can. While people can migrate, there are a LOT more barriers in place. And besides, if you're talking about the chip industry, which is the biggest and most complex industry on Earth spanning multiple continents, then humans really are a limited resource when it comes to expanding that industry. There are only so many people who can do some of those jobs.

        • Re: (Score:2)

          by ceoyoyo ( 59147 )

          Well, you can count your people as capital if you're running a certain kind of country, sure. Countries where the government tells you to do something and you do it.

    • You wonder whether or not these prophetic leaders know they are bullshitting or not.
      Every so often someone comes along at just the right time and rolls double sixes 100 times in a row. Musk. Trump. Now Altman. The man walks on water. When he picks his nose, journalists gasp and write that is the most amazing snot ever picked out of a nose.

      I do strongly suspect that he is just trying to leverage his influence to this purely ridiculous level of "vision" while he still has currency. I can imagine a lot of peop

      • Re: (Score:2)

        by RobinH ( 124750 )
        I definitely wouldn't put Musk in the same category as Trump and Altman. The latter are good bullshitters, and Musk is a terrible bullshitter. Musk's strength is that he understands engineering well enough to fund the right technical solution. But he's not a great swindler. Jobs was a good swindler, but he also had a relatively good grasp on what was possible, and the ability to convince people to do it.

  • The tech behemoths now want a nuke plant and a dedicated fab plant for each of their new AI data centers. Not happening, and even if it could happen it would take many years to assemble.

    Meanwhile a small fraction of that money could be invested in figuring out how to optimize existing resources. Make NPU's and TPU's that are much more computationally powerful and yet consume far less electricity. This has been the tech trend for CPU's and it would most likely be the best path forward for AI.

    • Other peopleâ(TM)s money. Nice try. Sam was just getting PR attention. Open AI can put up their cash they are currently burning see how far they get. Open AI can try selling more equity to help fund but these are very long term uncertain risks. In due time capacity will increase as market conditions favor and so will AI capabilities.
  • The word "bro" has become a fnord. If you think someone is acting evil, dangerous, or silly use precise language to describe what they're doing. We don't need another thought-terminating cliche or a manipulation of language.

  • One day, some day, some brainiac in a lab is going to find a way to perform the same AI calculations using some ingenious new mechanism with 1/100th the power demand and 1/10 the circuitry, if any circuitry at all - maybe biologics will be the next big thing. If that happens halfway into a $7-trillion chip plant build-out, a lot of investment money is going to be stranded high & dry. And I'll need a new graphics card.

  • Mock them! Mock them! Mock them like the fools they are. The Emperor has no clothing, AI has no profitability.
  • Two men build a house in one year. Four men build a house in half a year. 16 million men build a house in one second.

Slashdot Top Deals

Do not underestimate the value of print statements for debugging.

Close