6 comments on “Investing for a World Transformed by AI

  1. You’re misreading MS/OA. MS’s direct investment in OA is only a fraction, and that’s why they have also been investing in OA-run VC funds (https://openai.fund/about). The real gains is OA serving as a kind of R&D lab in tandem to MSR (still a major player in DL research). For MS, the gains include OA buying MS Azure (remember, the investment was in large part Azure credits) services, stress-testing its services as well, licensing the model source code (already done for GPT-3), integrating into their services like DALL-E into their graphics design or ChatGPT into Bing (https://www.theinformation.com/articles/microsoft-and-openai-working-on-chatgpt-powered-bing-in-challenge-to-google), visibility into hyperscaling (if you read about $1b capex on single datacenters stuffed with GPUs just to train+deploy DL models in the next 5-7 years, don’t be surprised), etc.

  2. Can you give a very brief summary of why you picked these?


    Or these?: ACLS, AMAT, INTC, LRCX, MU, and PLAB

    There are a lot of semi-conductor companies. My friends and I are interested in putting a lot of our net worth on these kind of bets and would appreciate more information. Even a very brief summary would help point us in the right direction.

  3. Sapphire, my choices are influenced by a variety of relatively ordinary considerations. PE, price/book, price / (earnings + R&D), etc.
    LSE:SMSN over TSM due to the war risk.

  4. You said REITs and property seem neutral. With the timeline you presented, I suppose I agree. I have a hypothesis that in the longer run–only slightly longer run–anything that can be manufactured or created that doesn’t have a fixed scarcity in some way becomes more valuable. You mentioned oil and gas, but in the long run, we’ll replace most of their applications with renewables and nuclear. Gold, other raw materials with intrinsic rather than instrumental value, and property in culturally desirable locations seem strong to me. Why property? There’s plenty of land in the world, but there’s only so much land in London, and New York, and San Francisco, and so on. I guess if you think metaverse and VR technologies can erode the proximity premiums those places possess, land might not be so worthwhile, but I’m sceptical place premiums will erode all that much.

  5. Ben,
    I think I mostly agree with you. Housing construction will be automated, but it will be one of the slower manufacturing processes to be automated. That seems unlikely to have much affect on REITs – scarcity of land will protect much of their business. Homebuilding profit margins will decrease, but that industry is already competitive enough that I don’t see a large effect.

    I would be bearish on oil and gas, for reasons I described in my Peak Fossil Fuel post. My AI forecast has altered that to a fairly neutral position: increased electricity use will keep natural gas demand from dropping much, but I’ve hedged my small positions in oil and gas stocks with short positions in long-dated crude oil futures.

  6. It occurs to me that I might want to quantify my advice. Here’s a portfolio that I think reflects what I intended when I wrote the post:

    Symbol #shares Jan 5 value Jun 30 value
    sehk:1799 1737 3165 3679
    acls 19 1566 3481
    amat 16 1567 2312
    amkr 123 3177 3658
    amzn 19 1579 2477
    aosl 113 3188 3704
    asml 6 3393 4349
    asys 383 3179 3642
    csiq 200 6358 7737
    cvlg 95 3177 4164
    dq 86 3193 3416
    googl 74 6379 8858
    intc 58 1598 1940
    invz 400 1586 1134
    jks 80 3197 3550
    klac 9 3359 4363
    lazr 367 1589 2530
    lrcx 4 1669 2570
    mram 288 1590 2606
    msft 7 1556 2384
    mtrn 36 3204 4112
    mtz 36 3174 4249
    mu 29 1587 1830
    myrg 34 3192 4704
    oss 2102 6348 6065
    oust 173 1589 855
    pick 76 3169 3075
    plab 95 1585 2450
    plpc 40 3141 6195
    prim 143 3184 4349
    scia 1716 6349 7293
    lse:smsn 3 4054 4990
    sym 125 1590 5351
    trt 709 3176 3418
    Total 101406 131490

Leave a Reply

Your email address will not be published. Required fields are marked *