The demand we are seeing for GPUs right now is basically driven by model building and very light weight 'consumer' experimental inference. When businesses actually start integrating gen ai into automated high volume business processes, running all day every day, the demand for gpus to do inference to support that is going to make current demand look trivial. That adoption is going to be slow--integrating ai into existing systems is not simple--but it is coming.
Awesome show. Love Michael and the Compound crew. I was talking to my wife and I believe that AI revolution (4th Industrial Revolution) will be bigger than the internet. Computers have all been about input with basic output. The way we will input will be simpler but the output much more complex and sophisticated. For example, imagine that you have a human assistant. You want to book a business trip to London for 5 days in July. Your assistant will arrange transport, lodging, meals, scheduling meetings, etc. based on your needs and tastes. Now imagine that an AI assistant can do the same. How much time, money, and resources have you saved as a result? This was just a simple example but can extend to everything that we do on a daily basis. The AI will know what we want better than we do and help us make decisions faster with more efficiency. As for Nvidia, you have to own the stock. I don’t care if you think that it is overvalued and will lose 90% of value. We simply don’t know how much run way AI has and if what I assume is true, Nvidia could double or triple in size within the next 10 years. I bought the stock last year worth about 3% of my portfolio and it has grown to 10% Most of my investment is in VOO and other ETFs but 30% in AAPL as well. Using a small percentage of your investment portfolio to hold momentum stocks like Nvidia will potentially pay off huge.
Great show. The biggest pressure point for Nvidia will come on the earnings calls of Microsoft, Meta and Amazon. Where do they guide Wall Street on capex spend for next year and then eventually 2026. Once that slows it will mark a turn in NVDA the stock.
16:55 Alex suggests that this is an accelerated computing server upgrade boom as opposed an AI specific boom. Similar could happen with edge computers.
@@Alex.kantrowitz Jensen Huang refers to “accelerated computing” as the function of Nvidia’s server chips. So he considers other uses than AI learning.