Our website is made possible by displaying non-intrusive online advertisements to our visitors.
Please consider supporting us by disabling or pausing your ad blocker.
His memos are very useful.
I agree with what he said in the memo. What he doesn't know is Nvidia is a semiconductor player, and every single semiconductor business is cyclical in nature. When the CSP are satisfied with the amount of compute that they have, we will witness a huge drop in terms of revenue for every single of them.
However, I think the reality is still far from that. Nvidia chips are still under supply at this moment. We shall see what will happen next once TSMC new facilities start running this year.
He's not an equity investor but a credit investor. His calls are mostly chun, but I disagree with him at this moment.
AI demand is not driven by business (profit and etc), it's driven by do or die. For example Google, if they fail to acquire and build the necessary infra within these few years, their search business (Ads) will no longer be relevant in the future.
It's also a race between US and China. Whoever that successfully invent the AGI will win the race. That's the reason why US doesn't allow Nvidia to sell their advanced GPUs to China. The demand/race/real world applications are REAL...
Just ignore the stupid market noise (recession, bubble, and etc)...
Imagine your product is sooo great that Tesla and Oracle are begging you for more supply... Of course the demand will start to drop once there's no more chip shortage. But it's still far from that "inventory digestion" moment...
Every innovation eventually ceases to be an innovation; in doing so,it follow a path of market penetration and eventual saturation. This is a time less truth. Initially , every new innovation (railroads,automobiles ,radios, televisions , conputers , the internet,etc) begin at a relatively high price point that only a small market segment can afford. Progress in Technology and Manufacturing gradually reduces the relative price of the new product.
This leads to market penetration in which more and more potential users can acquire the new product or service. At some point market saturation is reached.: all firm and households that will buy and use the new product have access to it.
CISCO System….The darling of internet era,from March 1989 through May 1993: CISCO System reported 15 out of 17 quarters of earnings increases of 100% or more,with the other two quarters at 92% and 71% respectively. Cisco’s stock price advanced more than 13-fold during this period……..
Starting in April 2000, CISCO systems plunged 90% from its high in 30 months.
有人会说，NVIDIA 又不是 CISCO Systems…. 做的東西也不一样，那里可以比。
但是，当market crash 時，订单开始大量少时…….history will repeat itself……..
During the dot com bubble S&P 500 PE peaked at 44.2 b4 burst. The current S&P500 PE is about 29.72 still not as extreme as during the dot com bubble. Need to monitor the US economy strength if it starts to show weakness then probably the burst will come. For the time being, the US economy is still in good shape.
上网找pdf就可以了。。。我是howard marks的fans，这本书我当然读过，所以我知道他每次的bubble prediction都是准的。我只是point out他在memo里面没提到的几个点，还有我不agree的地方，毕竟他也说了不清楚这个industry（他老人家一直都是focus on credit的investment）。
million dollar question是，到底CSP的AI accelator需求到几时才见顶，到几时他们才觉得oversupply？肯定会有这么一天的，不过是几时？就比如我也知道总有一天会发生熊市，不过是几时？
Mark zuckerberg 在 Joe rogan 的poscast说了，probably在2025年，他们的AI可以code like a mid level engineer. It's very expensive to run at first then slowly it will replace their roles.
它是一个刚需，每一个愿意拥抱科技的企业刚需。没有他，你根本没机会在科技舞台上跟别人竞争。。。