Put in $2000, then purchased:
$300 for BABA
$317.92 for OWL
$500 for META
$500 for PSUS
$700 for TSM
The AI infra play had a bloodbath these few days. Alphabet went down 12% after earnings, Intel went down 30% after earnings, META went down 7% after earnings, etc. The suppliers of the AI Infra also fared poorly, with TSM down about 5% from a month ago since my last portfolio update, NVDA down about 10%, SMH down about 10%, DRAM down about 20%.
Besides Intel, which has its own headwind from the softness PC market, the common bearish theme was that the AI capex spending did not look sustainable. The richest companies in the world, basically the hyperscalers like Alphabet, Meta, Amazon, have negative free cash flow, how much more can they spend?
On the other hand, the AI plays do not have a demanding valuation. The chip makers like Nvidia and TSM are trading at below 20 forward P/E. The hyperscalers, if you adjust for this and that, are trading at below 20 forward P/E as well. That means the market is pricing in a pretty bearish AI compute demand. Maybe it's not pricing in less AI compute, but it's definitely pricing in a big deflation, i.e. end users, consumers, will buy more compute for less.
With open models from China, there is a fair chance that people will buy more, both in terms of capability and quantity, for less. However, as a whole, I believe the demand for LLM AI is so vast in agents, robotics, media productions, etc, that the spending from end users of AI compute will grow fast enough to overcome any price decreases, so the AI infra as a whole will earn more money every year for at least 5-10 years. That makes their current valuation pretty cheap. That's why I allocate more to TSM in this portfolio update.
I also bought some BABA. It's been over two years since my last purchase of the shares. It went down quite a bit this year, at about 45% down from the $180 peak a few months ago. It's the ultimate AI cloud compute play in China. Unlike Alphabet and Amazon, other than the cloud business, Alibaba doesn't have a substantial business that has meaningful growth to fund its capital expenditure on AI. But I believe most capex can easily be passed on to its cloud customers, so it's just a timing lag between its spending and earnings. The high AI compute demand from the cloud will provide Alibaba a lot of earnings for at least 5-10 years. At a market cap of $250 billion, it is very cheap.
META is an interesting plan in tech. While it spends a lot in AI compute, it doesn't earn money from AI spending by other companies. It is pretty much an "end user" that uses AI to increase its productivity, the efficiency to operate its social and ads businesses. Meta's recent AI agent, Muse, is just an option value of the company. Meta doesn't need Muse to succeed for its earnings to keep growing. Cloud business is another option value that I wish Meta doesn't try, but hopefully it will not hurt it much. The capex in cloud will definitely hurt the company's free cash flow for years to come, but it doesn't hurt the fundamentals of the business in social and ads. People enjoy spending time on social media, so the mindless scrolling will not be replaced with AI agents.
Finally, coming back to the alternative managers. Their stocks still suck. Besides tech, my portfolio underperformed the general market because of the alts. While Blue Owl went up a few percent since a month ago, Apollo went down 8%, and my total return of the position went negative excluding dividends. Anyway, I still like OWL a little bit better given its lower P/DE (price-to-distributable-earnings), higher dividend yield, much less reliance on realization and equity, higher percentage of permanent or semi-permanent capital, etc.
Transactions
Recent and upcoming dividend distributions
Portfolio performance snapshot
Total return:
One-year return:
Portfolio IRR (calculation): 14.63%
Approximated IRR for a hypothetical SPY-only portfolio: 16.98%
Individual holdings:
Breakdown by categories (real-time):
Total returns for individual holdings:
Last prices:
Portfolio holdings conviction
The convictions in the table below reflects my current opinions and will guide the future contribution of additional investment to existing holdings. Stocks not inside the table are stocks with subpar return on equity that will be very unlikely to receive more contributions from new money (there can be exceptions for very cheap stocks). All of my writeups can be found here.
Conviction in long-term prospects means how much I believe a company would match or outperform the market (e.g. S&P 500) in the long run. Valuation matters so the conviction generally corresponds to the neutral rating of Valuation. It has the following ratings: weak, moderate, strong
Valuation: greatly overvalued, overvalued, slightly overvalued, neutral, slightly undervalued, undervalued, greatly undervalued
