treasurehunt

Won't FIH have to pay capital gains taxes on any sales of Anchorage? The Prosus sales of Tencent work so well because they are tax free. I guess if the discount is big enough, paying taxes is worthwhile.

Looks like Alibaba has joined the ranks of the AI maximalists. Here's the transcript (in Chinese) of Eddie Wu's keynote at the Alibaba Cloud Summit earlier this week: https://www.infoq.cn/article/Da7a5caUL75cqvmmZtKg I had ChatGPT translate the transcript for me. Here's the very high level summary: Thesis: Achieving AGI is now a near-certainty, but it’s only the starting line. The destination is ASI—Artificial Superintelligence. On this path, large models become the operating system, and AI Cloud becomes the next-generation computer. I didn't see any timelines mentioned, but it didn't seem like Eddie Wu was talking about multiple decades to get to ASI. Also, I didn't see any discussion of the dangers of developing ASI.

Trimmed some more BYD and will probably sell out of this position entirely. I think I overestimated their competitive advantages and underestimated the competition. The EV industry in China is mature enough that BYD's vertical integration is no longer a big edge. Foreign sales should be a bright spot for a while, but the competition will arrive there too in time. This, plus the China risk as a US investor and the fact that BYD's businesses are capital intensive, means that the risk level is too high for my comfort. Of course, now that I've said all this, the stock is going to zoom straight up.

Damn, I was hoping Milei would get at least four years to try out his libertarian approach. Seems unlikely now. I thought things were going better than many expected, but I guess not in the eyes of the Argentinian electorate.

Yikes! I took another look at its work. Market caps for NVDA, AMZN, TSLA and META are accurate, but ChatGPT is way off on AAPL, MSFT and GOOG. Didn't expect that. I just tried Gemini as well. ChatGPT actually does better than Gemini, which thinks the total market cap is $16.57 trillion.

All interesting questions. ChatGPT gave me the following numbers for the Mag 7. I verified some independently, so they are probably in the ballpark. Market cap: $17.5T GAAP Net Income: $551B FCF: $380B FCF - SBC: $288B Using these three measures of earnings, the P/E ratios work out to 32, 46 and 61. It seems to me that the market is expecting AI investments to work out well for these companies.

I'd say this is far from worthless; in fact, it's probably very valuable. Typically, it is much easier to check if an answer makes sense than it is to come up with an answer in the first place.

I don't think we are in agreement. I'm not saying the market is pricing in epic returns from capex; just that it is not pricing in low returns. There is a lot of room between low and epic. Also, as Eldad pointed out with excellent examples, P/E is just one measure of valuation. I'm sure you know that P/E can be misleading for various reasons. Finally, I would generally expect valuation measures based on sales or earnings (P/E, P to FCF, P/S etc) to decrease over time for growth companies, all else equal. For example, Google is almost five times as big today by revenue as it was 10 years ago. It will be tough for future growth to match past growth unless something big and new is on the horizon. The market seems to think AI is such a thing.

I don't know this. In fact, I don't have a strong view about this at all. You said this in your post though: "Everything I read seems to be along the lines that AI is a bubble and that the mag 7 are just dumping hundreds of billions into low returning cap ex." I was just pointing out that the market isn't pricing in this view that apparently permeates everything you read.

ChatGPT was released to the public on Nov 30, 2022. From that day till yesterday, the S&P 500 had a total return of about 58%. AAPL, which one might consider the most prominent non-AI tech stock, returned just over 63%. In contrast, here's how the five biggest tech companies that are usually considered beneficiaries of AI did over the same period. Ticker Nov 30, 2022 Sep 3, 2025 Total Return NVDA $16.91 $170.62 909% MSFT $249.65 $505.35 102% AMZN $96.54 $225.99 134% GOOG $101.45 $231.10 128% META $117.46 $737.05 528% It seems clear that the market is not pricing in enormous wasted capex at these companies.

I have a question regarding the calculation of underwriting leverage in this table. The table uses NPW (net premiums written) to calculate this number. But isn't the standard to use net premiums earned? Fairfax mentions specifically in their latest earnings release that they calculate the undiscounted combined ratio as underwriting expense expressed as a percentage of net premiums earned. Also, given the Q2 underwriting profit of $426.9 million and CR of 93.3%, net premiums earned works out to $25.5B. So underwriting leverage should be 1.0, and not 1.1 as claimed in the table. Obviously this is a small difference (0.8% pre-tax), but I'd like to know if I am understanding this correctly.

I am hardly an expert here, but since no one has responded to you yet, I'll chime in. I have friends and relatives who work in these companies in Bangalore and this is based on what I have heard from them. First off, these companies - Infosys, Wipro, HCL, TCS, Accenture, IBM, Cognizant, Tech Mahindra, LTIMindtree etc - employ upwards of 500K in Bangalore. There was a recent report that tech employment in Bangalore has exceeded 1 million, and most of the 1 million are likely software jobs. These folks are also relatively well paid, so they have a disproportionate impact on the city's economy. Now, Bangalore is a huge city, with a population of over 14 million. It's also pretty diverse, with a lot of employment in aerospace, biotech, finance etc. But if the software consultancy industry gets decimated, I think that will really hurt. Maybe in time, the other industries can make up for it. How are these companies doing right now? Not great. There have been some layoff announcements, and the rate of hiring of "freshers" has dropped off. I don't think there is a clear plan on how to deal with AI (not surprising, because the impact of AI is probably not entirely obvious). If AI turns out to be really disruptive in the short term, Bangalore's economy and BIAL by extension, will probably get hit.

That's an excellent template. A 4% investment yield leads to a 12.5% operating ROE according to this table. That is a lot closer to 15% than I would have guessed. It remains to be seen whether 93% is a good estimate for the long-term combined ratio.

Yes, good point. Even adding in a reasonable estimate for that, 4% is too low for 15% ROE, I think. Net earned premiums are around $25B I believe, so a combined ratio of 96 would add 4% to pre-tax ROE or 3% post-tax. But my estimate that the portfolio needs to return 6%+ is too high. Thanks.

How do you come up with 4%? The investment portfolio is about 3 times equity ($70B vs $25B or so). Even assuming zero expenses, a 4% return on the investment portfolio is only a 12% ROE. And realistically, you have deductions for interest on debt, corporate overhead, non-controlling interests, run-off and taxes. Non-controlling interests and run-off may not be relevant in the near future, but the rest are here to stay. I estimate that the investment portfolio will have to return better than 6% for Fairfax to hit the 15% ROE target. What am I missing?

Atlas seems to be operating well and Fairfax will do very well with this investment, I think. There is this bit in the 10-Q on transactions with Fairfax. During the three and six months ended June 30, 2025, the dividends paid on Series J Preferred Shares held by Fairfax were $5,250,000 and $10,500,000, respectively (2024 – $5,250,000 and $10,500,000, respectively). Fairfax remains a counterparty to certain indemnification and compensation arrangements related to the acquisition of APR Energy which occurred in 2020. During the three and six months ended June 30, 2025, the Company received $12,000,000 and $12,000,000, respectively (2024 – $42,500,000 and $42,500,000 respectively) from Fairfax related to these indemnification arrangements. The $300m of Series J Preferred pays 7% interest at the moment. The interest rate is scheduled to go up to 8.5% next year and then increase by 1.5% per year to a maximum of 11.5%. I think Atlas will redeem these in the next couple of years. How bad an investment was APR Energy? Fairfax is still paying out tens of millions per year in indemnification. Luckily 43% of these payments come right back in a sense.

Bought starter positions in LULU and BNTX.

Thanks for the link to the podcast. I learnt a lot listening to it.

I hope this doesn't get me banned from the board, but I sold some of my FRFHF. It's still my biggest position at just over 20% of my portfolio.

This is true if cash returns zero, like it did for most of the past 15 years. But if cash returns 4%, then you have six years to buy at 30% off and come out ahead. That doesn't seem like a terrible bet if you think markets are substantially overvalued.

Absolutely. I think gfp is the gold standard of posters here. It's a little disappointing to see a gratuitous slam against him.

Great post. Some of the posters here seem to have conflated the fiscal deficit and the trade deficit. Dimon's letter, for instance, was referring to the fiscal deficit being unsustainable, not the trade deficit.

Exactly how I'm dealing with my Berkshire position (I have owned it for 25 years now, albeit with some trading around a core position). I am all out of Berkshire in my tax deferred account as of late last year, and this year I have even pruned some from my taxable accounts. I'd sell more if it weren't for the tax hit; Berkshire is at least 20% overvalued in my estimation.

Yes, all of what you say here sounds good and sensible. I don't understand the approach they have actually taken. But I have said my piece on this issue; time to move on to other things.

So I tried Deep Research on this a couple of different ways, and did not get a very good answer. But Deep Research listed five industries where government influence is high - Defense, Healthcare, Infrastructure, Education and R&D - and estimated that around 60% of the job growth in these industries can be attributed to govt spending. From this I infer that 50% or so of total job growth last year can be attributed to the government. But I'm not really debating the need to reduce government spending. The question is whether firing a bunch of government employees so quickly is a good way to do it. I guess you could argue that the mass firing is a necessary prelude to cutting spending, but I'm not convinced; like I pointed out before, govt employment has gone down as a fraction of total employment over the last 20 years. I think this mass firing will lead to unnecessary casualties. Anyway, the point is mostly moot, since the firings have already happened. Now let's wait and see what happens with government spending. It will be very unfortunate if all these people lost their jobs and there is no significant impact on spending.