Manlobbi,

Agree with the future move to the GOOGL board. But also to wait a while for replies to the BRK thread. There are some similarities, but important differences, between how the PE firms may supply capital to AI versus BRK.

I agree that the analogue between using See's to buy BNSF and what Berkshire is doing now is strained, and mostly inaccurate. But the implication that Thompson doesn't understand Google is also a stretch.

I've spent the time to actually read the long pod link in the original post. I'm a novice in understanding what's going on in AI funding. I found the long Thompson link worth the time if this subject is of interest to others. And I intend to reread it, as I'm now doing. I've linked it at the end of this post.

From what I recall so far, Google and Nvidia represent pursuing two different strategies to funding the current round of of the large investments in data centers. In inflation adjusted money, the industry in 2026 total is about equivalent to the 1873 era investment levels in railroads. However, the big 1873 ones were backed by the federal government. Thompson discusses the more risky ones for Northern Pacific, which went broke several times, but ultimately became part of BNSF. Thompson refers more to the smaller bonds for Northern Rail which Cooke pushed to public investors, not institutions. Very risky. The current fear for many firms?

Basically Google is using cash flow plus issuing equity to raise needed capital. It is pursuing a lower cost investment strategy for the next round of data centers, using their cheaper - but less efficient - chips to minimize front end capital. Meanwhile, Nvidia is pursing an alternate strategy using their more efficient chips, assumed to have higher residual value, and partnering with private equity funds for capital. In turn, the PE firms are considering using pension money and life insurance float - patient capital. But at the cost that PE firms are going to want a good return on their money.

But how much of life insurance float and pension fund money will the regulators allow to be used for risky data center investments? New rules for float are going into effect in 2027 that allow the regulators to make their own assessments of risk - and impose them if they find this higher than the ones the PE firms, or their rating agencies, provide.

Manlobbi has shared an excellent post on the Brookfield board about how Brookfield is minimizing their risk. I don't think the Nvidia strategy is going to get any cheap money - i.e. they'll have to pay for the risk.

BRK is also using internal cash flow plus saved dollars and float to back their recent investments in Google. Being BRK, and with Buffett's blessings, they've concluded that the opportunity is worth the risk. BRK is so well capitalized that this doesn't raise any questions with the regulators. Are the PE firms? Their recent scrambles to come up with new capital to cover maybe higher risk mis-investments they made in themselves are causing a lot of stir and speculation. And maybe the sale of the LA Lakers.

So while PE firms and BRK are very different, both use float and earnings to back investments. How much will the NAIC regulators allow PE firms to do?

I'm still early in my learning process. But maybe BRK is taking a reasonable, relatively small, upfront risky investment judged to be a best bet on AI future in terms of risk/reward. Others seem to be making much bigger bets, with Google taking the lower risk approach. And maybe Microsoft watching from the sidelines - but not available to BRK because of past associations with Bill Gates by Buffett.

I hope this thread can run a bit more until we've heard from others about the BRK relationship to AI. I've also ordered the 1873 book to understand the possible analogue with past worldwide economic problems. If the Microsoft CEO refers to it, I want to try to understand why.

And I apologize if I'm not really accurately stating what's going on. Read the Thompson link and make your own judgments. stratechery.com - Nvidias risky business