Please read the Unison Asset Management Social Media Disclaimer

We recently attended the Berkshire Hathaway annual meeting in Omaha where succession was a key topic given the age of Buffett and the recent passing of Charlie Munger. Ajit Jain, Vice-Chairman of insurance operations, joined Buffett and his appointed successor, Greg Abel, on stage for half of the meeting. Ajit runs the insurance business at Berkshire which is its most important business. The insurance operation not only provides Berkshire with a large float to invest (currently $168 billion), but it has also contributed an average of $1.5 billion in underwriting profit over the last 10 years. As pointed out in the article, Berkshire has produced an underwriting profit in 18 of the last 20 years. This is a remarkable feat for an industry that in aggregate usually loses money in underwriting and makes its return on float. We have been pondering what happens to the crucial insurance operation if Ajit leaves or passes. While still young by Berkshire standards at 72, it is an important question. We believe the key advantages of Berkshire’s insurance division are its risk discipline and its fortress balance sheet. When the insurance market is soft and not providing enough premiums to compensate for risk, Berkshire has tended to significantly cut back underwriting. By contrast, when premiums are adequate as they were for CAT reinsurance in Florida last year, Berkshire will act in a big way (it made a multi-billion profit on its Florida contracts last year). Furthermore, Berkshire tends to favor re-insurance risk more as exposures to risks under these contracts are capped(think about the open-ended liability of being a monoline P&C insurer in Florida when a large hurricane hits for instance). While Ajit will be hard to replace, the culture of risk management is well ingrained at Berkshire, making us more comfortable that results will be good when Ajit (and Warren) are no longer there.

A recent study by The National Association of College and University Business Officers (NACUBO) provides an insightful view of the evolution of university endowment investment policies over the past 50 years. When surveyed initially in 1974, the 136 endowments that participated in the study were categorized into three broad investment approaches: total return, balanced, and income-oriented. Today, nearly every endowment follows what is dubbed the "Swensenian" model, which pays homage to Yale's late endowment chief David Swensen who advocated a heavy emphasis on alternative assets. When Swensen assumed his position at Yale in 1987, US universities allocated on average less than ~2% of their investments to alternatives. Nowadays, over 55% of endowment portfolios are comprised of alternatives, with private equity and hedge funds together accounting for ~23%. Has such a profound change led to better results? Hardly according to Richard Ennis a “doyen of institutional investor performance measurement and analysis” per FT. “Ennis found that the average US university endowment underperformed a blended benchmark in 12 of the past 15 years, for an average annualized relative loss of 0.9%.” It is worth noting that the benchmarks Ennis used are all liquid thus leaving us to ponder about an attending premium for the illiquidity of many of these alternatives. We suspect that the conventional wisdom regarding the new endowment model might be challenged as institutions increasingly look for liquidity to fund obligations in this new interest rate environment.

Yann LeCun, Meta's AI chief, has expressed skepticism about large language models(LLMs) like ChatGPT achieving human-like reasoning and planning. He has openly criticized LLMs for having a limited understanding of logic, lack of persistent memory, and inability to plan hierarchically, arguing that they can only respond accurately to prompts “if trained with the right data”. As such, LeCun is focusing his efforts on developing a new generation of AI systems aimed at achieving "superintelligence," which he believes could take around ten years to realize. Meta has been investing heavily in LLM development to compete with tech giants like OpenAI and Google. Despite this, LeCun's team at the Fundamental AI Research (Fair) lab is pursuing an approach known as "world modeling," aiming to equip AI with common sense and a human-like understanding of the world. This approach involves training AI by mimicking human learning processes, such as “feeding systems with hours of video and deliberately leaving out frames”, with the goal of “getting the AI to predict what will happen next. In a way, resembling how children learn from “passively observing the world around them”. While LeCun's vision is ambitious, it faces skepticism and significant challenges. Critics argue that teaching AI commonsense and causality is highly complex and has not yet proven effective. There has also been some internal tension in Meta, with some insiders attributing the company's late entry into the generative AI space to an overly academic culture within Fair. Despite these challenges, LeCun remains a key adviser to Zuckerberg, emphasizing that achieving artificial general intelligence is a scientific challenge rather than merely a technological or product design issue. Common sense has long been a persistent challenge for AI. As the AI development landscape grows increasingly competitive, the success of LeCun's ambitious vision could potentially give Meta a significant edge. Most importantly, it could redefine the future of AI as a whole.

Blackstone's launch of Blackstone Real Estate Income Trust (BREIT) in 2017 aimed to democratize access to the lucrative world of commercial real estate investment, traditionally reserved for large institutions and wealthy individuals. The promise of an “all-weather strategy” combined with a 4% annual dividend in a low-interest environment fueled BREIT's rapid growth. The fund has generated over $5 billion in fees for Blackstone since inception and today manages about$114 billion in assets (~8% of the firm’s entire fee-earning assets). Since its inception, the fund claims it has delivered an annualized net return of 10.5%—almost double the return of a comparable index of publicly traded REITs. However, recent scrutiny has cast doubt on the fund's strong performance amidst the downturn in the commercial real estate market following the pandemic. Critics argue that BREIT's returns are based on Blackstone's internal and potentially inflated valuations rather than real market conditions. These concerns are amplified by the fact that BREIT does not disclose crucial assumptions behind its NAV calculations to investors or regulators, and because these calculations are both unaudited and unregulated by the SEC. Chilton Capital Management, which invests in public REITs, concluded last April that BREIT “was overstating the value of its NAV by more than 55%”. But Blackstone argues that “it is unfair to compare BREIT” to publicly traded funds, highlighting that the fund’s portfolio is focused on top-performing asset classes such as data centers, logistics, and student housing, with minimal exposure to struggling urban office buildings. In any case, if BREIT's assets are indeed overvalued, investors may have been overpaying hundreds of millions of dollars in management and performance fees. Beyond concerns stemming from inflated valuation, there has been an even more worrisome revelation about the fund’s dividends not being fully sustained by operational cash flows. Not being able to cover a promised dividend can be seen as a Ponzi-like warning, because it means the money has to come from selling off assets, borrowing money, or attracting new investors — which BREIT acknowledges in its disclosures. This reality was put to the test recently in 2022, after a crash in the commercial real estate market led to a flood of redemptions in the fund, forcing the need for Blackstone to raise additional outside capital to meet those requests. And things might get worse. If market recovery continues to lag, it will be harder for BREIT to claim it is the exception. But so far, the BREIT saga has raised an important issue. Private equity firms like Blackstone have only recently begun catering to retail investors, with the fund serving as a litmus test for the entire industry. Will these structures with a mismatch between the liquidity of underlying assets and redemption terms stand the test of time? We are closer to finding out.

Technology journalist and co-founder of media brand The Verge, Nilay Patel, recently hosted Google CEO Sundar Pichai on his podcast Decoder. The central theme of their discussion was a concept Patel has been exploring over the past year- which he refers to as “Google Zero.” In a world where AI increasingly empowers search engines with overviews and summaries, there is a real possibility Google traffic could go to zero for some businesses. But Pichai explains how a similar "Google Zero" argument arose roughly ten years ago when people thought “the web was dead” as the mobile smartphone ecosystem grew. He explains how with every new technology, there is an initial period marked by uncertainty about how to best employ it. But Pichai reiterates that Google’s value-add remains the same–to provide users with a broader context during their searches, with the goal of“showing them something they had not initially thought of”. And pairing AI with search has enhanced that experience so far. One of these features that has received positive feedback has been AI Overviews, which leverages AI to summarize the results of a search for the user. According to Pichai, internal metrics indicate that “content and links within AI Overviews get higher clickthrough rates”. In his view, AI-driven results help the user understand better, which leads to higher interaction levels with the underlying content. This brings us back to the main topic of the conversation between Pichai and Patel. So far, the data has shown that AI is complementary to search, not replacing it. So maybe the future is more accurately described as “Google Enhanced” instead of “Google Zero.”