A cynical investor asks some tough questions on shorting and the markets.
The Steyn Capital Daily-Liquidity Retail Hedge Fund has returned 18.2% net of all fees over the last twelve months, echoing the performance of its older sibling, the Steyn Capital Qualified Investor hedge fund, which has compounded at 17.2% net of all fees since launch more than 15 years ago. Key to this performance is not only good stock picking and event driven investments on the long side of the book, but also a large and successful short book. The short book and its individual stock shorts have been crucial to protecting investor capital when markets are down, for example when our Pick n Pay short helped our retail fund to make money in Q1 2024, while the markets were down.
As I have been a successful short seller for the past 23 years, I often get questions from investors about shorting, our short book, and whether a particular stock or a bubble is about to burst. Recently, a short selling sceptic asked me some interesting questions, and I’m attaching an edited transcript of our Q&A below.
Isn’t short selling risky, given the potential for limited gains and unlimited losses?
While it’s mathematically correct that a short can only decline 100%, and in theory it could appreciate to infinity, I would start out by saying that I’ve seen a lot more shorts going to zero than infinity. It brings in the question about risk management. Our shorts are typically one third the size of our longs, in recognition of the risk that shorts running against us would grow larger (this is different to a long, where it becomes a smaller part of your portfolio, and thus a smaller problem, if you are wrong).
We manage our exposure to these target sizes. So, if we have a short with a target size of 2%, and it runs against us by 20% (resulting in a 2.4% short), we would cut the short back to 2%. While this is in essence a form of a stop loss, it is not automatic in nature, but considered in our weekly risk meetings.
In addition, we consistently monitor the cost and availability of stock borrow, as an indicator of a potential short recall, and we monitor the liquidity of our short book closely (we typically don’t short small-caps unless we think they are going to zero. A recent example is Accelerate Property, which we used to call Decelerate Property around the office, having shorted it from R4 to 70c).
Is Nvidia a short?
We never do valuation shorts. If something is crazily valued it can still double and just remain crazily valued. Instead, we look for catalysts: something which will bring the stock back to fair value. I’ve also learned with many years’ experience not to short into a lot of business or stock momentum. Having said that, there are good reasons to be highly cautious of Nvidia’s stock. Bubbles get formed by good ideas morphing into narratives that everything knows to be true, but aren’t. Mark Twain reputedly said “It ain’t what you don’t know which gets you into trouble, it’s what you know for darn sure which just ain’t so”. Well, right now everybody knows that Nvidia’s chips are indispensable for AI development, and that there is insatiable demand for AI development, which is increasingly leading to large scale AI usage. The bulls argue that, compared to the Dotcom bubble, there is real earnings and earnings growth here, and the stock is cheap at 54x when earnings are projected to grow 135% in the year ending January 2025, and 43% in FY 2026. If the bulls are right, this fabulous franchise is only trading at 30x 2026 earnings. I think there are bigger problems for the sustainability of the earnings growth. Firstly, there is the issue of someone building a ‘better mousetrap’. I still have a book on my shelf on competitive advantage, espousing the virtues of Intel, when they all but controlled the chip market two decades ago (spoiler: the prediction of Intel’s continued dominance was fabulously wrong). Last month, start-up chip maker Cerebras purportedly build a competing AI chip that runs 20 times faster than Nvidia’s new Blackwell chip, at a fraction of the cost (although scaling production might not be easy).
The bigger problem might be demand and price point. The demand for Nvidia’s super-fast chips is currently driven by both large AI incumbents and a plethora of venture capital funded AI start-ups who are using the chips to train AI models. Nvidia is the contemporary seller of shovels during the AI gold rush. This absolute rush for chips is leading to super high prices for Nvidia’s chips, driving not only a near 200% growth in revenues, but also a 76% gross margin and a 65% EBITDA margin. I’m not sure how long investors should expect this dynamic to continue.
The Nvidia earnings release watch party – not exactly bottom of the market kinda stuff!
The Financial Times reported quite an interesting dynamic in chip rental pricing last week. Their research has shown that it’s up to 40% cheaper to rent time on a Nvidia A100 processor base in China than in the US. Dear reader, recall that Nvidia AI chips are not allowed to be exported to China, so this price discount flies in the face of what you would expect normal economics to do. Alongside China’s proficiency in skirting export restrictions, the answer likely lies in the decimation of the Chinese venture capital industry, which has seen fundraising fall more than 90% over the last two years. Contrast this to the US, where AI funding has driven venture capital funding up 47% in Q2, to the highest level in two years. Customers who are spending as fast as they can to gain a “first mover advantage” don’t necessarily spend much time negotiating on price. While it’s true that Nvidia’s chips will also be used to run the AI processing once the models are built and sold, the price negotiations during this phase may very well be more rational, which is likely to be much less fun for Nvidia shareholders.
Article by André Steyn CA(SA), CFA, Chief Executive Officer and Chief Investment Officer, Steyn Capital Management.
Source of Financial Information: Steyn Capital Management – 31 August 2024
