“In the short run, the market is a voting machine but in the long run, it is a weighing machine.”
This is a quote from Warren Buffett in 1987 who paraphrased a passage from Graham and Dodd’s seminal book Security Analysis, first published in 1934.
Currently, the “voting machine” is having a disproportionately large impact on AI shares. Stock returns are being driven by investor sentiment rather than fundamentals. And return volatility is extreme, even for mega-cap stocks such as Nvidia.
This type of market environment can compromise the predictive power of factors with slow decay rates, particularly value factors. Given this, we are focussing more on our screens and factors which exploit short-term mispricing opportunities.
We’re currently focussing on AI as a risk factor rather than an alpha source. We’ve reduced our gross exposure to these stocks and are closely monitoring our net exposure. More details on why we believe AI is currently a risk factor for our style of investing are provided in the next section.
More broadly, it seems the marginal price setter for shares is more likely to be a retail punter than an actively managed mutual fund. Early this year, David Einhorn lamented “the value industry has gotten completely annihilated” and “the market is fundamentally broken” (Business Insider, February 2024). Putting aside the hyperbole, we believe that buy-and-hold value investing is structurally challenged and largely subscribe to this view.
We’ve adapted our investment process to suit the market environment. We’ve added new short-term screens to our rebalance process and have greatly increased the number of daily trades. We’re now doing well over 200 trades per day. These trades are typically small and are focussed on finessing position sizes to exploit short-term liquidity flows.
In the long term, the “weighing machine” will determine share prices. The long term, however, isn’t compatible with our investment time horizon. We endeavour to provide consistent returns, regardless of the market environment, rather than offering the vague promise of being right eventually. It’s the reason why the fund performed strongly in 2021 when the market was focussed on growth opportunities and in 2022 when value dominated market sentiment. We’re also pleased the fund generated consistent returns in 2023 and has registered a positive return every month in 2024.
AI as a Risk Factor
Risk factors explain cross-sectional returns but are not a source of alpha in our investment process.
Some risk factors are always strongly correlated with stock returns. Examples include market, sector and size exposure.
The importance of other risk factors varies over time. A good example is Yen exposure. It’s always a risk factor but its particularly important currently given recent Yen depreciation, government intervention to support the currency, and the likelihood the BOJ will tighten monetary policy.
Other risk factors are specific to the current market environment and aren’t captured by risk models calibrated using historical data. Covid is the obvious example given it was the first global pandemic in 100 years.
We believe there is a new risk factor which is driving investor sentiment but isn’t adequately captured in risk models: AI exposure. Many stocks which are beneficiaries of the AI revolution have outperformed. This list is large and extends far beyond NVDIA, TSMC, Micron, SK Hynix and other semiconductor stocks. For example, some REITs with data centre exposure have strongly outperformed YTD.
It’s not as manic as the Internet boom in the late 1990s, but some similarities are emerging. In late March 2000, Cisco became the most valuable company in the world, with a market capitalisation of more than $500 billion. In June, Nvidia hit this milestone with a market capitalisation of more than $3.3 trillion.
Much like during the Internet boom, the US is at the centre of the frenzy, attracting the strongest liquidity flows and the most extreme valuations. TSMC’s ADR currently trades at a 16.4% premium to its primary listing in Taiwan.
Other similarities include the outperformance of the Nasdaq Composite and the extreme level of global stock market concentration.
Another warning sign is that across Asian equity markets, liquidity flows rather than fundamentals are driving extreme share price divergencies. In Australia, NEXTDC and Goodman Group (whose growth outlook is contingent on data centre demand) are far more expensive than stocks listed in Singapore with significant data centre exposure (KDCREIT and Mapletree Industrial Trust). The contrast with Singapore Telecom is even more stark given its increasing data centre exposure and extremely cheap SOTP valuation. We believe this is because Australian investors tend to focus solely on Australian equities whereas Singaporean investors are more likely to have a regional focus, exposing them to more AI investment opportunities, particularly in Taiwan and Japan.
It’s hard to know whether the recent AI share price gains are a bubble which will pop, slowly deflate or prove to be resilient. What we do know is AI exposure is now a material risk factor which can’t be mitigated by minimising net sector exposures. Even within Semiconductors & Semiconductor Equipment the level of AI exposure varies widely.
To deal with this, we’re manually reviewing the AI exposure in our long and short portfolios to ensure we’re not materially exposed if there is a sudden reversal in investor sentiment. This is a big advantage we have over systematic quant funds which analyse risk based on generic risk models that don’t deal effectively with what is currently a significant driver of investor sentiment.
Comments