Dhaval Joshi, chief strategist of the well-known British research company BCA Research, released a bold and interesting research report this week, some of which coincides with many of the arguments in our previous article 'Is there a connection between the surge of the yen and the plunge of technology stocks?' The report also analyzed in depth the correlation between the closing of yen arbitrage trades and the turmoil in technology stocks in recent times.

Here is an abstract of the report, let us have a quick look:

Joshi said that recently, the market has been turbulent, and many market commentators quickly pointed the finger at two culprits: the closing of so-called 'yen carry trades'; and the initial retreat of the 'artificial intelligence bubble'. However, these two apparent culprits are actually one - yen carry trading and the artificial intelligence bubble have become the same type of trading, and the convincing evidence is that their price trends are identical.

Correlation does not equal causality, but...

In mathematics, complexity quantifies information. The more complex the model or time series, the more information it contains. In financial markets, price complexity quantifies the information provided by different types of investors, who are defined by their different time spans.

Simply put, long-term investors provide valuation information, while short-term investors provide momentum information. The specific price complexity of any investment is the result of the combination of long-term and short-term investors in that investment at a particular moment in time.

A most vivid metaphor is that the complexity of investment prices is like fingerprints. Therefore, when the 'fingerprints' of two seemingly completely different trades come together, it means that the investors driving these two trades have become a synchronized group of investors. Trades that seem different are actually the same trading type.

Like fingerprints, the complexity of price for different investment products is also drastically different, and yen and AI stocks are usually such cases. However, in the past few years, their price complexities have merged, especially at turning points of complexity collapse.

Correlation does not necessarily mean causality. However, we can say that when investors sell yen, they will buy AI stocks. But they do not buy AI stocks for current profits. At the same time, they also raised the valuation of artificial intelligence stocks relative to bonds (as shown below).

Joshi believes that although those who sell yen may not be the same people who buy AI stocks, selling yen and 'inflating' the artificial intelligence bubble are actually two ends of the same trading process. The sale of yen (borrowing yen for carry trades) has led to the expansion of the artificial intelligence bubble, while the artificial intelligence bubble has promoted the sale of yen. Therefore, this relationship is more of a bidirectional, reflexive relationship than a causal one.

The combination of yen carry trades and the artificial intelligence bubble is significant, because all bubbles require leverage. In this round of artificial intelligence bubble, leverage has been coming from yen finance (carry trades).

"Three-legged stool"

Joshi pointed out that yen carry trades are like a three-legged stool, and now each leg is bending...

Carry trades involve borrowing currencies with low (preferably zero) interest rates and investing in high-return investments. This means that carry trades are like a three-legged stool, and each leg must be stable. Specifically, which three legs?

The interest rate of the financing currency cannot rise;

The exchange rate of the financing currency cannot appreciate;

The high-return investment must continue to bring high and stable returns, because if the financing cannot be used for safe profitable purposes, then financing is meaningless.

If any one of the three legs is broken, the stool will collapse. Unfortunately, Joshi believes that within a cycle (6-12 months), all three legs of yen carry trades are easily broken:

Firstly, the Japanese interest rate faces the risk of rising. Given the recent instability of the market, the Deputy Governor of the Bank of Japan downplayed the possibility of further tightening earlier this month. However, the underlying meaning is that once the market returns to calm, the prospect of further interest rate hikes will be reignited, which will once again disrupt the stability.

Moreover, even if the Bank of Japan does not further raise interest rates, the yen is likely to appreciate if other central banks cut interest rates.

Thirdly, there is always speculation about how quickly the gold rush for artificial intelligence can "dig out gold" and who might shatter it.

The winners of Web 2.0 are unlikely to be the winners of the AI era.

It should be clear that generative AI can indeed improve productivity, like all previous general-purpose technologies, such as PCs and the internet. However, there are two major misconceptions in this regard.

The improvement in productivity due to AI will not be greater or faster than that of previous general-purpose technologies.

Each generation of general-purpose technology wave has a productivity S-curve, and the overall economic productivity is determined by the sum of these contiguous S-curves. However, the steep part of the newest general-purpose technology S-curve often overlaps with the flat part of the previous general-purpose technology S-curve. The result is that the overall economic productivity will eventually rise at a stable and roughly constant rate.

In addition, companies that became big winners in the previous generation of general-purpose technology waves are rarely the big winners in the next wave of general-purpose technology waves.

For example, the superstars of technology stocks in early 2000s, such as Microsoft, Intel, IBM, and Oracle, were big winners of the previous generation of general-purpose technology (PC). At the time, Cisco was the company that made "picks and shovels" for the next technology gold rush, the internet.

Today, the superstars of technology stocks and quasi-technology stocks, such as Microsoft, Apple, Amazon, Meta, Google, and Broadcom, have become the big winners of the previous generation of general-purpose technology, Web 2.0. In addition, there is Nvidia, which produces "picks and shovels" for the current technology gold rush, AI.

Joshi believes that Nvidia is unlikely to be a long-term winner among today's super AI stars, because the demand for "picks and shovels" occurs in the early stages of the gold rush and then tends to rapidly wane. At most, one or two companies among other super AI stars will transform into big winners of the AI era.

This means that the high total valuation of super AI star stocks will eventually become precarious, especially in the case of previous yen carry trades that helped inflate valuations.

Conclusion

In summary, Joshi pointed out that none of the three legs of yen carry trades are likely to remain stable within the cyclical range (6-12 months). Once one of the legs collapses, the entire chair will collapse.

The conclusion that Joshi draws is:

Reduce holdings of US superstar stocks; underweight Japanese rates (assets); overweight yen.

But it should be noted that these three trades are actually a huge correlated trade.*

Finally, in terms of investment advice, Joshi offers a somewhat surprising option: tactically overweight Global X MSCI China Consumer Discretionary ETF (CHIQ).

Joshi says that Chinese stocks have been lukewarm in the past few years. Nevertheless, they have provided excellent tactical opportunities for long and short positions, from short-term upward and downward trends to the collapse of complexity in the price of the S&P 500 index for 65 days, which has provided a good tactical entry point for directly long or increasing holdings of Chinese stocks.

Joshi pointed out that we tend to be overweight CHIQ relative to the S&P 500 index, and set profit targets and symmetric stop losses at 10%.

Editor/Lambor