The AI of the Beholder

Talk of whether we are or are not in an equity market bubble feels like a waste of energy. After all, we cannot know for sure, and even if we could, we don't know if we're at the 1996 or 1999 stage. But there is something about extreme valuations that can give markets a kind of weightlessness, where the fundamental rationale for prices becomes almost binary: it depends on one's beliefs about the efficacy and impact of AI.

This can lead to a sense of randomness to security pricing, as shown by the large swings in the post-earnings share price for Meta, Microsoft, Alphabet and others towards the end of the month, and by Nvidia adding US$2.5 trillion to its market cap in six months to become the first company worth more than US$5 trillion.

Of course, there is more going on in the world than AI, but AI feels like the right thread to pull. The investment in large language models (LLMs), growth in datacentres, and development of other AI-tangential services are the driving force behind much of the expansion of credit lending, both public and private.

In public markets, AI-linked companies now account for nearly 20% of the global convertible bond market, up from less than 4% at the start of 2023. In private markets, lending is supporting much of the build-out of services across the AI ecosystem, since many of these companies are yet to turn a profit.

Last month, JP Morgan’s Jamie Dimon made headlines talking of potential 'cockroaches' in private credit markets. The implication is that one visible problem suggests many more are hidden, not, as sometimes misreported, a comment on the market's unsanitary nature. Either way, the blend of hidden lax lending standards combined with a currently loss-making industry is potentially combustible.

One can also bring many of the geopolitical and macro themes back to the AI narrative. The much-celebrated trade agreement between the US and China (the precise details of which remain unclear at the time of writing) had a key focus on US access to rare earth metals from China, which are critical to the continued build out of the AI infrastructure. And taking a further step back, the success (or not) of the AI revolution has meaningful implications for GDP growth, productivity, and inflation.

My learned colleague Henry Neville has frequently referenced Amara’s Law with respect to AI: that we overestimate the impact of a technology over the short term and underestimate it in the long term. In the maelstrom of uncertainty around AI, this feels like a modicum of solid ground.

The overestimation in the short-term is centred on real-world resource and time constraints. How do the huge number of promised datacentres get built? How do we generate enough electricity to power them? How does every customer get to use the most powerful AI? How do we build enough robots to make a dent in the quantum of human manual labour? Furthermore, there is a big difference between ‘is it possible?’ and ‘is it profitable?’, particularly at the early stages of new technology. There is a reason we last sent people to the moon in 1972.

But the underestimation of the long-term impacts also feels true. It is a tired but real observation that the bursting of the Dot-Com bubble didn’t kill the internet. Similarly, if we are currently in an AI bubble then it doesn’t negate the long-term benefits. Incrementalism is key here. Over time, many of the resource constraints listed above may be resolved, albeit slowly. Furthermore, anyone harbouring concerns that AI will never make the leap to Artificial General Intelligence (AGI), or that LLMs will never make the leap to large reasoning models, would do well to read the recent pre-print publication by Hendrycks et al from the Center for AI Safety.¹

This paper suggests that GPT-4 was 27% of the way to ‘matching the cognitive versatility and proficiency of the well-educated adult’, and that GPT-5 is 57% of the way there. Solving the remaining 43% of this more pragmatic definition of AGI seems just a matter of time.

For hedge fund managers, trading in this environment is a double-edged sword. Volatility typically begets alpha, but this level of uncertainty can mean that investors may be reticent to take as much risk as previously, and deleveraging typically begets losses. Hedge fund industry performance has broadly stayed strong over the last few months, but some concerns are beginning to surface. There have been notable headlines about outsized losses in some quantitative equity-market-neutral strategies in September and October.

Two pieces of evidence suggest that the situation is potentially more fragile: firstly, the Goldman Sachs Most Shorted Index (an equally weighted index of the 50 most shorted stocks in the Russell 3000) has seen a step-change in volatility since mid-September, rallying 34% from 10 September to 15 October. Secondly, for the first three weeks of October, market-neutral factor measures of US stocks were negative for all of Value, Quality, Momentum and Low Beta – the four typical risk factor building blocks of fundamental market-neutral strategies. However, both metrics have calmed during the second half of October, and there appears to be little sign of contagion to other active strategies, or to broader measures of market liquidity.

So, regardless of your views on AI, the consensus move from here appears to be for equities to go higher. Markets are pricing five more rate cuts from the Federal Reserve before the end of 2026 to go with the October one, despite no credible forecast of inflation showing any hope of reaching the 2% target over the next few quarters. Such loose monetary policy risks amplifying the bubble chat for some time yet.

Key drivers of hedge funds’ performance: an early October snapshot

Equity Long/Short (ELS):

• Equity markets saw continued strength in October, but what is concerning is that the narrowness in market leadership has seemingly spread outside the US and into international markets. For instance, Japanese and Korean equities (as represented by the Nikkei and KOSPI) posted double-digit monthly returns; however, the gains have been concentrated in a handful of stocks. Similarly, the S&P equal-weighted index was down on the month while the S&P 500 was positive
• Given policy and geopolitical uncertainty is still having a large influence on market sentiment, positioning at both the sector and factor level has continued to play a major role in ELS fund performance. Crowded long positions have underperformed while crowded short positions have outperformed. There have been significant sector rotation dynamics rooted more in themes and optimism rather than fundamentals. All of these factors serve as headwinds to long/short fund performance
• That said, index and prime brokerage data indicate modest performance for ELS funds across regions, albeit with considerable dispersion. Larger and longer-biased funds have outperformed due to beta and size dynamics. Market-neutral funds have underperformed due to deteriorating short alpha and the underperformance of quality-linked factors
• From a risk standpoint, prime brokers have not reported any meaningful changes in gross- or net-leverage though funds have continued to steadily cover shorts

Credit:

• Despite concerns over lending practices at US regional banks, and two higher-profile default situations (Tricolor and First Brands), credit spreads were largely unchanged during October due to the broad risk-on move across markets
• Convertible bond markets continue to perform well, supported by the large levels of issuance of AI-linked names. Outright convertible bonds are one of the best performing asset classes globally year-to-date, buoyed by retail involvement in the names
• In cross-border deals, Switzerland’s Novartis announced plans to buy Avidity in the US for US$12 billion
• For hedge fund managers, the high level of issuance and volatile nature of these convertible bonds has been a significant source of alpha over the last two months
• Elsewhere in credit markets, managers are generally taking a more cautious approach, waiting for wider spreads and more default activity that would provide opportunities to deploy capital

Event Driven:

• It was a mixed month for event driven strategies. Merger arbitrage strategies were generally positive, but broader event funds saw more dispersion
• Deal volumes continued their positive year-to-date trend, with several interesting US transactions announced in October, including the Blackstone and TPG US$16 billion bid for Hologic; banking mergers between Comerica and Fifth Third Bancorp (US$10 billion) and Cadence Bank and Huntington Bancshares (US$7 billion), Essential Utilities merger with American Water (US$11 billion); and Qorvo and Skyworks US$9 billion merger in the semiconductor space
• In cross-border deals, Switzerland’s Novartis announced plans to buy Avidity in the US for US$12 billion
• Two prominent mergers failed to materialise in October. The first was somewhat expected as Core Scientific shareholders voted down an all-stock acquisition offer by CoreWeave, in line with recommendations from some activist holders. The second was more surprising, with Spanish bank BBVA’s offer for rival Banca Sabadell only receiving 25% backing from shareholders

Quantitative strategies:

• Quantitative equity-market-neutral strategies appear to have had a mixed month in October. As ever, given the secretive nature of these managers, the true picture only becomes apparent later. From early signs, it appears that more fundamental models struggled in October, in contrast to the more technical difficulties seen in September
• Available manager performance in Statistical Arbitrage suggests a wide range of returns, without clear commonality between winners and losers. There were some media reports of significant losses mid-month, although some of the worst-performing factor families (such as short interest) recovered significantly in the second half of the month

Macro strategies:

• Trend-following strategies had a broadly positive month, with strong returns in the first half tapering toward month-end. The best performers were long positions in equity indices and precious metals, although the latter saw a significant reversal in the second half of the month
• Discretionary macro managers continue to see a strong opportunity set in emerging market, rates and FX, particularly around the tightening cycle in developed markets rates and its global implications

On the radar:

• Over the short term, we are watching for any continuation of single stock volatility around the AI theme. The early announcements of both Amazon’s US$38 billion deal with OpenAI and Microsoft’s US$9.7 billion deal with IREN in the first few hours of trading in November suggests that there is plenty more deal-making to come
• With the holiday period less than two months’ away, we are also watching for the behaviour of intra-stock dynamics during this period of typically lower market volumes. Losses from crowded positions and forced derisking could lead to renewed issues in equity-market-neutral portfolios
• Over the longer term, the focus remains on the impact of easing monetary policy across developed markets, despite few signs that inflation targets have been achieved, and the risk that excessive stimulus could trigger further inflationary pressure

1. arXiv:2510.18212 ‘A definition of AGI’