Whatever anyone thinks about the AI trade, it's the not the table-thumping buy it was, says Citrini Research
The AI trade is evolving.
It's the end of the AI trade as we knew it, says one well-regarded research team whose new report was reflected in Asian stock markets on Thursday.
The idea put forward is it's better to exploit the sticky inflation, higher-yield environment with overlooked real-economy stocks.
Overnight moves in Asian stock markets certainly reflected this shift in approach. AI bellwethers like Korean and Japanese semiconductor plays recoiled as traders rotated into consumer and internet plays listed in Hong Kong and China.
The Kospi KR:180721 benchmark index plummeted 5% while the Hang Seng Index (HSI00) rallied 3.6%. Memory-chip makers like Samsung Electronics (KR:005930) and SK Hynix (KR:000660) tumbled as Tencent (HK:700) and Alibaba (BABA) both spiked more than 6%.
In U.S. premarket trade, Intel (INTC) and Iren (IREN) shares fell 2% as Nvidia (NVDA) lost 1%.
Citrini Research is a well-regarded independent boutique whose thematic and topical reports have made a big splash on several occasions in 2026. With nice timing, its latest output, "State of the Themes," dropped Wednesday, in the immediate aftermath of the midterm elections. It made one bold prediction, of "peak capex growth."
Reducing this argument to simple numbers, the note points out annualized hyperscaler investment is approaching $1 trillion while annual recurring revenue in the whole AI ecosystem sits around one-fifth of that. Clearly something has to give, Citrini observes, and it probably involves a bit of both: reducing the capex and lifting the revenues.
There are limits to how much investment hyperscalers can tap from capital markets. The trend has to be for deceleration
The problem for investors now, according to Citrini, is that many of the bets they are being forced to make in AI are becoming longer-term, gambling on the total addressable market of 2030.
Moreover, hyperscalers are sucking all the available capital out of the market though huge equity and bond offerings. Now, after the Democrats' commanding victory in the midterms, there are additional physical risks to the execution of the AI trade. Data-center construction will, at the very least, become more expensive and complicated so the bottlenecks trade in power and chips has more question marks placed beside it, said the research firm's report.
Citrini is also deterred by the token price wars between OpenAI and Anthropic that began this summer. This confirmed their fears of commoditization and consumer price elasticity, or how sensitive buyers of products are to a change in the cost. At the same time, Citrini's analysis finds that AI is now starting to present headwinds through competition and disintermediation that is undermining the prospects of the software-as-a-service sector.
Price war between OpenAI and Anthropic confirmed that the frontier models are now commoditized
So, how should investors adjust their portfolios then to reflect these new realities? Citrini's flagship Citrindex model portfolio has generated a total return of 257% since May 31, 2023, more than 2.5x times that of the S&P 500 SPX. The selection is now being adapted in favor of reversals, pair trades, lower beta plays and the overlooked real-economy sectors like "Boomer software," financials, entertainment and energy.
The team thinks it's time to look for relative value. They suggest pair trades like a long position in Nvidia (NVDA) against a short in Advanced Micro Devices (AMD), Oracle (ORCL) paired against Nebius (NBIS), Samsung against SK Hynix. and Meta (META) versus Alphabet (GOOG).
Given the relative strength of the economy, better wages and sticky inflation, Citrini is looking at consumer plays like Victoria's Secret $(VSXY)$ and e.l.f. Cosmetics (ELF). Entertainment is represented by media stocks like IMAX (IMAX) and Cinemark $(CNK)$ and live events by companies like MGM Resorts (MGM).
-Jules Rimmer
Comments