Rashomon, Akira Kurosawa's film about one event and multiple conflicting accounts of it, is the clearest comparison for three brokerage notes about one artificial intelligence (AI) trade. The brokerage notes from Goldman Sachs, Jefferies and CLSA, published on October 8 and 9, strike three different tones.
Goldman sees Asian earnings exploding. Jefferies' Chris Wood says the AI spending backing this earnings explosion is "increasingly debt funded rather than cash funded." CLSA's Weekly Note identifies the extremes, from the S&P 500 being top-heavy to the worst long-run Treasury returns in a century.
Goldman: Earnings Justify Rally
Goldman's paper begins with a caution regarding treating Asia as one trade: "Investing in Asia is not a monolithic investment decision." Goldman predicts earnings in Asia to increase by 34% annually through 2028 "after achieving only 6% compound annual growth over the past decade and a half."
Memory chips represent a large portion of that increase. Goldman states that "the memory cycle is apt to last longer and be stronger than it has been previously."
Goldman states that the rally has not increased the valuation of tech. Goldman believes that recent increases in valuations are representative of "underlying profit growth rather than excessive optimism regarding AI."
Goldman concludes that hyperscalers (large cloud providers building AI data centres) will be capable of financing the build-out. The primary risk identified by Goldman is concentration. The top 10 stocks represent a record 41% of the index.
Korea Nearly Doubles, Taiwan Jumps 68%, India Drops 16%: How AI Split Asia in Two
Jefferies' tables demonstrate how unevenly the boom has impacted Asia. As of September 30, 2024, MSCI Korea has risen 95.5% in dollar terms and Taiwan has risen 68.4%. Conversely, MSCI India has dropped 15.6%.
According to Goldman, Taiwan and Korea now collectively comprise half of the Asia-Pacific ex-Japan index. India's weighting has decreased from 19% in July 2024 to 10.3%.
CLSA's chief strategist Alex Redman observes similar divisions within emerging markets: Korea contributes 75% of emerging markets' earnings growth and the remaining 22 markets contribute only 10%. Redman wrote that the current episode "is best described as a powerful tech cycle, not an EM supercycle."
India falls into the other half of that division. Jefferies' tables estimate that India's 2026 earnings will increase approximately 12% on an April-March fiscal-year basis, compared with 58% for Taiwan and over 300% for Korea.
Despite this disparity, Jefferies continues to maintain a slight overweight position on India based on "evidence of a cyclical rebound in domestic demand".
Jefferies: Higher Yields Test AI Build
Chris Wood's Asia Maxima Note states that equities continue to receive support from "stellar earnings growth" resulting from the AI spending boom.
However, Wood notes that the bond market represents a threat to equity markets; the US 10-year yield closed September at 5.28% and the Federal Reserve has resumed raising rates.
Wood wrote that since spending is increasingly debt-funded, "the cycle is most likely to terminate once markets refuse to fund these investments."
Five hyperscalers have sold $229 billion in bonds this year - more than double 2025 sales. Wood predicts that "massive capital destruction in the US" is the long-term "end game".
Even so, Jefferies continues to remain overweight Asia ex-Japan, but underweight Korea and Taiwan. Jefferies states that it lost money due to its underweight positions in major semiconductor stocks. Jefferies states that its absolute return portfolio for Asia ex-Japan is up 13.6% this year, compared with 26% for the benchmark index.
CLSA: Extremes Everywhere, And A Potential Pause In Yields
CLSA's Weekly Bits & Pieces Note identifies "historically extreme readings" on "multiple fronts", many of which are quoted from other analysts and social media posts.
A chart credited to Charlie Bilello of Creative Planning displays Nvidia, Apple and Microsoft as comprising 21% of the S&P 500 - the highest percentage in data dating back to 1980.
Regarding bonds, a Bank of America chart displayed in CLSA's note illustrates the worst long-run Treasury returns in a century.
CLSA's technical analysts Laurence Balanco and Yi Zhou believe that a potential turn could occur. The rise in yields "appears increasingly stretched", they wrote, and there exists "additional evidence for a pause in yields".
Who Pays For AI? Where the Three Notes Differ
Regarding who pays for AI spending is where the sharpest division occurs. Goldman believes that debt funds approximately one-third of hyperscaler spending and that the build-out is financeable. Jefferies believes that the cycle ends when markets stop lending.
CLSA quotes Apollo's Torsten Slok regarding a contradiction in forecasts: analysts covering tech anticipate that tech's cash flow will more than double by 2028; analysts covering customers of tech anticipate less growth. "Both cannot be right at the same time," Slok wrote.
Yields also divide them: Jefferies believes that there is a structural bear market in government bonds; CLSA's technical analysts believe that there will be a pause.
All three depend upon the same spending, which Jefferies states accounted for half of US growth in Q2.
Essential Business Intelligence, Sharp Market Insights, Practical Personal Finance Advice, Daily Fuel, Gold and Silver Prices and Latest Stories — On NDTV Profit.