South Korea’s KOSPI hit an all-time high
(+175% YoY), led by semiconductor giants like Samsung and SK hynix, driven by soaring chip exports (+134% YoY) fueled by global AI demand. Retail investors are shifting from crypto to domestic AI and semiconductor stocks, as the “Kimchi Premium” shrinks. Source: Bull Theory
Goldman Sachs Winner and Looser
Goldman Sachs sees AI winners in hardware, cloud, infrastructure, and security (e.g., Nvidia, TSMC, Microsoft, Cloudflare), while traditional software and consulting (e.g., Salesforce, SAP, Accenture) lose as AI commoditizes code and cuts billable hours. The key: compute, power, and security beat software and platforms, reshaping the tech stack. Source: Michael Fritzell (Asian Century Stocks) @MikeFritzell
JPMorgan's Bob Michele compared the move in CLO equity to ABX indices from 2005 through 2007.
Two decades ago, traders dismissed some of those ABX moves as noise. Some are doing the same today with this move in CLO equity. (1/2) Source: Lisa Abramowicz @lisaabramowicz1
This chart is fascinating, not only because of where the number stands today, but also how it has changed over time in these countries.
Which one stands out most? Source: Michel A.Arouet
Silver scores as the number 3 "bubble" according to BofA's Bubble Risk indicator (data per last week).
Source: BofA, TME
OPENAI IS ON TRACK TO BURN $218 BILLION AND THAT SHOULD MAKE PEOPLE PAUSE
OpenAI is projected to spend $218B from 2026–2029—far more than Uber, Netflix, and Tesla combined (~$39B)—while top AI researchers note that more compute now yields marginal gains (Source: Hedgie). Unlike companies with clear revenue paths, OpenAI is losing money on $200/month plans, chasing scale in a field with diminishing returns, driving GPU shortages and higher costs. Meanwhile, AI’s actual contribution to U.S. economic growth was minimal, raising questions about sustainability once investor optimism fades.
Software is deeply oversold (see below $IGV ETF) - Deutsche Bank seems to agree
1. DB argues Anthropic’s Enterprise Agents event reinforces that model providers are more likely to act as orchestration layers on top of incumbent software systems, not replace them, given the deep data, workflows, and metadata embedded in existing platforms (“Claude is only as useful as the data it connects to”). 2. This supports DB’s prior view that AI displacement risk to core software is over embedded in current multiples, making the event incrementally positive for the software sector. 3. Risks remain, including pressure on software development costs, potential changes to the interaction layer that could lower switching costs, and increased competition around the “control plane” for agentic AI, though DB continues to see this dynamic as supportive for infrastructure and compute demand.
P/E multiples tell the story
Software & Services used to rank as the 3rd most expensive industry group, it now sits 9th (and has fallen from 3rd to 13th in Europe). Multiples are down roughly 5.8x globally (around 5x in Europe), a re-rating unmatched by any other industry group. Source: DB, TME
