Amazon’s AWS acceleration and higher hyperscaler capex keep upstream AI demand alive, but cash conversion, memory demand destruction and RDDT acquisition risk split the tape. Refining constraints—and an unconfirmed weekend Iran strike plan—make energy the immediate binary.
AWS's 37% growth and Amazon/Google's full-stack advantages support real demand, while hyperscalers raised aggregate capex. The disagreement is cash conversion: separate unaudited DD posts challenge Microsoft's true FCF and Amazon's tax-adjusted operating cash flow and liabilities. The cut favors verified bottlenecks, not blanket AI exposure.
KOSPI's unprecedented 18% rebound and the fund postmortem point to leverage—not the AI thesis—as the immediate break. Avi Felman disclosed buying a little MU, while Jonah Van Bourg advised avoiding South Korean equities. That selective split sits beside unaudited evidence of handset demand destruction and rising HBM stacks per accelerator.
The quarter looked strong, yet both X and Reddit focused on Google referrals and AI answers interrupting new-user discovery. A counterview sees usage-based data licensing as a scalable offset, but no purchase was disclosed in that post. The selloff is a dispute about growth-engine longevity, not a denial of current revenue or profit.
Exxon's CEO says refinery capacity—not crude—is now breaking the usual pump-price link, converging with an investor's no-slack refinery map. China Beige Book supplies the counterweight through stockpiles, electrification and weaker consumption. A Reddit-relayed CBS strike plan adds weekend gap risk, but the same report said no final go order existed.
Serenity disclosed a completed add after the earnings drop; the thesis is explicitly weighted to 2027-28 ramps.
Kaizen disclosed a recent BE add and said he was not selling; Filtronic and PL were future intent, not fills.
An unaudited Reddit author said 'I entered today' and set a 100 target; pipeline execution and competition are the stated decision risks.
Labubu's NBIS, GOOG, MSFT, AMZN and ORCL calls are current holdings; IREN is a conditional add with no disclosed fill.
Serenity called them favorites and said he wants to scale further; that is existing exposure plus intent, not a fresh execution in this post.
Current numbers were strong, but referral risk is structural and usage-based licensing remains hypothetical. No retained source disclosed a fresh RDDT purchase.
Jonah Van Bourg said he was out of Meta after a prior bull thesis, while favoring Amazon's execution; treat this as an exit, not a new short.
Avi Felman is selectively buying MU, but Jonah Van Bourg advised avoiding South Korean equities; consumer memory elasticity keeps the regional call unresolved.
The relayed strike plan is market-moving but unverified, and the report itself says no final go order had been issued.