ChiNext V-shaped recovery on a rather flimsy premise, but sometimes that works out in this market. Local SPEs and semis bouncing hard after 7 straight down sessions. 688012 688120 688256 6809

TENCENT $700: AI catalysts last couple days. "Workbuddy" AI agent combined with new HY3 model launch seeing usage growth & news of ex OpenAI researcher becoming new head of "Hunyuan" AI division. Also, obviously a cheap stock and certain parallels w/ MAG7 rotation.

CAMBRICON 688256: Macquarie initiates BUY & stock flat. US$140bn stock with 1/100th of NVDA's revenue scale even 2 yrs out. Must beat rev estimates by high DD to surprise over time, admittedly quite possible. China's #2# ASIC designer after Huawei: not always a comfy place to be, given how many other #2s# Huawei has completely destroyed or maimed over time, using mafia tactics and regular lawfare. Hygon 688041 another comp, tho not as growthy. Forced to choose b/w Cambricon and Chinese SPE companies as expression of the China semi localization theme, I'd opt for the latter, largely because they don't have to compete w/ Huawei and because memory might be lower hanging fruit than ASICs/GPUs. All that said, I'm small long this doggy.

JAPAN MACHINE TOOL ORDERS: May +38% YoY, moderating slightly from April but still close to ATHs in China (auto-driven) and US (data center driven). Some TPMACH outperformers today, inc. Daifuku 6383

JAPAN STRAT--FACTOR REVERSALS: Nomura analysis concludes that factor shelf life has shortened in last decade and that recent factor performance has largely depended on leadership of a small # of stocks. Looking for reversals in: 5706, 5713, 5801, 6594, 6762, 4004, 6383, 4186, 6506 inter alia

$QCOM: GS stays at HOLD but is more optimistic/trusting of the mgmt vision presented at QCOM's investor day. TBH, it does sound a bit fluffy, but stock will react to future hyperscaler wins.

$QCOM: grudging u/g SELL to HOLD by MS, citing "credible" 27e $5bn AI guidance -> $15bn by 2030. $230 TP hwvr. Questioning the logic of this guidance when co sees no server CPU rev till 28 and is late to the accelerator game. Fake it till u make it.

PANASONIC $6752: day-old news, but interesting report on Panasonic increasing supercapacitor capacity to support BBU production. Panasonic held up well in recent turbulence, steady institutional bid here for vertically integrated new energy player.

CHINA AI GPU TAM ESTIMATE RAISED 37%: MS getting more bullish on local players Cambricon 688256; Hua Hong 1347; AMEC 688012; NAURA 002371 et. al., estimating recent US policy shifts on AI semis and China response will lead to larger 2030 $91bn TAM. Arms races benefit the suppliers more than the combatants (CSPs).

SPE ESTIMATES RAISED AT MS: Raising 26/27 DRAM WFE spending by about 10% each, while not tweaking NAND much at all. Outcome favors AMAT, but maintaining pref for LRCX. Given that the YoY deltas will be much higher for Chinese names CXMT (DRAM) and YMTC (NAND), Chinese local SPEs (benefiting from captive clients) are clearly leveraged to theme, with AMEC 688012 best positioned.

QUANTS--WE ARE IN A KOIZUMI TYPE RALLY: Nikko quants differentiate current Japan rally from dotcom rally, pointing to concomitant strength in Topix (bank-heavy) and momentum (growth names) in current (7/25-present) and Koizumi ('05-'06) rallies. Earnings-driven rallies are characterized by shallower drawdowns whereas multiple-expansion driven rallies have historically collapsed more rapidly and unpredictably.

JAPANESE RETAIL INVESTORS SACK UP: interesting observation from Nomura: Japanese retail investors, notorious for buying dips and selling rips, have changed their behavior this year and are buying on the way up as well as down. Foreign instos, meanwhile have been reducing Japan last 2 months, though I suspect this is about to change with BoJ catalyst and SPCX IPO behind us.

PASSIVES, MLCCs: NIKKO report today. +ve outlook on AI-driven demand cycle. Qualitatively different than the '17/'18 EV cycle or the '13-15 smartphone cycle. "Assuming the current upswing is driven by AI, then the tech upcycle this time is different to past ones that were driven by consumer demand. We think the current cycle is driven by government investment into AI given increased concern over national security. The number of end-products is smaller, but there are more electronic components per end-product, and these components are often high in margin. This means gains in unit prices will contribute more to profit growth than in past cycles, and demand from DCs has been strong enough to offset the slowdown in consumer demand. As a result, OPMs are approaching the peak levels recorded during the IT bubble period." ...However, if you are buying these names here, you probably take the view that consensus estimates are still underestimating the operating leverage in these businesses and we are likely to see big overshoots in the next 2 years.

BMW, A CAUTIONARY TALE: Just a reminder that you don't hate Europe enough. Humiliating guidance cut on China and implications of declining China mkt on EM/Europe competition (Chinese brands dumping inventory). BMW will never come back in China, nor will VW, Benz. Massive strategic failure coming home to roost.

US MACHINERY: +ve outlook on focused equipment spending (ROK, TT, RAL, AYI, VRT, ETN & JCI). "US producers have been able to quickly ramp capacity + output post 2025 Liberation Day because the initial wave of spend has been focused on efficiency + productivity investment, smaller projects w/ quicker payoffs relative to large capacity expansion.... Automation led efficiency projects allow facilities to ramp production in months vs greenfields which take years to come on line - this signals urgency from the customer to drive domestic production & mitigate tariffs, a marked difference when compared to the slow moving & incentive laden "Mega Project" wave in 2021-23. But ultimately, we do believe greenfields will follow - we are already seeing the incremental unit of production shift into the US market in response to changing supply chain costs (ie tariffs), a leading indicator for where larger capacity projects will land when they are ultimately needed."