United States stocks staged a narrow semiconductor-led rally, recovering on dip buying after CoreWeave bond yields, CXMT, and the CPI had squeezed chips last week. On an equal-weighted view only chips rose while the rest fell, with sector rotation repeating and questions over AI funding and payback persisting. CoreWeave CDS and bond yields repriced default risk into a direct hit on AI shares, and Wall Street is now offering one buying reason at a time. Goldman lifted CoreWeave from Neutral to Buy while cutting the target from 139 to 106 dollars, a paradox enabled by how far the price had fallen. Through a rental-shop analogy, secured bank funding runs at 8 to 10 percent unlike 15 percent unsecured paper, and rising usage fees plus residual value lift returns from 15 toward 20 to 25 percent. The CXMT note sees short-term shock as overdone given the advanced DRAM and HBM gap plus qualification and ramp time, calling Samsung and Hynix attractive below a PBR of 1. Despite Treasury interest above 5 percent, a Bank of America survey puts full fund shifts at 10-year yields of 5.9 to 6 percent, yet Nvidia and Broadcom CDS gains say to watch credit warnings too. All directional views are the speakers' conditional judgments and do not confirm any short-term price forecast.

Maekyung Live

A narrow chip rally and repriced AI risk

Neutral to Buy with the target cut from 139 to 106 dollars

Funding rates and usage fees through a rental-shop lens

The CXMT technology gap and below-book memory

Fund-shift rate levels and a credit-market warning

[Hong Jang-won Bull and Bear] Targets cut yet stocks to buy? Rechecking the CoreWeave and CXMT fears

In a SamproTV global interview, Hanwha Investment & Securities senior research fellow Han Sang-hee says fears of an AI investment collapse are excessive. Amid deja vu identical to July (month 7), the S&P 500 sits just 1.5% below its high and the Nasdaq and Dow 4% down, so the US market — 65% of global equities — is not broken. Noting the Nasdaq rose over 1% on a day rates climbed 6bp, he says to read it as a bonds-versus-stocks fight rather than a rates story, with the November (month 11) 3rd midterm split-versus-blue-wave outcome as the variable. The CoreWeave stock and CDS premium spike reflects balance-sheet trouble, not AI demand, and rising rents on old GPUs plus stable asset-backed debt rates make the crisis talk excessive. He rebuts the Gartner report warning of 2030 write-offs of 3 trillion 5 thousand hundred-million dollars and 80% neocloud shutdowns as close to executive-targeted consulting: the 2027 to 2030 outlook for 6 trillion dollars of investment is itself bullish, and CoreWeave's March (month 3) cash-flow model holds up to 16.5% rates. The Anthropic markdown from 2 to 1.5 trillion dollars reflects doubt about the 1 thousand hundred-million-dollar annualized-revenue goal; he walks through the 168 hundred-million-dollar third-quarter (quarter 3) result and 20x PSR math and says quick listings must deliver clear numbers. For big-tech earnings, cloud revenue and operating profit — not FCF — are the checkpoints, and KOSPI 5700 sits at the low end of a 5200 to 7200 box, a place to buy with the 30% cash raised above 7000. Expanding Fed assets and falling bank card delinquencies mean short-term liquidity is not bad.

SamproTV 3PROTV

Is the AI investment collapse fear excessive? From the CoreWeave crisis story to big-tech earnings | Han Sang-hee, Hanwha Investment & Securities senior research fellow [Global Interview]

Amid volatile markets, the broadcast sees no evidence yet to judge AI finished, explaining with a bean-noodle analogy that data-center demand persists. SemiAnalysis forecasts over 50GW added in 2027 alone, a record year against claims that half that, 25GW, would already be huge. 50GW matches all humanity built through 2026, citing satellite photos of 24.3GW under construction across 40 sites and before-after shots of wasteland turned giant factories. The report's proprietary leading indicators confirm 1 trillion dollars in non-cancellable leases as the core basis for the aggressive 2027 to 28 outlook. Citing Factfulness, it argues smooth construction never makes news while problems get amplified. Quarterly megawatt-scale starts surged in Q1 and Q2, led by Hyperscalers 1 and 2, whose identities Q3 earnings will reveal. Construction that took 19 months in 2024 now takes 3 months in 2026 via modular methods, while the non-US share fell from 35 to 28 percent, leaving 70 percent built in America. All directional views are the speakers' conditional judgments and do not confirm any near-term price forecast.

Lee Hyoseok Academy

They said AI data centers failed, yet wasteland became giant factories: the real America in satellite photos

The broadcast centers on Google's 780MW data-center plan in Pine Island, Minnesota, where the AI power race meets an aging grid. With 70 percent of transmission lines past 25 years, it presents the Google-Xcel Energy deal as a model that avoids passing cost hikes to ratepayers. As centers grow from 25 to 50MW toward 1,000MW-class scale, off-grid gas emerged as an alternative, but the Memphis Colossus case raises pollution and reliability questions. It reports 2,600 planned projects across America adding 400GW, alongside Gemini surging from 4 hundred-million to 10 hundred-million users. It explains Google backing 2 times the power it needs with 1,400MW of wind, 200MW of solar and a 300MW battery, plus minimum-demand charges and peak flexibility protecting ratepayers. It covers 1 hundred-million dollars in support pledges against unanimous local opposition, pending commission approval, and the excess-capacity case with criticism of utility profit structures. It relays both the claim that bills rise less than without Google and the risk of hikes if conditions fail. All directional views are the speakers' conditional judgments and do not confirm any near-term price forecast.

CNBC

How the AI race for power is reshaping America's electric grid

The discussion compares the 8-day, over 20 percent consecutive decline with the IMF, financial crisis and COVID episodes in Goldman Sachs long-run statistics, diagnosing last Thursday and Friday as the bottom zone given no domestic internal crisis. It cites KOSPI PBR around 1 and won strength at a 2-year high as bottom evidence, and lays out the conditional path that a halt in AI investment could sink investment, wealth effects and consumption together into a US recession. It flags foreign selling of 20 trillion won and a weak rebound of only hundreds of billions of won after options expiry as a flow burden, stressing position sizes that let you sleep at night plus spare money and a long horizon. It presents Chinese firms' reported US withdrawal as a reflected-benefit case for 2nd batteries while also flagging halved lithium prices and Europe's approach to China as risks. It relays the interpretation that Tesla's flying-car demonstration sets up a SpaceX merger ahead of the December (month 12) shareholder meeting, voicing concern about 1-person decision-making concentration. It sees limited further downside in Korea trading under 5 times versus Taiwan and Philadelphia Semiconductor at 18 to 19 times and China STAR50 at 40 times. It reads the crash in leveraged neo-cloud names like CoreWeave against resilience at Amazon, Alphabet, Meta and Microsoft, with bond markets also favoring strong borrowers, as a selection signal rather than a collapse of all investment. All directional views are the speakers' conditional judgments and do not confirm any near-term price forecast.

CoreWeave and Nebius plunge 25 percent yet Big Tech holds: what happened in the AI cloud market | Myeong Minjun, Seo Sujin, Song Jaegyeong [Beginner Rescue]

Sosumonkey reviews a sea of red: Nasdaq down 5.5%, semis down 11.8%, 8 straight sessions lower. He explains CoreWeave's business and the FT founder interview's insurer and chip-loan trigger. He relays weakest-link fears: 15% funding rates, the 1000 default line breached, and 60% 5-year default odds. He reports Oracle's credit gauge worsening from 200 to 300 plus Nvidia ties and Jensen's soothing. He covers Broadcom and Intel capital warnings, Gartner's 80% wipeout thesis, and Musk's memory attacks. He urges short-term caution on pre-midterm watchfulness with the VIX at 20 and his crisis gauge. He presents below 20 extreme fear and single digits as the contrarian split-buy zone. All directional views are the speaker's conditional judgments, not confirmed short-term price forecasts.

Sosumonkey

Will AI infrastructure's weakest link finally break? Entering extreme fear — another opportunity this time?

In the 10/19 broadcast, the KOSPI opened in fear of a crash as Samsung Electronics and Hynix plunged over 6 percent, then reflected the rebound in United States overnight futures to rally into the close, defending the prior low with a meaningful bullish candle. The speakers argue the market looks too cheap to push lower here, citing a KOSPI PBR of 1.2 to 1.3, ROE of 25 to 30 percent, and a PER in the 6 range excluding Samsung and Hynix. They welcome the return of foreign spot buying of 6,900 and pension buying above 2,000 in hundred-million-won units, but warn against calling a V-shaped rebound as foreigners sold nearly 1 in futures against individual selling above 2 in trillion-won units, with Samsung possibly boxed between 20 and 23 in ten-thousand-won units. On the AI funding gap, they note CoreWeave contracted 4.3 to 4.5 gigawatts against only 1.5 operating, Nebius operating below 1 gigawatt, and fewer than a handful of 50 neo-cloud firms operating 1 gigawatt or more, similar to the Oracle shock episode. Secondary batteries (2) snapped back after Samsung SDI's premarket plunge of nearly 7 percent on news of a collective China exit from the United States via Taiwan DigiTimes, with EcoPro, POSCO group names, lithium plays, and cell makers strong and LG Energy Solution at a short-term high. Strength spread to autos, robots, HVM superalloys, space, bio, solar, and ESS, alongside Tesla damage, hopes for orders shifting to Korea, and the view that a United States-China compromise would hurt Korea. They flag Thursday-dawn Meta earnings and investment plans, the expansion of Muse beyond North America, neo-cloud CDS premia, a memory rebound, Treasury yields, and the midterms as variables, and for tomorrow watch SK Hynix follow-through, defending 5,500, and possibly testing 5,900 to 6,000 on United States strength. All directional views are the speakers' conditional judgments and do not confirm any short-term price forecast.

815 Money Talk

[Real Market Talk] KOSPI rebounds on 7,500 of foreign buying in hundred-million-won units, but it is too early to relax — the signal to check tomorrow | Representative Lee Kwon-hee

A Money Never Sleep New York editor unpacks the Tesla-SpaceX merger talk sparked by the Roadster reveal, centering a Barclays report. With the regular shareholders' meeting due before 12/6/2026, concluding this year requires a formal announcement between October (10) and mid-November (11). Three merger rationales are named: physical-virtual AI vertical integration and governance cleanup for the 1,190 hundred-million-dollar Terra chip project. Most decisive is the control math giving Musk 56% of votes from a 37% stake in the combined firm. Assuming SpaceX at 2 trillion 1 hundred billion dollars and Tesla at 1 trillion 5 hundred billion, an all-stock takeover looks likely. Tesla holders must weigh indirect investment effects against sharing long-term funding burdens. Security reviews, China data issues and shareholder consent stand as barriers, with 20s-of-percent year-end odds and three checkpoints to watch. Every directional outlook is the speaker's conditional judgment and does not settle any near-term price forecast.

Money Never Sleep

Is a Tesla-SpaceX merger near? Barclays on Musk's endgame

Rsanmoo's day strategy for Tuesday, October (month 10) 20th. Yesterday's bounce brought relief, but he sees funds unwinding into an early winter amid a 20% monthly slide shock. The US session was no strong tape despite the Nasdaq's over 1% gain, fading into the cash session and dawn, while Korea's premarket tracked US futures. He likes index levels — buying 5500 toward 7000, a 1500-point 22% run — but draws the line at single names. Despite a Blue House press-room mention of stocks, he rates government or Trump pump odds low, and names the core issue as shaken faith in AI itself over IB outlooks. Rail and drone analogies argue consensus futures never arrive, and he warns of debt and bailout hangovers if AI frustrates. After asides on old-money patience exhaustion and immortality-conspiracy musings, he prescribes a probabilistic response: up or down, buy 2% of total weight near the close, a little each day. An institution would short — meaning hedge — here, but already trimmed with no hedge benefit left, he buys; a 5200 to 5300 break opens the fear zone. He adds buying Samsung Electronics and SK hynix only on plunges, KOSDAQ at 666, trimming some buys on a Friday 5% KOSPI gain, buying a 7% intraday drop, and flash-crash caution.

Rsanmoo

He believes cash must never be fully exhausted. As the video says, this is a phase where asset managers mind hedges more. And no articles have yet appeared saying Nasdaq 100 futures shorts rival longs.

A breather, but no strong tape

Levels say buy the index, pick the names

Low stimulus hopes, shaken AI faith

Consensus futures never arrive

A fan's eyes and old money's patience

Today's play: buy 2% into the close

Single-name tactics and flash-crash watch

Day strategy (1020 Tuesday)

A learning note on the narrow chip rally, Goldman's mixed CoreWeave call, the CXMT technology-gap debate, fund-shift rate levels, and credit warnings.

[Hong Jang-won Bull and Bear] Targets cut yet stocks to buy? Rechecking the CoreWeave and CXMT fears

Timeline

Was this helpful?

Questions & opinions