Daily Alpha · X
· Premarket Alpha · by Buzzberg Research
The X crowd split on the same facts: one camp read the Korean memory crash and Anthropic's print as proof the AI trade is cracking, the other argued the selling is macro-driven and contract-backed suppliers are being conflated with leveraged neoclouds.
Themes on this desk
Korea memory unwind
Samsung and SK Hynix down 9–11% and a KOSPI circuit breaker; shorts pressed Hynix as the weakest memory name.
TSMC results
Q3 revenue above guidance, EPS up 68.7%, Q4 guide above consensus and capex maintained; management sees demand outstripping supply.
AI credit and neocloud differentiation
CoreWeave down 15–20% on financing fears while analysts argue prepayment-funded builders and contracted optical suppliers deserve separation from levered neoclouds.
Ticker heat
?
Top voices by smart followers and alpha score
Market Radar →Jukan: Anthropic's ARR 'miss' is demand shifting to OpenAI, not demand disappearing
The post argues the market is misreading Anthropic's lower-than-expected ARR as slowing demand when customers moved to OpenAI's Codex after the Astra launch; expectations were simply too high. He contends the relevant metric is combined OpenAI and Anthropic ARR — and that the trade was always about how much the two labs spend, not their own profits.
A good chunk of the market seems to want to read Anthropic's lower than expected ARR as a sign that
If demand is shifting rather than shrinking, the negative semiconductor read-through from the Anthropic print is a positioning event, not a fundamental one.
Watch OpenAI's next revenue disclosure and whether combined frontier-lab spend keeps rising into the IPO window.
Source →Serenity (@aleabitoreddit): the selloff is macro-driven — separate hyperscaler-funded optical demand from financing-hungry neoclouds
The post attributes the drop to CPI and rate-hike anticipation, notes the difficulty of mapping knock-on effects if CoreWeave or frontier labs stumble, but argues indiscriminate selling creates opportunities. It stresses separation: LITE is backed by take-or-pay contracts sold out to 2029 and hyperscaler financing from Microsoft and Amazon, unlike a neocloud needing billions in future financing at high rates.
Clearly a macro related drop after CPI numbers came out today and further US rate hike anticipation.
A high-hit-rate supply-chain voice arguing the market is over-punishing contract-backed AI suppliers relative to leveraged compute landlords.
Watch Whether contracted optical names reclaim pre-selloff levels while neocloud credit spreads keep widening.
Source →Ren's take: Gartner says 80% of neoclouds won't stand alone by 2030 — and Nebius is built to survive
The post agrees with Gartner's forecast that 80% of neocloud providers will stop operating independently by 2030, arguing the capex burden means most get acquired, merged or starved of capital. It says Nebius is different because customer prepayments covered 50–60% of Q2 deal capex and gross margins are about 77%, versus peers borrowing against GPUs.
Gartner says 80% of neocloud providers will stop operating independently by 2030 and I agree.
Reframes AI-credit risk as a consolidation question — fewer providers with a bigger share means survivors could capture outsized economics.
Watch Nebius's Q3 disclosure on prepayment-funded capex and any neocloud acquisition or restructuring.
Source →Bubbleboi backs Musk's view that China will make 'a LOT of RAM' by 2028
Replying to Elon Musk's claim that China will be making a lot of RAM in 2028, the widely followed AI-hardware commentator said simply, 'It's true,' endorsing the forward-supply argument feeding the memory de-rating.
It’s true > Elon Musk (@elonmusk) > China will be making a LOT of RAM in 2028
A respected hardware voice validating the longer-run supply threat caps how much memory cycle the market will pay for.
Watch Chinese DRAM capacity announcements and whether 2028 supply materializes into pricing pressure.
Source →TSMC guides Q4 sales to $52.3–54.3B, beating consensus, and keeps 2026 capex at $60–64B
The post relays TSMC's Q4 forecast of $52.3–54.3bn sales versus a $50.35bn consensus, gross margin of 65–67% and operating margin of 56.5–58.5%, alongside Q3 results including NT$763.1bn net income and unchanged 2026 capex of $60–64bn.
$TSM (Bloomberg) -- TSMC forecast sales for the fourth quarter; the guidance beat the average analyst estimate.
The guidance beat is the cleanest single datapoint on AI demand durability and keeps the buildout funded through 2026.
Watch Whether the guidance beat moves TSMC shares and any 2027 capex quantification on the call.
Source →