Daily Alpha · X
· Premarket Alpha · by Buzzberg Research
Serenity (@aleabitoreddit) relayed Coherent's call that 2027 is sold out, Meritz via @jukan05 sees a tighter memory market in 2027, and Daniel Koss argued against Citrini's short-Nebius, long-Oracle pair.
Themes on this desk
Data-center ballots
@jasonschips counts data-center opponents winning 40 of 41 local ballot measures across 10 states; Serenity (@aleabitoreddit) argues the new measures affect only a tiny share of planned compute buildout.
Memory pricing
Team Group's CEO sees the memory shortage worsening from end-2026 to end-2027 with spot-contract spreads of 30-50%, and @PhotonCap reports record Q3 DRAM revenue of $200B, up 369% year on year.
Ticker heat
?
Top voices by smart followers and alpha score
Market Radar →Citrini says it cut AI hardware from more than 50% to under 10%
Citrini says the firm reduced AI hardware from more than 50% to less than 10%, with Nvidia the majority of that, and that NVDA has massively outperformed SMH since it was overweighted.
@alwayslearni 1) Like I said, we reduced AI hardware from >50% to less than 10%.
A closely followed thematic investor has sharply cut AI hardware exposure.
Source →Meritz sees the memory order fulfillment rate falling to about 80% in 2027
Meritz says 2027-2028 memory orders far exceed supply capacity; the fulfillment rate has stayed in the low-to-mid 80% range this year and should fall to around 80% in 2027. Sam Altman asked SK hynix and Samsung for HBM for 2027-2030 in volumes beyond what they can accommodate.
Meritz Securities (South Korea): Genuine demand reflected in orders received by memory chipmakers for 2027 and 2028 far exceeds their
Meritz says Democratic-sweep delays are unlikely to change the supply shortage materially.
Watch The 2027 memory order fulfillment rate.
Source →Serenity's Coherent call notes: 2027 sold out and no laser oversupply risk by 2028
Serenity's (@aleabitoreddit) summary says Coherent sees no oversupply risk by 2028, its internal InP production is consumed by its own products, customer LTAs carry prepayments used for capacity, 2027 is completely sold out and LTAs extend to 2030 with customers placing orders into 2028.
$COHR earnings transcript TLDR: Q: On $LITE + $MRVL merger. A.
Customer prepayments answer financing concerns about Coherent's capacity expansion.
Watch Customer orders placed into 2028.
Source →Daniel Koss argues the opposite of Citrini's short-Nebius, long-Oracle pair
Koss cites Oracle's BBB- rating, about $118B of long-term debt and OpenAI as its largest customer, against Nebius's backlog on investment-grade Microsoft and Meta contracts. He notes Oracle's Stargate footprint is heavily in Texas while Nebius has no disclosed Texas sites.
Short $NBIS, long $ORCL. This Citrini trade seems absolutely INSANE to me. The exact OPPOSITE trade seems much smarter.
Texas moratorium risk falls on the long-Oracle leg, in Koss's view.
Watch Delivery of Nebius's Microsoft and Meta tranches and Oracle's OpenAI site ramps.
Source →AT&S posted record Q2 revenue of €682M, up 52% year on year
AT&S Q2 revenue was €682M, up 52% YoY, with a 35.8% EBITDA margin versus 23.3% a year ago and EPS of €3.05 versus €2.08 expected. The full-year guide of 45-55% revenue growth was held, and customers added €421M of prepayments in six months.
The AI trade is NOT dead. $ATS.VI ER just dropped, they posted a record quarter. > Q2 revenue €682M.
AI substrate pricing is now showing up in the numbers, per the post.
Watch H2 needs €683M to €773M a quarter to meet the guide.
Source →