$AVGO just gave the most bullish semiconductor outlook you’ll hear all year.
Exactly the kind of thing you want to hear for a company like $AEHR.
And it closed deep red yesterday anyway.
Broadcom said XPU shipments went 3.5x year over year and now make up 73% of AI revenue.
mSix customers. Q3 AI at $16.7B, Q4 guided to $21.7B, FY26 at $58B. Then Hock laid out $115B in FY27 and $230B in FY28, with supply secured for both.
Look at the gigawatts behind that.
Google is a multi-year deal for multi-tens of billions in TPUs annually. Anthropic goes 1 GW this year, 5 GW next year, 10 more in 2028. OpenAI's Jalapeño does 1.3 GW in 2027 and 5 GW+ in 2028. Meta has 3 GW of MTIA through 2028.
That's not a slide deck. That's silicon that has to be screened this year and next.
And custom XPUs don't ship raw. These are 600W to 2kW multi-die packages where an early-life failure kills the whole part. That's a burn-in problem, and burn-in is what $AEHR does.
Here's the opportunity.
In April, Aehr got a record $41M production order from its lead hyperscale customer for package-level burn-in of their custom AI ASIC. Deliveries start in fiscal 2027, which began June 27.
In February, that same customer placed an initial order for the next-gen device, which has twice the power per package.
Management said the current device is ramping now, the next-gen moves to production later this year, and the customer is forecasting a substantial expansion of Sonoma purchases.
Overlapping ramps. Both landing in the fiscal year that started ten weeks ago. First print is early October.
$AEHR never names the customer. But Broadcom just put the gigawatt schedule in public. You don't commit 1 GW, then 5, then 10 of custom silicon and wait until 2028 to burn it in.
Wafer-level is the same story. $22M follow-on in August from the lead AI processor customer, nine 300mm wafers per system, shipping to Taiwan over six months.
Separate follow ons in June and August from the lead silicon photonics customer building optical I/O for the same clusters. Aehr is one of very few vendors doing both wafer-level and package-level burn-in for AI processors.
FY26 was the silicon carbide hangover at $50M. Backlog is already $100M+. FY27 is guided to $130 to $150M. Contract manufacturing now supports 20-plus extra Sonoma systems a month. Mix is roughly 70% AI and 15 to 20% silicon photonics.
Stock is down 48% from its $147 high on basically no fundamental change.
The customer who publishes the demand schedule just said it doubles twice. The vendor who tests it sold off.
J'attends que le gap se comble à 72$ et d'avoir un RSI qui diverge bullish pour envoyer une grosse cartouche. Le $WTI a baissé de 8% suite à une annonce bullshit de Trump. Ma théorie : ils veulent combler le gap avant de l'envoyer sur la lune
Je participe au concours #ROGPGW25 pour tenter de gagner un ROG Strix G18 à la Paris Games Week !
Tu peux aussi tenter ta chance avec ce lien 👇 https://t.co/Ez8SQt3dfr #
🇫🇷 FLASH - Marlène #Schiappa va participer à "Tous en cuisine" avec Cyril Lignac ce mercredi sur #M6. La ministre déléguée à la Citoyenneté préparera des œufs moelleux sauce meurette et un pithiviers jambon-fromage. (Le Parisien)
🏟️🇫🇷
Dans #Stade2 retrouvez "Le geste" un nouveau No exceptionnel avec @melrobertmichon L'art du lancer du disque ou comment devenir vice championne olympique à Rio 2016
Reportage signé @philippelafon
What an incredible win! 🔥 Simply an amazing job by @redbullracing and @HondaRacingF1 🙏🏻 Let’s keep this up! Insane to see the Orange Army out in full force, this one was for you 🧡 #AustrianGP 🇦🇹
@DinozzoFlantier pas mieux qu’un athlète pour commenter de l’athlé. J’aimais bcp Bernard Faure et Diagana est bien aussi. Souvent très pertinents dans leurs analyses. Sinon la voix de Montel est indissociable de l’athlé pour moi même si ça fait bcp débat
L'homophobie c'est d'associer systématiquement la sodomie aux homosexuels, une femme peut très bien se faire enculer d'ailleurs je pense que ça te ferait du bien ma chère Marlène