Backing Huang
Gavin Baker said a petition had been signed by roughly 20,000 people including 2,000 Nvidia employees, and @altcap says the letter to Nvidia's board reminds it not to rush to judgment.
Nov 13, 23:00 - Nov 15, 23:00 Lisbon
Post-Market AlphaJukan relayed a court ruling letting a DRAM antitrust suit proceed and reports of SK hynix HBM4 expansion, Gavin Baker defended exporting less powerful GPUs to China, and Serenity says lasers are sold out for LITE past 2029.
Gavin Baker said a petition had been signed by roughly 20,000 people including 2,000 Nvidia employees, and @altcap says the letter to Nvidia's board reminds it not to rush to judgment.
Per Grok and Claude, Baker says the US-to-China AI FLOPs ratio grew from 2.8x in 2023 to 8.8x in 2026 and is modeled to reach 14.1x in 2027. He argues Chinese labs prefer these American GPUs to domestic accelerators from Huawei and others, which has slowed China's hardware industry.
Both the Biden and Trump administrations allowed the export of less powerful GPUs to China while reserving the most advanced
Baker calls the Biden and Trump policy a clear national security success.
Watch Any change to the policy allowing exports of less powerful GPUs to China.
Open originalA federal court in California found the plaintiffs' allegations plausibly suggest the three companies agreed to restrict DRAM supply, and denied their motion to dismiss.
A federal court in California found that the plaintiffs’ allegations plausibly suggest Samsung, SK hynix, and Micron agreed to restrict
The antitrust lawsuit against all three memory makers can now proceed.
Watch Next steps in the DRAM antitrust lawsuit.
Open originalKorean media report SK hynix is moving ahead despite the Huang scandal, with demand from AMD and Broadcom still strong. Nvidia's lower HBM4 bandwidth target for Vera Rubin eases the manufacturing burden and reportedly informed the decision.
Korean media report that SK hynix plans to further expand its HBM4 production capacity, aiming to add at least 20,000
Demand from other customers remains strong even if Nvidia's AI chip business faces disruption, per the reports.
Watch Confirmation of the HBM4 capacity addition.
Open originalSerenity says the same components, such as lasers, are used across CPO and NPO, and that if NPO comes first, FAUs will still ship through TSMC's COUPE. NPO laser ranges are adjusted down from 400mW to about 150mW-250mW, with lower ASPs than CPO from what Serenity has seen.
So first to address this point:
Serenity says Huang drove and financed supplier capacity expansion ahead of time, such as SPIL's 15x optical engine capacity by H2 2027.
Watch Whether Nvidia's interim CEO keeps investing in supplier capacity expansion and reservations.
Open originalThe H100 started shipping in 2022 and now rents for 58% more than a year ago. @ren_stocks argues capacity locked at last year's prices will be repriced 50% or more higher when those contracts roll off.
$DELL $NBIS $CRWV @MichaelDell said it best. Old GPUs are renting at a premium, and a HUGE premium at that.
They say the bear case for a GPU depreciating in four years is pretty much out of the picture.
Watch Repricing of AI cloud contracts as they roll off.
Open originalNo publications match your search.