Publicar
taishiro
The HBM trade is becoming a capacity allocation story, not just an AI demand story.
When memory inventory gets this tight, chipmakers face a choice: prioritize high margin HBM or supply conventional DRAM. That competition can push memory prices higher, but it also makes supply normalization the biggest risk to the trade.
SK Hynix has the advantage today, but Samsung's progress shows that leadership is not guaranteed.
#SamsungQ3ProfitOutlook
ARRIWYKS
#SKhynixQ3HBM4Earnings
#Hynix Cloud +AI HBM4 Cloud = Memory bottleneck now, not GPU
HBM market $38B→$58B +52% 26E, $85B 27E. SK = 62% HBM share, *∼70% of NVDA Vera Rubin HBM4* (TrendForce) vs 50% est. 16 stacks = 576GB/stack vs HBM3E 8Gb → 2x BW.
Why bullish: Inventory <10 days (first ever). Wafers 150-200k/mo 3-4x Micron, 50% → HBM4 = DDR5 squeezed → Q4 DRAM +10-15% NAND +15-20%. JPM sees 20%
*Risks:* Samsung $8.7B HBM Q3 > SK $7.6B first time*
*OKX play: Long HBM leverage, short PC DDR.*
Aviso legal: o conteúdo do OKX Orbit é fornecido apenas para fins informativos. Saber mais
Respostas
Ainda sem comentários. Sê a primeira pessoa a responder!
Novidades do mercado de hoje
1#Rumores de espionagem chinesa envolvendo Jensen Huang e controvérsia sobre a permanência do CEO da Nvidia
2#As vendas de casas usadas nos EUA em outubro atingiram o nível mais baixo em 16 anos
3#Anthropic inicia roadshow de IPO com avaliação a subir para 1,8 biliões de dólares
