#Micron10BAIResearch

Micron’s planned investment of about $10B over 10 years in a Boise research lab is more than a capacity story. Its focus on next-generation memory, Memory+Compute, advanced packaging and future chipmaking suggests the competitive edge in AI infrastructure may increasingly sit in integration, not simply in selling more memory. My measured read: the strategy could support steadier demand and stronger margins if research converts into differentiated HBM, data-center memory and AI storage. But the sequencing matters; heavy spending before revenue catches up would leave cash flow, and storage valuations, exposed. Not advice, just analysis. #Micron10BAIResearch

Marvell’s Google agreement matters less as a one-day share-price catalyst than as evidence that custom AI infrastructure is broadening beyond compute alone. The scope spans TPUs and related accelerators, storage controllers, networking and memory interfaces, giving Marvell several potential routes into Google’s buildout. The warrants, covering up to 58.97M shares at $206.58, are not a completed equity purchase. Their real analyti#BTCBreaks72K #FOMC9To3Split #PopMartEarningsWatch

#StorageValuationSplit Investors remain divided over how to value Sandisk and other storage companies after extraordinary gains driven by AI demand. Sandisk recently reported $8.97 billion in quarterly revenue and adjusted EPS of $39.25. Data-center storage demand, higher NAND prices and long-term customer agreements support the bullish case. However, skeptics argue that current earnings reflect unusually favorable memory-market conditions. A low price-to-earnings ratio does not automatically mean a cyclical company is cheap. If NAND pricing is close to a peak, future profits could decline even while current results look impressive. Bulls believe AI data lakes and enterprise SSD demand have permanently improved the industry’s economics; bears expect new capacity and normal competition to pressure margins. The next evidence will come from contract pricing, supply additions, data-center revenue and customer commitments. Investors should value the company using several cycle scenarios instead of assuming either today’s profits or historical downturns will continue indefinitely.

CXMT’s breakout listing has sparked optimism across the domestic artificial intelligence hardware space, supporting China’s strategic ambitions to challenge the US in AI development

JUST IN: Micron CEO Sanjay Mehrotra pushes back on DRAM bear thesis, citing AI capex cycle tailwinds.

Interesting bullish read-through for $MRVL from SK hynix $SKHY today. SK hynix's long-term CPO roadmap shows multiple AI accelerators accessing a shared memory pool through optical links. That looks very similar to the architecture Marvell is pursuing with Photonic Fabric and optical shared memory. Even more interesting: SK hynix and Marvell are already collaborating on CXL-based memory-centric AI infrastructure. Not a CPO customer win. But pretty strong validation of the direction Marvell is betting on.

ROADCOM IN TALKS TO RAISE MORE THAN $60 BILLION IN DEBT FOR AI CHIP FINANCING, WITH TOTAL FUNDING POTENTIALLY REACHING $100 BILLION. || DEAL COULD HELP ANTHROPIC AND OTHERS SECURE AI CHIPS AND DATA-CENTER CAPACITY, UNDERSCORING THE MASSIVE BORROWING REQUIRED TO FUND THE AI BOOM.

Well, part of the mystery surrounding why we have been seeing CDS spreads widen across the semiconductor sector may have been solved. Today, it was revealed that Broadcom is looking to raise up to $60 billion in debt to fund its AI buildout. The credit market has been right to worry about increasing debt loads, and it will be interesting to see whether spreads continue to widen now that this news is out. I'm guessing they will widen further, and not just for Broadcom.