The Six Five with Patrick Moorhead and Daniel Newman

The episode contains genuine industry-specific detail on Qualcomm's Hot HBC architecture, Micron's margin trajectory, and Cerebras's structural problems, but these patches of substance are diluted by significant filler - family anecdotes at the open, investment banter, and Cathie Wood jokes - and many observations are surface-level restatements of consensus ('it's AND not OR on compute,' 'best time to raise is when you don't need it').

“instead of having a separate package for the memory, it actually um, is layered on as a metal layer, uh, just like you would do a monolithic chip”

“China DDR5 production ramps in 2H26. Um, you have Lenovo already using it, uh, you have HP and Dell who are qualifying it”

The structured 'flip' debate on memory-as-commodity vs. memory-as-infrastructure surfaces a few genuinely distinct arguments (JEDEC compliance as the definitional test for commodity; vendor-financing embedded in SCAs flattering optics), but the majority of framing - heterogeneous compute as 'AND not OR,' OpenAI needing to control its supply chain, raise-money-when-you-don't-need-it - is widely circulated analyst consensus.

“When you're building to a JEDEC standard, it is by default a commodity”

“these customer deposits are not cash flow. This is vendor financing embedded in these agreements. It flatters the optics”

There are no guests; this is a two-host analyst discussion between Patrick Moorhead and Daniel Newman, both firm founders and industry commentators. Patrick's 21 years working inside memory manufacturers gives his commodity-cycle arguments some practitioner credibility, but neither host is a current operator who has executed at scale inside the companies being discussed.

“I've lived through nine memory cycles. I worked for manufacturers for 21 years. I have crawled through memory factories, PCA, PCB chip, chip flipping”

“Founder and CEO of More Insights and strategy”

The episode is above-average in citing hard numbers: Micron's 1,215% YoY EPS growth, 84.9% gross margin, 16 multi-year agreements totalling $22B; Qualcomm's $5B FY27 and $15B FY29 targets; Cerebras's 47% to 36-38% margin compression; and a detailed serdes roadmap (800G copper/optical in 25/26, 1.6T in 26/27, 3.3T optical in 2028). Some claims remain unsubstantiated (ByteDance speculation, Anthropic FCF margin 'rumor').

“$25.11 per share earnings, 1,215% year over year growth, um, beating 2049, uh, expectation 84.9% record record margins”

“800 gig UM copper, 800 gig optical and EECs in 26 and 27. It's 1.6 T and then in development for 2028 UM 3.3 uh point terabit optical”

The structured 'flip' format on the memory debate is a reasonable mechanism for surfacing disagreement, and there are moments of genuine pushback (Patrick challenging Daniel's strategic-infrastructure thesis with the JEDEC standards argument). However, the hosts mostly validate each other, follow-ups rarely force deeper precision, and extended personal banter displaces potential probing.

“I saw Big Dan there. He got this. So Daniel, you are saying the era of memory as a commodity is over?”

“I can't argue with anything that you've said”