Funny story
Micron insinuated a few days ago that Apple was aggressive with pricing negotiations a few years back that got memory guys in the doldrums
I just recalled an incident in 2022
Remember the guy in the photo? Tony Blevins
He was Apple's top negotiator as VP of procurement, with a direct reporting line to Tim Cook, around the time when the last memory negotiations happened
Blevins was so good at his job that he became notorious for his ruthlessness in negotiating favorable contracts on behalf of Apple, where he was nicknamed "The Blevinator"
He would often schedule meetings with various suppliers on the same date at Apple's headquarters, so that they would become aware of their competition while waiting in the lobby
He was fired in 2022 over a viral TikTok video when he was asked what he did for a living, while exiting his fancy Mercedes SLR McClaren
His reply, and I quote:
“I race cars, play golf and fondle big-breasted women. But I take weekends and major holidays off. Also, if you’re interested, I’ve got a hell of a dental plan.”
That's a joke of course. But in poor taste
Let's hope Apple has a better person in charge of procurement now
Must read for everyone
I agree with most points written in the article
Although, I disagree on CXMT points (they are quite active in 3D DRAM, and we may see results in 2028, capacity will also be greater than 500k WPM)
And lots of work is happening to reduce KV cache footprint (DeepSeek v4),
Expect the size of KV cache/token to go down but the volume of tokens generated/processed will be go up so much that it will make up for it
Jevon's paradox
The 800V transition discourse has descended into full retail psychosis. Watching people rank Navitas vs. Infineon vs. onsemi by who has the lowest RDS(on)·Qg product is genuinely hilarious because you fucking retards think a 15% advantage over one supplier or the other matters when that ‘edge’ doesn’t survive the first round of price negotiations.
If your investment thesis requires you to cross-reference reverse recovery charge curves, zero-voltage switching loss at 1MHz, or whether GaN HEMT enhancement-mode beats SiC MOSFET trench in a specific 11kW OBC topology…. congratulations, you’ve already lost. If you need to LARP as a power electronics engineer to justify a position you probably shouldn’t have the position at all.
People got rewarded once for being “technical” about HBM stacks and now think every semi cycle resolves by reading spec sheets harder. It doesn’t. The 800V transition isn’t a horse race and margin compression in WBG power discretes is already accelerating, the Chinese GaN entrants are coming in 30-40% under incumbent ASPs. Also, ‘GaN vs. SiC’ isn’t one trade, it’s four trades where traction inverter, OBC, DC-DC, and server PSU each have different winners and different timelines, and anyone pitching a single device vendor as ‘the 800V play’ is telling on themselves for how little they know about anything. The Infineon bulls and the Navitas bulls are both going to get cooked because they’re fighting over who supplies the picks during a gold rush where the actual scarcity is somewhere else entirely.
There’s a vastly cleaner way to express this thesis with better unit economics, less competitive intensity, and exposure that compounds through the entire 800V buildout regardless of which device wins which socket. But watching people brute force datasheet comparisons into a thesis is genuinely entertaining, so I’m content to let the AI psychosis play out a bit longer.
Explains why everyone keeps underestimating and is constantly surprised by the ability of $BE to scale.
$INTC has also had Bloom installations for a while.
@magicsilicon A little known connection with Bloom $BE is that KR Sridhar directly credits Andy Grove for their manufacturing (no idea if in person or spirit) -
“On thinking about mass manufacturing and bringing costs down like semiconductors. Andy Grove from Intel”
https://t.co/cxZksswWT7
I interviewed @bubbleboi about his ratings of AI supply chain bottlenecks.
We talked about DRAM, advanced packaging, CPO, HBF, PCBs, power delivery, etc.
0:00 HBM, DRAM, the cartel
7:24 Silicon photonics, CPO, Lumentum lasers
11:35 Advanced packaging: TSMC vs Intel
13:49 HBF: Sandisk/SK monopoly window
17:02 Memory accelerators + TurboQuant
22:35 PCBs + Unitika
27:27 Power: transition from 48V to 800V
31:32 Outsourced assembly + test, fiber coupling
36:23 HBM vs DRAM vs NAND in 2-3 years
42:03 Where will hardware founders come from?
44:10 Alt accelerators: Etched, Taalas, MatX
48:43 Emerging tech: CPO bearishness
49:23 Hyperclouds
49:53 HBF timeline + CXL
51:48 Voltage + cooling wall
56:30 Rapid fire: Intel, Nvidia, TSMC, Alphabet
1:05:25 ASML, Hynix, Lumentum, Wolfspeed
1:13:25 Building conviction in Intel
1:19:37 Pitching Intel to funds
1:22:39 X accounts, analysts + why care about any of this
High voltage SiC and GaN is the most miscovered sector in semis right now.
The alpha from public datasheets is incredible.
I am huge solid state transformer bull now.
People keep confusing a bubble with “stocks go up and get overvalued”. A bubble is when when a prevailing trend and a prevailing misconception about that trend interact reflexively, each reinforcing the other until the gap between perception and reality becomes unsustainable.
A bubble is not when everyone realizes that right now every iota of AI demand eventually, at some point upstream, must move through memory OEMs. Nor is it when estimates continue rising because things are better than expected. And it’s not just when stocks trade expensive to historical valuations.
The reason behind the moves in the AI infrastructure layer so far have been simply that we don’t have enough. They’ve been driven by the fundamental reality more than the perception of the future. It’s why the bulk of the most bullish parts of this cycle have been lumpy and centered around earnings season when companies uniformly come out and confirm there’s still not enough. In the bubble, the reality is driven by the market - not the other way around.
Everyone keeps saying “people are gonna freak out if it’s not a bubble!”. I think that’s silly, we have a transformative new technology that needs crazy capital to fuel it coming to fruition, that has and always will result in a bubble as long as we have financial markets.
But if you want to call the top in a bubble, you need a much stronger view on what the misconception is and what negative catalyst forces broad perception to align with realizing it than you do on valuation.
This article was written in collaboration with a current member of Intel Foundry’s research organization. All content was prepared in full compliance with NDA obligations.
Intel is coming back.
Many people explain this through separate pieces: product momentum, process roadmap, foundry strategy, AI server demand, and advanced packaging. I think all of those pieces matter.
But if someone asked me what Intel’s strongest technical moat is today, my answer would be four letters.
STCO.
In this article, I break down what STCO actually means, why it could become Intel’s unique technical edge in the AI era, and which companies may benefit from this structural shift.