Yes ONTO is good. I have a bunch (~3% net position, largest semicap position). Only ONTO and CAMT feel like safe semicap right now. AMAT and KLAC need another 10-15% hit before I start aggressively buying.
Depends on the timeframe. Within next two years, 0 lol. Good question. I find the claims by Credo that their LRO solution is easily extendible to 224G to be very implausible.
Thanks for your insightful articles. Question: is the AMD MI350 locked into Samsung HBM3e, or did AMD take the time to revamp the silicon and use a different memory supplier? Any ideas?
As an analog designer myself, it's just amazing to read your articles. Wish I could find you earlier! A few questions here: 1. You seem to assume crdo has a chance to go much higher because they may gain a lot of revenue from LRO. But the management clearly said that they only expect revenue from optical to be ~10% FY25. The main driver is still from AEC. 2. What is your thought about alphawave, theire valuation and marketcap is way too cheap...
I am a little confused on the Marvell/Alchip point. You are basically saying that Marvell has not developed 224G SerDes IP - at least not what AWS needs in Trainium 3? Is it clear to you that Alchip has won the Trainium 3 order? Why or why not?
Irrational, I'd love to know why you're so confident Samsung won't get it's act together (99% is huge). 10-15 years ago heads would have rolled, new sub-leaders would have come in, and the problem would be fixed. What has changed?
Many of my biases are from how Samsung Logic Foundry has behaved throughout my career over the past decade. There has been zero improvement. DS-Memory failing Nvidia qualification four times in one year looks just like what Samsung Foundry does. Engineer flight to SK Hynix indicates the rank-and-file have given up too.
I’ve been nibbling on ONTO right now since the stock is a bit repressed. Someone I know in the industry likes their software so that’s a plus I guess.
Yes ONTO is good. I have a bunch (~3% net position, largest semicap position). Only ONTO and CAMT feel like safe semicap right now. AMAT and KLAC need another 10-15% hit before I start aggressively buying.
What do you think the probability of LRO being successful at 1.6T is?
Depends on the timeframe. Within next two years, 0 lol. Good question. I find the claims by Credo that their LRO solution is easily extendible to 224G to be very implausible.
Thanks for your insightful articles. Question: is the AMD MI350 locked into Samsung HBM3e, or did AMD take the time to revamp the silicon and use a different memory supplier? Any ideas?
Hi IA,
As an analog designer myself, it's just amazing to read your articles. Wish I could find you earlier! A few questions here: 1. You seem to assume crdo has a chance to go much higher because they may gain a lot of revenue from LRO. But the management clearly said that they only expect revenue from optical to be ~10% FY25. The main driver is still from AEC. 2. What is your thought about alphawave, theire valuation and marketcap is way too cheap...
I am a little confused on the Marvell/Alchip point. You are basically saying that Marvell has not developed 224G SerDes IP - at least not what AWS needs in Trainium 3? Is it clear to you that Alchip has won the Trainium 3 order? Why or why not?
Looks like fund managers got back from ski vacation and put on the The Great HBM Trade this morning.
Irrational, I'd love to know why you're so confident Samsung won't get it's act together (99% is huge). 10-15 years ago heads would have rolled, new sub-leaders would have come in, and the problem would be fixed. What has changed?
Many of my biases are from how Samsung Logic Foundry has behaved throughout my career over the past decade. There has been zero improvement. DS-Memory failing Nvidia qualification four times in one year looks just like what Samsung Foundry does. Engineer flight to SK Hynix indicates the rank-and-file have given up too.