11 Comments
User's avatar
AY's avatar

Congratulations on the win at ur day job.

Fair to say then it is highly unlikely MRVL wins 2nm Trainium? When do you expect this decision? I hear estimates varying from 4Q 25 to middle 26.

Expand full comment
Peter W.'s avatar

About that Morgan Stanley analyst and his softball question: is Morgan Stanley handling any of Softbank's activities, maybe in the M&A space? Maybe he didn't want to upset that lucrative apple cart 💰😁.

Expand full comment
Anmol's avatar

Still ARM is a lifetime stock. Game will change when Nvidia will enter the PC market with her arm chips with nvidia graphics. As we know Nvidia has a whopping 90% share in the PC graphics. but I think Nvidia will not have much success in the beginnings due to not much apps. and games available natively on arm but that will eventually change in just 1 to 2 years if Nvidia forced the game developers to move on from x86.

Expand full comment
Warren's avatar

How can MRVL be channel stuffing when days sales outstanding is dropping?

Expand full comment
Vrangr's avatar

It's incredible how bullish Harlan (JPM analyst) is on Marvell. He is really really sticking his neck out to defend Marvell's Trainium 3 and Maia 3 wins...

Surprising his checks are so different from most buyside folks...

Expand full comment
Japdongsanee's avatar

So... even more long Astera Labs here?

Expand full comment
Peter W.'s avatar

As @AY wrote, congratulations on the win at your day job!

And here some questions about "Marvell botched porting their 224G SerDes from N5 to N3P.". If an experienced foundry like TSMC fabs a chip that doesn't meet specifications, is that the fault entirely the customer's (here: Marvell)? Especially if the design was performing well enough in an older node (N5), and the foundry is famous for supposedly industry-leading PDKs and execution? So, I wonder if TSMC is getting a free pass here? Or did they tell Marvell about possible problems, and Marvell told them to go ahead regardless?

Expand full comment
Ray Wang's avatar

What if MS analyst really get fired later 😭😭😭😭. Great post. Always enjoy your article

Expand full comment
Peter W.'s avatar

The only new-ish chiplet design with ARM cores is Fujitsu's Monaka. That 144 core design is quite interesting, but (per Fujitsu) not expected before 2026.

And, regarding Samsung Foundry's "4nm" node, Samsung has definitely plenty of available capacity in that node, so is desperate for customers. After initial rumors that AMD would have their next I/O die fabbed in that node, AMD clarified just recently that this won't happen. I guess they got a more convincing offer from TSMC.

Expand full comment
Mr_V's avatar

Do you think Qualcomm will buy Sifive as they pivot to RISCV, especially with recent acquisition rumors: https://www.semiaccurate.com/2025/04/02/is-someone-looking-to-acquire-sifive/

Expand full comment
Jay's avatar

Qualcomm are already targeting Alphawave which, among many other things, will give Qualcomm access to the RISC-V IP from Alphawave's previous acquisition of SiFive's custom silicon division OpenFive.

Expand full comment