I had a quick chat with SIMO IR yesterday to clarify the NVDA Bluefield project and how SIMO fits in.
SIMO is qualified for the system's boot storage. This is some type of Ferri module.
Each Bluefield 3 system can connect to up to 30 128TB SSDs. SIMO is trying to help enterprise SSD makers develop 128TB SSDs which use the Mon Titan controller. SIMO itself is not making enterprise SSDs in this system.
I asked Perplexity - Nvidia is developing something called the bluefield 3 DPU. What's a reasonable expectation for the number of Bluefield 3 DPUs that might sell in 2026?
perplexity.ai
A reasonable expectation is that Nvidia could sell between 400,000 and 2,000,000 BlueField-3 DPUs in 2026, depending on market growth, adoption rates, and average selling prices
hmmmm
So lets say NVDA sells 1 million Bluefield-3 DPUs in 2026. If the Ferri-based boot storage module is $5 (??), that's $5 million revenues. And if SIMO's Mon Titan gets a 20% share of the enterprise controller chips, that's
1 million units x 30 enterprise chips per unit x 20% x $50 price per Mon Titan = $300m sales of Mon Titan in 2026 just due to Bluefield 3 DPU sales.
----
Hmmm, that seems optimistic, but I don't know the Bluefield 3 DPU market segment well enough to tell what's off or whether that expectation is reasonable.
One thing is Bluefield 3 DPU can accomodate up to 30 128TB SSDs, but perhaps they sell with 2 128TB SSDs, and more SSDs get added as need each quarter?
I don't know, but this was fun! |
