NVIDIA officially unveiled the new generation AI platform Rubin at CES, with a hardware matrix comprising six new chips—Vera CPU, Rubin GPU, NVLink 6 switch, ConnectX-9 super network card, BlueField-4 DPU, and Spectrum-6 Ethernet switch.
The key highlight is the Rubin GPU. This chip integrates the third-generation Transformer engine, with NVFP4 inference computing power reaching 50PFLOPS, directly surpassing the previous Blackwell solution—performance increased by 5 times. It’s important to understand what this level of computational upgrade means for data centers and AI infrastructure.
More importantly, the production schedule. NVIDIA has advanced the Rubin platform to full-scale production, with plans to market it through partner channels in the second half of 2026. In other words, a new round of computing power competition is about to begin.
View Original
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
17 Likes
Reward
17
5
Repost
Share
Comment
0/400
OfflineValidator
· 01-08 23:19
50PFLOPS? Oh my, are they planning to move the data center to the sky... Coming to market in the second half of 2026, this pace is a bit intense.
View OriginalReply0
FrogInTheWell
· 01-06 16:50
50PFLOPS? Claiming this performance has increased fivefold sounds almost too good to be true, but mass production isn't until the second half of next year... Fine, I'll just keep stacking Blackwell.
---
Is Rubin for real? They were still hyping Blackwell in the first half of the year, and now they’re launching new products? A compute race? Honestly, it’s just a squeeze-the-tube situation.
---
Wait, six chips launching together? Are they aiming to dominate the entire AI infrastructure? Competitors’ days might be even tougher.
---
Can only buy in 2026? Well, that’s another story by then... But a 50P inference computing power is indeed impressive.
---
Rubin GPU sounds very auspicious, but the question is: what about the cost? No matter how powerful the compute, it has to be affordable.
---
With this hardware matrix combo, data centers should upgrade and replace their equipment. My mining rigs can retire now, haha.
View OriginalReply0
MevTears
· 01-06 16:44
It's NVIDIA again, with 5x performance... Is it really true? It won't be on the market until 2026. What are we using now?
AMD and other manufacturers should have started earlier; otherwise, everything will be monopolized by NVIDIA.
These performance figures are a bit outrageous, it feels like another hype.
50PFLOPS... can they really mass produce it in the second half of the year? I don't buy it.
The computing power arms race has become too intense; the real key is who can actually bring products to market.
View OriginalReply0
nft_widow
· 01-06 16:33
5x performance boost? Won't arrive until the second half of 2026? The gap is just too big
---
Once again Blackwell dominates, and GPU is invincible. Honestly, I'm a bit tired of the hype
---
50PFLOPS sounds impressive, but what’s it good for to someone like me, a retail investor... still two years away
---
NVIDIA is draining its competitors with this pace. I just want to see who dares to compete next
---
Every time they say "a new round of competition," but in reality, it’s just NVIDIA playing alone
---
Rubin hasn't been released yet, and Boss Huang is probably already hyping up the next generation
---
Data centers are going crazy. If this upgrade continues, electricity costs could create a black hole
View OriginalReply0
just_here_for_vibes
· 01-06 16:26
50PFLOPS? This performance boost is really outrageous, five times... but it won't be released until 2026, so it still feels a bit far away.
Another round of "stuck" competition, who will be the last to laugh this time?
Rubin sounds impressive, but in the end, it still comes down to price and capacity.
If they can truly stabilize supply this time, they will win. The previous generations of cards have been blown out of proportion.
The computing power race has started again; miners probably can't sit still.
Six chips together, hardware infrastructure is really being upgraded comprehensively.
50PFLOPS, I wouldn't even dare to imagine it. The graphics card market will change in two years.
NVIDIA officially unveiled the new generation AI platform Rubin at CES, with a hardware matrix comprising six new chips—Vera CPU, Rubin GPU, NVLink 6 switch, ConnectX-9 super network card, BlueField-4 DPU, and Spectrum-6 Ethernet switch.
The key highlight is the Rubin GPU. This chip integrates the third-generation Transformer engine, with NVFP4 inference computing power reaching 50PFLOPS, directly surpassing the previous Blackwell solution—performance increased by 5 times. It’s important to understand what this level of computational upgrade means for data centers and AI infrastructure.
More importantly, the production schedule. NVIDIA has advanced the Rubin platform to full-scale production, with plans to market it through partner channels in the second half of 2026. In other words, a new round of computing power competition is about to begin.