Uploaded September 2026 | Updated September 2026, 1 day ago
Every headline about NVIDIA's Vera Rubin says the same thing: more compute, more memory, more throughput. The bigger change is what NVIDIA is now selling — the rack. Mike San Miguel from Luxor's GPU/CPU Hardware team breaks down what the Vera Rubin platform is, how the reference architecture changes what you're purchasing, and what it takes to actually deploy it.
190 to 230 kilowatts per rack. Mandatory direct-to-chip liquid cooling. A 12 to 18 month retrofit window. This is a planning decision, not a procurement one.
CHAPTERS
0:00 The real story isn't more compute
0:45 What Vera Rubin actually is
1:41 The third processor: Groq LPU
1:58 NVL144 vs NVL72 vs VR200
2:22 Reference architecture and why it matters
3:02 The rack is the new unit of compute
3:32 Hopper, Blackwell, Vera Rubin
4:41 Three generations side by side
5:23 Why NVIDIA split the rack
6:19 The Groq deal and the LPX rack
7:47 Your facility is the gate
9:05 Is it worth buying for inference?
10:04 Where your money may be better spent
10:40 The takeaway
📧 Talk to our team: sales@luxor.tech
🔔 Subscribe for GPU market updates & hardware deep dives.
#nvidia #verarubin #gpu #aiinfrastructure #datacenter
Every headline about NVIDIA's Vera Rubin says the same thing: more compute, more memory, more throughput. The bigger change is what NVIDIA is now selling — the rack. Mike San Miguel from Luxor's GPU/CPU Hardware team breaks down what the Vera Rubin platform is, how the reference architecture changes what you're purchasing, and what it takes to actually deploy it.
190 to 230 kilowatts per rack. Mandatory direct-to-chip liquid cooling. A 12 to 18 month retrofit window. This is a planning decision, not a procurement one.
CHAPTERS
0:00 The real story isn't more compute
0:45 What Vera Rubin actually is
1:41 The third processor: Groq LPU
1:58 NVL144 vs NVL72 vs VR200
2:22 Reference architecture and why it matters
3:02 The rack is the new unit of compute
3:32 Hopper, Blackwell, Vera Rubin
4:41 Three generations side by side
5:23 Why NVIDIA split the rack
6:19 The Groq deal and the LPX rack
7:47 Your facility is the gate
9:05 Is it worth buying for inference?
10:04 Where your money may be better spent
10:40 The takeaway
📧 Talk to our team: sales@luxor.tech
🔔 Subscribe for GPU market updates & hardware deep dives.
#nvidia #verarubin #gpu #aiinfrastructure #datacenter
