Fetching from the wire…
01
02
03
04
05
06
07
08
Public story · 2026-08-24 · high
The chip talk comes two days before Nvidia's Q2 FY2027 earnings call on August 26, with Rubin GPU details the same afternoon.
Why now: The disclosure comes two days before Nvidia's Q2 FY2027 earnings call on August 26, 2026.
NVIDIA detailed the Vera CPU's 88 custom Olympus cores at Hot Chips, with a Rubin GPU talk the same afternoon. The numbers give data center buyers their first efficiency baseline for chips that won't ship for months.
NVIDIA says the Vera CPU completes tasks 1.8 times faster than a traditional x86 chip and runs at twice the efficiency. Those are the company's own numbers, not an independent benchmark, and Nvidia hasn't said which x86 chip it used for the comparison.
The full rack, Vera Rubin NVL72, packs 72 Rubin GPUs and 36 Vera CPUs into one system, alongside ConnectX-9 SuperNICs and BlueField-4 DPUs for networking and data movement. Operators will use that spec to plan power and cooling before the platform ships.
Publishing chip architecture months before shipping, two days ahead of Nvidia's Q2 FY2027 earnings call on August 26, points to timing chosen for investor attention as much as engineering interest. Watch whether the call leans on Vera Rubin NVL72 to justify next year's capital spending guidance.
Each link below shares sources, entities, or timing with this story.
Jensen Huang disclosed 100% of NVIDIA uses Claude Code, calling it "the first agentic model." Vera CPU: 88 custom Olympus cores, 1.2 TB/s LPDDR5X, paired with Rubin GPUs at 1.8 TB/s coherent bandwidth. 22,500+ concurrent CPU environments per rack. Dell, HPE, Lenovo, Alibaba, B...
NVIDIA's Blackwell successor is in production ahead of schedule. The NVL72 rack (72 GPUs) delivers 3.6 exaFLOPS for inference, with 288GB HBM4 per GPU. NVIDIA claims 10x lower cost-per-token versus Blackwell. The Rubin CPX variant — purpose-built for million-token inference —...
NVHBM, announced August 26, relocates NVIDIA's custom memory controller from the compute chip into the HBM base die, claiming up to 30% more bandwidth than standard HBM4E, 15% lower HBM power, and up to 25% more freed area on the XPU compute die (NVIDIA). Next-generation Train...
Announced at Hot Chips on August 24, it's an interactive inference accelerator extending the Vera Rubin NVL72 platform, aimed at the token-generation phase that determines how responsive an agent loop feels (NVIDIA). The cited figure runs Gemma 4 31B at 100,000-token context,...
Portable Computer launched August 26, running the orchestrator LLM, subagent LLM, planner, tool router, scheduler and local search index locally, with local work consuming no billing credits and each cloud escalation requiring separate approval (VentureBeat). Launch platform i...
On the August 26 earnings call Huang said "AI has reached its inflection point. It's doing useful work. Its tokens are productive and profitable. Now, compute is revenue," and dismissed the AGI debate as "kind of senseless" in favor of whether AI does profitable work (CNBC). N...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.