cd /news/artificial-intelligence/nvidia-touts-vera-rubin-performance-… · home topics artificial-intelligence article
[ARTICLE · art-67182] src=ca.finance.yahoo.com ↗ pub= topic=artificial-intelligence verified=true sentiment=· neutral

Nvidia touts Vera Rubin performance ahead of rival AMD's Advancing AI event

Nvidia is touting performance gains for its Vera Rubin platform ahead of AMD's Advancing AI conference, with CoreWeave seeing 10 times more tokens per watt running DeepSeek's R1 model compared to the prior-generation Grace Blackwell system. Nvidia also claims Vera's CPU provides up to 1.9 times better agentic AI performance versus AMD's Epyc Turin processor in benchmarks. The announcements come as AMD prepares to unveil its Helios rackscale server, a direct competitor to Nvidia's NVL72, at the conference in San Francisco.

read2 min views1 publishedJul 21, 2026
Nvidia touts Vera Rubin performance ahead of rival AMD's Advancing AI event
Image: Ca (auto-discovered)

Nvidia (NVDA) is touting the performance gains its Vera Rubin platform offers over its prior-generation Grace Blackwell system, saying CoreWeave (CRWV) is seeing 10 times more tokens per watt while running DeepSeek's R1 model with the new setup.

Tokens are a unit of measurement in AI systems representing parts of words and phrases you input into and receive from an AI model.

The company is also boasting about the Vera central processing unit's (CPU) standalone capabilities, saying it provides up to 1.9 times better agentic AI performance in benchmarks versus AMD's (AMD) Epyc Turin processor.

The announcements come ahead of AMD's Advancing AI conference, set to kick off in San Francisco on Wednesday. CEO Lisa Su will hold a keynote address on Thursday.

AMD will likely talk up its upcoming Helios rackscale server platform, its first such offering for AI data centers and a direct competitor to Nvidia's NVL72 rackscale server.

Helios features 72 of the company's Instinct MI455X graphics processing units (GPUs), along with its Epyc Venice CPUs, Pensando data processing units (DPUs), and ROCm software.

Nvidia's NVL72 includes 72 Blackwell or Rubin GPUs, its Grace or Vera CPUs, NVLink networking system, and CUDA software. Nvidia is also preparing a new NVL144 setup that will feature a whopping 144 Rubin GPUs.

Nvidia continues to dominate the data center market, bringing in $215.9 billion in revenue in its fiscal 2026, well ahead of the $34.6 billion AMD generated in its fiscal 2025. But new competitors continue to threaten its reign.

In addition to AMD's Helios, Nvidia's own customers, including Google (GOOG, GOOGL) and Amazon (AMZN), rent out their custom AI accelerators to third parties, including Anthropic (ANTH.PVT). Meta (META) also has its own chips, which it uses to reduce its dependence on Nvidia, as does Microsoft (MSFT).

OpenAI (OPAI.PVT) and Broadcom (AVGO) also recently announced their own AI processor.

But Nvidia regularly updates its AI chips and systems, debuting a new lineup in such regular cadences that it could prove difficult for its competitors to keep pace.

That said, the AI market is rapidly evolving, and the rise of AI agents, which forced companies to focus more on CPUs, is proof of that alone. And it's unlikely to slow down anytime soon.

Email Daniel Howley at dhowley@yahoofinance.com. Follow him on X at @DanielHowley.

Click here for the latest technology news that will impact the stock market**.**

Read the latest financial and business news from Yahoo Finance

── more in #artificial-intelligence 4 stories · sorted by recency
── more on @nvidia 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
Live at https://your-agent.zahid.host
Get free account → Pricing
from €0/mo · no card required
LIVE [news/nvidia-touts-vera-ru…] indexed:0 read:2min 2026-07-21 ·