Cisco-Nvidia Rack-Scale AI Factories Move to Production as Dell and Salesforce Spotlight Enterprise AI at Six Five Summit

CSCO and NVDA are expanding their Secure AI Factory partnership, adding a new hardware track built with Supermicro that pairs liquid-cooled compute with Cisco's own liquid-cooled AI networking gear to support NVIDIA's upcoming Vera Rubin NVL72 and HGX Rubin NVL8 platforms. Cisco frames itself as the only NVIDIA technology partner running its own switches and network operating system end to end, using Silicon One on the front end and Spectrum-X on the back end, unified by Nexus One. The Supermicro-based offering targets enterprises, neoclouds, and sovereign clouds, and Cisco says it will become available in October 2026.

The announcement lands during the Six Five Summit: AI Unleashed, a three-day virtual event running August 25-27, 2026, headlined by CRM CEO Marc Benioff and also featuring the CEOs of Snowflake, Marvell, and Applied Materials. DELL's Doug Schmitt, president of Dell Services, used his summit appearance to discuss enterprises' shift from AI pilots to production workloads and the operational questions that come with running agentic systems at scale.

Benioff's core argument, laid out in a summit interview, is that most enterprises are still early in AI adoption and should not rip out core systems like SAP, Salesforce, Microsoft, or their existing data lakes to get value from AI. Instead, he argued for layering AI on top of those platforms as an architectural middle layer that unlocks "trapped value" while preserving existing security and governance controls, a middle path he expects to play out over the next 12 to 18 months. As evidence the approach is already working, Benioff pointed to Casey, a Salesforce customer-service agent that has resolved 5 million customer service issues, and to Slackbot, which now supports tens of millions of Slack users.

Separately, NVIDIA unveiled the Jetson Orin Nano 2, an entry-level edge AI robotics computer delivering 2x the inference performance of its predecessor while consuming 40% less power . NVIDIA says more than 3 million developers already use its robotics stack, with early adopters including Cognex, Doosan Bobcat, Matic, and Wing ; the new module and developer kit are expected to be available in the first half of 2027 .

Rounding out the summit's enterprise-AI theme, SNOW CEO Sridhar Ramaswamy argued that trusted, well-governed data, not model benchmarks, will determine which companies get lasting value from AI, since "the smartest model in the world cannot make sense of truly bad data" . He pointed to one customer capping spend on a critical AI application at $30 per user per month as an example of the cost discipline enterprises are applying as pilots move toward production . Together, the week's announcements suggest infrastructure vendors, software incumbents, and data platforms are converging on the same problem: moving AI out of pilot mode into reliable, governed production use, with hardware availability, integration cost, and data quality as the factors to watch.

Related Stocks

• NVDA
• CRM
• DELL
• SNOW