TLDRocket
Sign in

‘Now We Can Know Everything and Do Anything,’ Jensen Huang Says at Dreamforce

NVIDIA Blog Brian Caulfield Covered by 5 sources

Jensen Huang joined Marc Benioff at Dreamforce to back Salesforce’s new Koa CRM model. It was built on NVIDIA tech, and Salesforce says it runs without using customer data.

Based on reporting by NVIDIA Blog, Brian Caulfield — read the original for the full story.

Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error

Jensen Huang took the Dreamforce stage with Marc Benioff, but he also spent part of the appearance moving through the Moscone Center crowd, mic in hand. The timing mattered. Salesforce used the event to announce Koa, its first CRM reasoning model, built on NVIDIA Nemotron 3 Super.

Huang used the moment to sell a familiar NVIDIA thesis: each big era expands what companies can do. Electricity powered everything, the internet helped people find anything, and AI, he said, is becoming an infrastructure layer that lets us “know everything and do anything.” Benioff leaned into the same idea, calling the future “agentic” and saying Trailblazers would be the ones to shape it.

The technical pitch behind Koa is unusually specific. Salesforce said it fine-tuned and ran the model entirely inside its own infrastructure because Nemotron is open. It controls the weights, and no customer data was used for training or inference. Rohan Kumar, Salesforce’s president of platform and engineering, said not a single byte of customer data was used.

Koa was trained on synthetic enterprise scenarios drawn from nearly three decades of CRM deployments and spanning more than 14 industries, including manufacturing, financial services, healthcare and travel. Salesforce says the model used supervised fine-tuning and reinforcement learning with NVIDIA NeMo RL, NeMo Gym and NeMo AutoModel. In its CRM Bench, which tests real tasks like updating opportunities, routing cases and scheduling follow-ups, Koa already matches or beats leading models on CRM actions while making three times fewer errors.

It’s already in use inside Salesforce, where it powers an employee agent in Slack. Customer pilots are set for October, with Formula 1, UChicago Medicine, Baxter Credit Union, 1-800Accountant, Engine and Xero among the first names listed. General availability is expected in winter 2026 in U.S. regions.

My take — AI-written commentary, not fact-checked reporting

The real story here isn’t the keynote bravado, it’s the quiet admission that open models now matter because closed ones aren’t the only game in town. That’s where the industry is heading: companies want AI they can actually control, not just rent by the minute and pray over. The funny part is that the safest pitch in AI may be the least mystical one: keep the data home, keep the weights in hand, and skip the cosplay.

Read more about this at: NVIDIA Blog

Related stories

The daily briefing

Every AI story that matters, in your inbox by 8am.

TLDRocket reads all relevant sources, removes duplicate coverage, and summarises the day in two minutes. Follow companies and topics for alerts, or get the briefing in Slack. Free, no spam, unsubscribe anytime.