Dreamforce Attendees Say Older AI Models Are Good Enough
Business leaders at Salesforce's Dreamforce say last year's AI models deliver enough value, cooling the frenzy over cutting-edge upgrades.
The AI arms race looks different from the trenches. At Salesforce's massive Dreamforce conference in San Francisco, business leaders threw cold water on the idea that every company needs the freshest, most powerful AI models available. The consensus on the floor: last year's models are doing the job just fine.
That's a meaningful signal for anyone trading AI-adjacent stocks. The hype cycle has been built on the assumption that enterprises will perpetually chase the bleeding edge — newer chips, newer models, bigger compute budgets. Dreamforce attendees are quietly challenging that narrative, and the market should pay attention.
Read more Beauty, Health and Wellness Merge Into One Big Consumer Battleground →
The AI safety debate also collided with business reality at the event. While researchers and regulators spar over existential risk, the people actually deploying AI inside companies are focused on something far more mundane: does it work, does it save time, and can we trust it enough to build on top of it? For most attendees, the answer is yes — but with the models already in their hands, not the ones still on the roadmap.
This creates a real tension for the hyperscalers and model providers betting that corporate America will keep spending aggressively on AI infrastructure upgrades. If enterprises are satisfied sitting a generation behind, the upgrade cycle slows — and so does the revenue ramp that Wall Street has priced in. That doesn't mean the AI trade is broken, but it does mean the "must have the latest" assumption deserves a hard look.
Continue reading at US Top News and Analysis.
