$NVDA is pushing AI agents as the next computing platform — not just smarter chatbots, but systems that pick models, use tools, and execute real work.
At AI Day Seoul, Jonathan Cohen highlighted Switchyard cutting costs to roughly one-third while holding success rates near 80%. Nemotron 3.5 runs about 5x faster. Korean partners like Naver Cloud and SKT are already building localized models on the stack.
The thesis: enterprise AI only scales when cost and control improve together. If agents can meaningfully reduce operational expense while maintaining output quality, adoption accelerates.
This is about infrastructure leverage — the kind that compounds when enterprises deploy at scale. Watch how quickly regional cloud providers integrate and whether cost savings translate into broader rollout timelines. 🤖
At AI Day Seoul, Jonathan Cohen highlighted Switchyard cutting costs to roughly one-third while holding success rates near 80%. Nemotron 3.5 runs about 5x faster. Korean partners like Naver Cloud and SKT are already building localized models on the stack.
The thesis: enterprise AI only scales when cost and control improve together. If agents can meaningfully reduce operational expense while maintaining output quality, adoption accelerates.
This is about infrastructure leverage — the kind that compounds when enterprises deploy at scale. Watch how quickly regional cloud providers integrate and whether cost savings translate into broader rollout timelines. 🤖