NVIDIA AI Updates: April 16, 2026

1. NVIDIA Launches Ising: First Open AI Models for Quantum Computing

NVIDIA. NVIDIA released Ising, the first family of open-source AI models for quantum computing, addressing processor calibration and error correction. The suite includes a vision-language model for automated calibration that reduces time from days to hours, and 3D CNN models for real-time error-correction decoding that are 2.5x faster and 3x more accurate than traditional approaches. Available on GitHub, Hugging Face, and build.nvidia.com. Source

2. Jensen Huang Says Mythos Shows Need for US-China AI Dialogue

NVIDIA. In an interview on the Dwarkesh Podcast, NVIDIA CEO Jensen Huang stated that Anthropic’s Mythos breakthrough demonstrates the need for greater US-China cooperation on AI safety. He warned that China has sufficient energy and mature chip manufacturing to train competitive models even with export controls limiting them to 7nm processes. Source

3. Rethinking AI TCO: Why Cost per Token Is the Only Metric That Matters

NVIDIA. NVIDIA published a blog post arguing that traditional data centers have evolved into “AI token factories” where inference is the primary workload in the agentic AI era. The post makes the case that enterprises should evaluate AI infrastructure by cost-per-token rather than peak chip specs, highlighting that Blackwell architecture lowers cost per million tokens by 15x versus the prior generation. Source

4. Adobe Premiere Color Grading Mode Accelerated on NVIDIA RTX GPUs

NVIDIA. Ahead of NAB Show 2026, NVIDIA announced a new color grading mode for Adobe Premiere Pro that runs entirely on RTX GPUs. The feature offers six luminance adjustment zones beyond traditional highlights/midtones/shadows, 32-bit color depth precision, and GPU-accelerated playback in a dedicated grading environment. Source