NVIDIA paying half the capital expenditure to expand supplier fabs for critical thin-film material.

Announced Nemotron 3 Nano Omni, an open multimodal model processing video, audio, images, and text.

Nvidia invests $2 billion in Marvell Technology for NVLink Fusion interconnect development

Nvidia trained a billion-parameter LLM without backpropagation or full-precision weights, using zero gradients.

Released V4-Pro and V4-Flash open-weight models with up to 1.6 trillion parameters

Released DeepSeek-V4 with 1M-token context at 10% KV cache cost

Open-sourced Kimono, a motion diffusion model for humanoid robots

Expanded partnership with Google Cloud to advance agentic and physical AI infrastructure

Initiates first external fundraising round, targeting at least $300M at a $10B+ valuation to set a benchmark for employee stock options.

Delays to DeepSeek-V4 model due to Huawei Ascend chip compatibility work