NVIDIA CEO Jensen Huang said at NVIDIA GTC in March 2026 that open and proprietary artificial intelligence will coexist, stating, “proprietary versus open is not a thing. It’s proprietary and open.” He argued that the global AI ecosystem will require different models for different workloads, industries, and security requirements rather than one universal system. Frontier proprietary models may continue leading tasks demanding maximum reasoning performance, managed security, and enterprise support, while open-weight models can give developers, businesses, and governments greater control over customization, data processing, and deployment. Huang’s view has implications for crypto and decentralized AI: if advanced model weights become more widely available, the next bottleneck will be the infrastructure required to train, customize, and operate them. This creates a potential opening for decentralized GPU networks, blockchain-based AI markets, and protocols such as Bittensor, but it does not guarantee that every decentralized AI project or related token will succeed. The central question is whether decentralized networks can provide useful compute, inference, storage, payments, and verification at a level that businesses and developers are willing to purchase. Technical performance, network reliability, token economics, and real customer demand will determine whether crypto becomes an important infrastructure layer or remains a speculative narrative around AI growth.