Gradient CEO Eric Yang believes the next major shift in artificial intelligence will be driven by distributing training across a global, permissionless network of compute rather than centralized corporate supercomputers. Speaking with Yellow.com, Yang said dominant AI labs like OpenAI, Google, Anthropic, and xAI assume foundation models can only be trained inside massive, centralized infrastructure. Gradient claims it has already achieved successful reinforcement-learning training runs distributed across independent data centers, with performance rivaling centralized RLHF workflows. Yang says this opens the door to trillion-parameter model post-training conducted by thousands of compute providers worldwide. He describes a global “bounty-driven” marketplace where GPU operators, data centers, and small independent providers compete to contribute compute, earning rewards for supplying compute at the lowest price while training costs fall below centralized alternatives. Yang also argues decentralized AI infrastructure provides security and trust advantages, noting that if inference runs on user-owned hardware like MacBooks, desktops, or home GPUs, personal data never leaves the device. “Today we’re leaking far more sensitive data into AI systems than we ever did into Google,” he said. He adds that training data provenance recorded on-chain can show which environments and contributors shaped a model, countering biases and opaque editorial control. Yang envisions a “sea of specialized models” trained and owned collaboratively, stating, “Every company will run AI just like they run analytics today. When that happens, a global decentralized compute network becomes the only model that scales.”