At a high-stakes fire-side chat during the AMD AI Developer Day in Shanghai, 01.AI CEO Lee Kai-fu and AMD CEO Lisa Su delivered a sobering message to the corporate world: the era of 'experimental AI' is coming to an end. Lee, a veteran of both Silicon Valley and the Chinese tech ecosystem, argued that any AI deployment that does not fundamentally alter a company’s financial statements is essentially a 'waste of money.' This perspective marks a significant shift from the hype-driven cycles of 2023 and 2024 toward a ruthless focus on return on investment (ROI).
The dialogue outlined a clear trajectory for the industry, predicting that 2026 will serve as a watershed year. While 2024 focused on simple task completion and 2025 will likely see AI managing complex workflows, the leaders expect 2026 to be the year AI begins replacing entire functional departments. Lee noted that the breakthrough will come as AI coding capabilities cross a critical threshold, enabling 'autonomous agents' to manage end-to-end digital behaviors that were previously the domain of human teams.
Central to this transformation is the transition from single-agent AI to multi-agent architectures. Drawing a parallel to the 'Medici Effect,' Lee explained that when diverse AI experts—such as a 'recruitment agent' and a 'performance agent'—are networked together, their collective output exceeds any single model's capability. Su reinforced this by revealing that AMD’s own engineers are already utilizing AI agents to compress work that once required entire teams into the hands of a single professional equipped with the right compute power.
However, the transition faces a significant internal hurdle: corporate governance. Lee posits that the traditional bottom-up approach led by Chief Information Officers (CIOs) is destined for failure. Because AI transformation requires redefining core business logic rather than just managing software, it must be a 'Top-Down' initiative led by the CEO. For AI to be effective, it must penetrate high-stakes areas like supply chain management, dynamic pricing, and innovation—sectors where executives are often most hesitant to relinquish human control.
Addressing the hardware-software divide, the two leaders officially launched 'Cube01,' a private compute node powered by AMD Ryzen processors. This hardware is designed to alleviate the high costs and privacy concerns associated with public cloud AI, offering businesses a 'private car' experience for their sensitive data. Lee encouraged the next generation of developers to embrace the 'Directly Responsible Individual' (DRI) model, where a single human orchestrates a cluster of AI agents to deliver quantifiable business outcomes, potentially birthing the era of the 'one-person billion-dollar company.'