Regarding the Token economy, Huang Shan, Strategic Technology Director of Lenovo China's Infrastructure Business Group, recently offered an observation: what enterprises are truly anxious about is not the technology itself, but the inability to clearly calculate the economics. This observation precisely captures the core proposition of the current industry transitioning from "technology exploration" to "value realization."
Huang Shan pointed out that a Token factory is not an intelligent computing center, but a complete system encompassing "top-level planning 鈥?data readiness 鈥?intelligent computing center 鈥?model services 鈥?workflow alignment." A Token factory is a way of producing intelligence that can be standardized, scaled, traceable, predictable, and billable. The standardized and billable characteristics of a hybrid token factory allow computing power to be "instantly accessible" like water and electricity, lowering the barrier for thousands of industries to use AI, echoing the strategic goal of computing power inclusiveness under the "East Data, West Computing" initiative.
The biggest variable in Token cost is computation, and computing efficiency optimization accounts for more than one-third to one-half of that. Huang Shan breaks this down into a "nine-layer tower": the computing layer has 7 layers of optimization, including operator libraries, communication libraries, memory semantics, inference framework parallel strategies, and more, with each layer stacking roughly a twofold improvement; the hardware system layer sees efficiency differences of 30%-40% across different model and chip combinations, and GPU composition methods create another twofold difference; the operations layer includes model gateway routing, agent orchestration, Token Hub, FinOps, and anti-time-fragmentation scheduling; and the infrastructure layer covers IDC construction, power selection, liquid cooling, and more.
"The cost of a single Token can go from ten yuan down to one yuan," Huang Shan said. "Our tagline is: unleash every ounce of computing power's efficiency." Huang Shan judges that AI penetration still follows previous technological pathways, and not all scenarios are suited for AI-native approaches 鈥?systematic planning and industry know-how are required. "AI algorithms cannot replace all algorithms; they only play a tremendous new role in certain specific scenarios," Huang Shan believes.
Huang Shan holds that super nodes will become the new form of public inference services, replacing traditional cluster computing forms. He cited analysis from Huatai Securities showing that super node revenue in 2025 will still be relatively limited; however, as inference scenarios begin adopting super nodes, the market will experience explosive year-over-year doubling growth. The fundamental driving force behind this is that Scaling Law is still at work, making models smarter.
Huang Shan emphasized that super nodes themselves are still evolving rapidly: "Super nodes will become the new form of public inference services in the future, and over at least the next 3 years, super nodes will continue to evolve." From a three-year perspective, inference-specific chip pathways are also worth watching. TPU, NPU, and LPU each have their own advantages and disadvantages. Autonomous driving chips are already a clear track, and research and other application areas may also produce specialized chips.
Against the backdrop of national strategies like "AI Plus" and "East Data, West Computing," computing power is transforming from a scarce resource into inclusive infrastructure. The prerequisite for inclusiveness is that someone drives down costs, clarifies the economics, and lands solutions in industries. What Lenovo Group Ltd (ASX: 00992) is doing is exactly this: supporting computing power inclusiveness with systematic technological capabilities, enabling every ounce of computing power to be converted into tangible productivity. "Calculate costs clearly, understand input-output clearly. Unleash every ounce of computing power's efficiency, and let every ounce of computing power be converted into tangible productivity." This is Huang Shan's judgment, and also Lenovo's commitment in the Token economy transformation.
Comments