What happened
Small and medium-sized enterprises (SMEs) in China are facing significant financial pressure from rising AI computing costs, hindering their digital transformation efforts. Many report that despite recognizing AI's potential, the high expenses make them hesitant to invest.
The price surge is attributed to a medium- to long-term supply-demand imbalance, driven by insufficient high-end chip production, restricted overseas supply, and rising storage and energy costs. Policy tools like computing vouchers offer only short-term relief.
Various local initiatives are underway, including computing supermarkets and banks that allow resource sharing, and new models like Guizhou Mobile's 'token factory' that charges based on actual usage, reducing the need for costly infrastructure.
Why it matters
AI computing is a critical factor for SME innovation and competitiveness in the AI era. Without affordable access, many SMEs risk falling behind, unable to leverage AI for efficiency gains.
The challenge is not just cost but also a lack of technical expertise and clear standards, making it difficult for SMEs to assess the return on investment. This calls for a systemic solution involving policy, infrastructure, and services.
Coordinated efforts from government, state-owned enterprises, and private firms are essential to create a supportive ecosystem, including standardized vouchers, shared computing pools, and tailored services for diverse industry needs.
Key facts
A Beijing AI startup reports monthly per-capita token usage costs exceeding 6,000 yuan for R&D.
In Nantong, Jiangsu, a company received a 50% subsidy on a 12-month computing service, reducing costs from 504,000 yuan to 252,000 yuan.
Guizhou has built a national hub scheduling platform for integrated computing networks, offering low-cost, reliable services to eastern regions.
What to watch next
Whether national standards for computing vouchers and cross-regional subsidy mutual recognition will be established to streamline support.
The expansion of computing supermarkets and banks, and the adoption of usage-based billing models like per-card-hour or per-token pricing.
How state-owned enterprises enhance their role in providing affordable computing pools, pre-processed datasets, and consulting services for SMEs.
