China’s capital expenditure on AI is projected to reach between 600 billion yuan (US$83.8 million) and 700 billion yuan (US$97.8 billion) in 2025, according to a report by Bank of America.

This marks a potential 48% increase compared to 2024.

Government investment is expected to account for up to 400 billion yuan (US$55.8 million), while major internet companies may contribute as much as 172 billion yuan (US$24 million).

The rest will come from telecom operators and special-purpose bonds.

The rise in AI spending follows the emergence of DeepSeek, a startup in Hangzhou that launched two open-source AI models, V3 and R1.

These models were developed at lower costs and with less computing power than typical LLM projects.

🔗 Source: South China Morning Post

🧠 Food for thought

1️⃣ China’s AI spending surge represents a strategic shift from planning to implementation

China’s projected $98 billion AI investment in 2025 marks an acceleration of a long-term strategy that began years ago with ambitious policy frameworks.

The 2017 New Generation Artificial Intelligence Development Plan initially set the goal of building a $150 billion national AI industry by 2030, establishing China’s aspirations in this space well before the current surge 1.

Government funding is expected to provide the majority of this investment, up to $56 billion, reflecting China’s state-directed approach to strategic technology development 2.

This represents a substantial increase from previous years, as research from 2019 estimated China’s public AI R&D spending was only in the few billion dollar range, indicating a dramatic acceleration of investment 3.

The timing aligns with China’s goal to achieve full AI development independence by 2030, suggesting this increased spending is part of a coordinated effort to meet that timeline 4.

2️⃣ DeepSeek’s cost-efficient approach challenges assumptions about AI development economics

DeepSeek’s breakthrough demonstrates that effective AI development may not require the massive capital expenditure previously assumed necessary in this field.