China AI spending expected to hit 700bn yuan this year, global to reach $800bn by 2030: BofA

China AI spending expected to hit 700bn yuan this year, global to reach $800bn by 2030: BofA

China's AI spending is projected to hit 700 billion yuan this year, with global AI spending expected to reach $800 billion by 2030, according to Bank of America (BofA) [1]. However, Nvidia's dominance in China's AI chip market has faced significant challenges, as its market share dropped from 66% to 54% in 2025 [NUMBER:1]. This decline is primarily due to U.S. export controls and the rise of local competitors such as Cambricon and Huawei.

Cambricon, once a struggling IP licensor, has transformed into a formidable rival, reporting a 4,300% revenue surge in H1 2025 [NUMBER:1]. Its Siyuan 690 chips, now in mass production, are being adopted by Baidu and others for inference tasks [NUMBER:1], while its $5 billion fundraising plan signals aggressive ambitions in LLM chip development [NUMBER:1]. Meanwhile, Huawei's Ascend 910B and 910C chips power 85% of new infrastructure in China Mobile's cloud data centers [NUMBER:1], with AI-related revenue surging 240% year-over-year [NUMBER:1].

Nvidia's B30A chip, six times more powerful than the H20, faces regulatory uncertainty, while China's 55% localization target by 2027 threatens its long-term dominance [NUMBER:1]. The Trump administration's 15% revenue-sharing agreement for H20 sales underscores a strategic pivot, leveraging economic incentives to keep Chinese developers tethered to the American tech stack [NUMBER:1]. However, this approach risks alienating a market that is increasingly prioritizing self-reliance.

Nvidia's response to these challenges hinges on two pillars: supply chain agility and product innovation. By securing TSMC's CoWoS packaging technology ahead of demand [NUMBER:1], the company has maintained a performance edge in AI training. Modified chips like the A800 and H800, designed to comply with U.S. export rules, have preserved a foothold in China, albeit at reduced capacity [NUMBER:1]. The B30A chip, if approved by the U.S. government, could reassert Nvidia's dominance in high-performance training, a segment where domestic alternatives still lag [NUMBER:1].

The AI accelerator chip market is projected to grow at a 33.19% CAGR through 2032 [NUMBER:1], but Nvidia's ability to capture this growth in China depends on navigating a volatile geopolitical environment. The company's Q3 2025 revenue forecast of $54 billion highlights its global strength, but China's localization ratio is expected to reach 55% by 2027, signaling a structural shift in demand [NUMBER:1].

For investors, the key risks are twofold: the U.S. government's willingness to approve advanced chips for China and the pace at which domestic alternatives close the performance gap. Cambricon's focus on software compatibility [NUMBER:1] and Huawei's cluster computing strategies [NUMBER:1] suggest these firms are not merely imitators but strategic innovators. Nvidia's ecosystem advantage, particularly CUDA's entrenched role in global AI development, remains a moat. However, the B30A's approval could tip the balance, but until then, the company's long-term value in China will remain contingent on a fragile interplay of geopolitics and innovation.

References:
[1] https://www.ainvest.com/news/nvidia-strategic-position-china-ai-market-regulatory-turmoil-2509/