The Magnificent Seven – Apple, Microsoft, Alphabet, Amazon, Meta, Nvidia and Tesla – invested more than $200 billion collectively in data centres and computing capacity in 2025, with hyperscale AI infrastructure driving the bulk of the spending.

China’s top tech giants, led by e-commerce giant Alibaba, earmarked at least CNY380 billion ($54 billion) in compute power and AI infrastructure this year.

While both countries are investing huge sums to emerge as a leader in the segment, they have vastly different goals.

Huawei CEO Ren Zhengfei last month pointed out the US is exploring artificial general intelligence (AGI) and artificial superintelligence to solve big issues, such as “what it means to be human”.

This is clearly demonstrated by Meta CEO Mark Zuckerberg’s bold ambition, outlined in July, to bring “personal superintelligence” to billions of users. It has allocated to up $27 billion to AI compute this year.

Real-world applications
In contrast, Ren said China is studying the adoption of AI in real-world scenarios, aiming to create more value and drive growth, with a focus on improving industrial efficiency, safety and profitability.

The CEO said the key technical requirement for AI is having sufficient electricity and a well-developed information network, two areas where China has a significant advantage over the US. “China has excellent power generation and grid transmission systems, and its communication network is the most advanced in the world.”

He told a gathering of students last month at its Lianqiu Lake Campus AI’s sensing and control capabilities will depend on data transmitted across thousands of kilometres, which will require advanced networks.

Leonardo Dinic, an adviser on geopolitics and international business, highlighted in an article in China-US Focus that while both countries’ AI strategies are globally ambitious, there are clear differences. 

He suggested the US “explicitly promotes the export” of end-to-end AI technology, covering hardware, models, software, applications and standards to allies and firm geopolitical partners, to prevent competing adversaries from free-riding on innovation. 

Meanwhile, China is positioning itself as a partner for Global South nations by emphasising “open technology exchange, cross-border cooperation”, which is less conditioned by each actor’s political alignment, Dinic noted.

Lagging in AI chips
Despite restricted access to the most advanced AI chips, Chinese AI companies, demonstrated by start-up DeepSeek introducing its R1 model in January, are keeping up with US leaders in many categories.

TrendForce analyst Frank Kung told Mobile World Live Chinese AI suppliers have adopted an approach of using a larger number of chips compared with US-based GPUs, such as Nvidia’s. 

He explained: “This entails disadvantages such as higher costs and a much larger data centre footprint when assembling equivalent systems, but it is also a breakthrough strategy that they are effectively compelled to pursue.”

The country’s AI suppliers face disadvantages in upstream supply, such as wafer fabrication and high-bandwidth availability, as well as in process technology, Kung added. 

As a result, their development path is mainly focused on two directions.

First, they adopt a server-cluster approach: while individual chips may underperform compared with leading alternatives, system-level performance is achieved by aggregating a larger number of chips.

Second, they place greater emphasis on AI inference applications. Although their capabilities in large-scale LLM training may be less competitive, they aim to compensate by developing more diversified AI application scenarios, he noted.

The pod advantage
Huawei rotating chair Eric Xu acknowledged at the chip level, the performance of an individual AI Huawei chip is not as good as Nvidia’s, “but at a super pod level, we can deliver the most powerful systems in the market”.

In September, he detailed the company’s roadmap for AI computing platforms, with plans to release the world’s most powerful single system in Q4 2026 and then double the performance with a new launch a year later.

Looking ahead, talk of an AI bubble rages on not only in the US but globally, with downside risks coming from over-investment (possibly repeating Meta’s metaverse overreach) to a lack of electricity to power data centres in some locations.

From a global perspective, Kung noted the major source of uncertainty stems from geopolitical risks. “An unstable political and economic environment – such as export controls, bans or tariffs – could lead enterprises to adopt a more cautious stance, increasing the risk of tightened procurement spending and slower adoption.”

Ren earlier suggested the development of AI will take decades, even centuries, advising “don’t worry.”

Such a “long march” view leaves plenty of time for shares of AI companies to soar to new peaks and for the market to sustain a wrenching correction, and the two countries to continue to battle for the claim of AI champion.