BEIJING—As China’s tech behemoths like Alibaba, Tencent, and Baidu charge ahead in the artificial intelligence race, they’re harnessing explosive growth in cloud computing to fuel their ambitions. Yet, this surge comes at a cost: severe constraints on advanced AI chips due to U.S. export restrictions, coupled with mounting margin pressures from intense competition and high infrastructure costs. Industry insiders say these hurdles are reshaping strategies, forcing innovation in domestic alternatives while prioritizing internal needs over external services.

Recent earnings reports paint a vivid picture. Tencent Holdings Ltd., China’s largest social media and gaming company, reported a 15% quarterly revenue growth in its latest results, but executives highlighted chip shortages as a key bottleneck. ‘One constraint of cloud business growth is availability of AI chips because you know, when AI chips are actually in short supply, we actually prioritize internal use as opposed to renting it out externally,’ said a Tencent representative during the earnings call, as reported by Reuters.

The Cloud Computing Surge

Cloud services have become the backbone of China’s AI ecosystem, with demand skyrocketing as enterprises integrate generative AI into everything from e-commerce to autonomous driving. According to a report from the South China Morning Post, analysts at JPMorgan predict that firms like Tencent and Alibaba will lead AI growth into 2026 by embedding AI features into their core applications, benefiting downstream sectors. This integration is expected to drive cloud revenue, even as chip shortages persist.

Alibaba Cloud, a dominant player, has seen similar trends. The company is investing heavily in AI infrastructure, but faces the same supply issues. Baidu, another key giant, recently unveiled new AI processors—the M100 for inference and the M300 for training and inference—set for launch in 2026 and 2027, respectively, alongside upgraded Ernie AI models, per The Economic Times. These moves signal a push toward self-sufficiency amid external pressures.

Navigating Chip Constraints

The U.S. export controls on advanced semiconductors have created a ‘chip crunch’ that’s stoking prices and spurring panic buying, as detailed in a Reuters analysis. Chinese firms are turning to homegrown options like Huawei’s Ascend chips, which are gaining traction in data centers. Government intervention is ramping up, with officials overseeing the allocation of high-end AI chips and prioritizing domestic alternatives, according to Tom’s Hardware.

Posts on X (formerly Twitter) reflect industry sentiment, with users noting that GPU supply constraints are forcing Chinese AI companies to innovate. One post highlighted Kai-Fu Lee’s observation that these limitations enable training frontier models for $3 million versus $1 billion in the U.S., achieving inference costs at 1/30th the price. Another discussed an oversupply of Nvidia H100 chips in China despite sanctions, paired with popularity of Ascend chips for lower capex, as seen in various X discussions.

Margin Pressures Mount

While cloud growth is robust, profit margins are under siege. The capital-intensive nature of AI infrastructure—building massive data centers and securing power—combined with chip shortages, is squeezing profitability. A NextBigFuture report contrasts U.S. hyperscalers’ $400 billion capex with China’s efforts, noting that Beijing is building 250 public AI data centers amid an oversupply in some chip categories.

Analysts warn of ‘margin pressures’ as companies like Tencent prioritize internal AI chip use, limiting cloud rentals to outsiders. This internal focus, while strategic, caps revenue potential from external clients. Furthermore, the Economist predicts that in 2026, Chinese firms like Huawei and SMIC will advance in AI chip design despite restrictions, potentially alleviating some pressures but not without ongoing costs.

Innovation Amid Adversity

China’s response to these challenges is multifaceted. The government is mandating the use of domestic chips in new state-funded data centers, requiring the removal of foreign chips from projects under 30% completion, as reported in X posts and corroborated by industry sources. This pivot accelerates the adoption of chips from Cambricon and Huawei, with Baidu deploying a 30,000-card Kunlun P800 cluster for LLM training, according to AI Supremacy.

Cambricon Technologies reported a staggering 4,347% year-over-year revenue increase in H1 2025, driven by demand for its Siyuan 370 chips. Baidu’s Intelligent Cloud secured a 1 billion yuan contract from China Mobile, showcasing the viability of domestic alternatives. As one X post from a tech analyst noted, ‘China is ‘flushed’ with AI chips & building 250 public AI DC,’ highlighting how sanctions have spurred a parallel ecosystem.

Geopolitical and Economic Ripples

The broader implications extend beyond China. U.S. restrictions aim to curb Beijing’s AI advancements, but they’re inadvertently fostering innovation. A Merics report emphasizes China’s pursuit of self-reliance across the AI stack, from chips to models, viewing it as critical for national security. Europe could learn from this in its digital sovereignty efforts.

Comparatively, U.S. firms like OpenAI are pursuing ‘brute-force capital’ with massive power requirements—26GW for recent chip orders—while China focuses on efficiency, as covered in an X cover story reference to Caixin. This divergence could redefine global AI leadership, with China aiming to become the leader by 2030, per a RAND perspective.

Strategic Shifts in Play

Tech giants are adapting by vertically integrating. Huawei’s accelerators are retraining cloud providers away from Nvidia, as per IEEE Spectrum. Shortages are so acute that Beijing is intervening in distribution, prioritizing top AI developers, according to the Asia Economy and echoed in a Wall Street Journal piece by Lingling Wei: ‘Shortages of advanced AI chips are so acute that Beijing is intervening and tech companies are resorting to workarounds.’

Looking ahead, the World Economic Forum notes that China’s AI surge, from DeepSeek to MiniMax, stems from coordinated policy and chip innovation despite bans. As X users observe, this could lead to cost-effective models challenging U.S. dominance.

Power and Infrastructure Challenges

Beyond chips, power constraints loom large. China’s AI data centers require immense energy, with U.S. clusters far outpacing in scale. Recent X posts peg 2025 AI capital expenditure in China as significant but trailing the U.S.’s majority control of global compute. Yet, innovations like lower-cost training are closing gaps.

Industry statistics from SQ Magazine highlight market growth, with top manufacturers pushing performance benchmarks. For China’s giants, balancing cloud expansion with these constraints will define their trajectory, potentially surprising the world as forecasted by the Economist.