China's AI ambitions put to the test: Why billions in investments are going to waste

More than just a chip war: The real reason why China's AI offensive is stalling

The People's Republic of China is pursuing its goal of becoming the world's leading artificial intelligence superpower by 2030 with dizzying determination. While official pronouncements conjure up a bright future in which 90 percent of the economy operates using AI and intelligent systems permeate every aspect of society, a far more complex picture is emerging behind the scenes. China's AI offensive is grappling with fundamental structural problems that extend far beyond the much-discussed American restrictions on chip exports. A talent gap of over five million skilled workers, a fragmented technological infrastructure, dramatic regional inequalities, and impending market consolidation pose existential challenges to Beijing's ambitious plans.

The parallels to Germany's energy transition problems are striking. Just as Germany risks failing its digital future due to a lack of grid capacity, China is grappling with a different kind of infrastructural imbalance. While data centers cannot be built in Frankfurt due to a lack of power connections, state-of-the-art facilities in western Chinese provinces stand largely empty because the downstream infrastructure, human capital, and practical demand are lacking. In both cases, a fundamental truth of modern technology policy is revealed: Gigantic investments in individual components are rendered ineffective if the overall system is not developed consistently.

Suitable for:

• China and the Neijuan of systematic overinvestment: State capitalism as a growth accelerator and structural trap

The talent trap

Perhaps the most critical weakness of China's AI strategy is the dramatic shortage of skilled workers. The Ministry of Human Resources and Social Security estimates the gap at over five million people, with a staggering supply-to-demand ratio of one to ten. In the first half of 2025, job postings for AI-related positions exploded by 37 percent compared to the same period of the previous year. Robotics engineers and algorithm developers were particularly in demand, with job postings for these positions increasing by over 50 percent. These figures do not document healthy expansion, but rather a desperate race for scarce resources.

McKinsey predicts that China's demand for AI professionals will rise to six million by 2030, while domestic universities and returning overseas Chinese can provide, at best, two million. This creates a gap of four million highly skilled workers, and it is likely to widen further as China's birth rate has been declining for years. The working-age population is projected by the UN to shrink by 180 million by 2050 compared to 2023, while the population ages rapidly. The average age of the workforce will rise to over 45. China is thus finding itself in a demographic bind between emerging economies like Vietnam and aging industrialized nations like Japan.

A superficial glance might lead one to assume that China has an abundance of graduates. Indeed, Chinese universities produce around 1.4 million STEM graduates annually. However, reality reveals a qualitative discrepancy. Truly cutting-edge research and the development of frontier models primarily require doctoral candidates. The output of AI-trained PhD students remains relatively low, leading to intense competition for the available top talent. Experienced machine learning scientists at tech giants now command seven-figure salaries in yuan. Smaller startups report that critical research and development positions remain vacant for months, massively delaying product development.

The problem is exacerbated by the specific nature of AI integration. Unlike the mobile revolution of the 2010s, when the core technologies were already functional and capital was primarily needed for user acquisition and logistics expansion, AI implementation requires continuous, context-specific research and development. A hospital cannot simply install ChatGPT and talk about AI-powered healthcare. It takes months or years of development to address medical workflows, regulatory compliance, and integration with existing systems. Without patient capital willing to fund these multi-year development cycles, most AI-plus projects stall before resolving the core implementation challenges.

The lack of interdisciplinary expertise is proving particularly problematic. A 2024 study by Renmin University found that China is suffering from a shortage of top talent, especially AI scientists and professionals with cross-industry expertise. Integrating AI into traditional industries requires individuals with both deep technical understanding and in-depth industry knowledge. An agricultural AI system needs developers who understand agronomy. A financial AI requires experts familiar with regulatory requirements. These interdisciplinary skills are scarce globally, but especially so in China.

Companies are responding with various strategies. Some are aggressively recruiting abroad, relaxing hukou restrictions, and trying to bring back talent from overseas. Others are investing heavily in internal training programs. The government is promoting the expansion of AI curricula at universities. Over five hundred Chinese universities have established AI degree programs since 2018. However, cultural and educational shifts take time. Even with accelerated efforts, the talent gap will burden the Chinese AI ecosystem for the next decade.

The geopolitical dimension further exacerbates the problem. While Chinese universities are making substantial progress in AI education, global tech hubs continue to attract top talent. Uncertainty stemming from government regulation, ideological control, and perceived limitations on academic freedom prompts some talent to migrate abroad or remain there. Although China boasts 47 percent of the world's leading AI researchers and 50 percent of AI patents, these impressive figures cannot mask the fact that the sheer scale of demand far exceeds any available resources.

Infrastructure crisis despite massive investments

China's AI infrastructure presents a paradox of monumental proportions. On the one hand, the country announced or built over 250 new artificial intelligence data centers between 2023 and 2024. Public and private investors pumped billions into expanding the digital backbone infrastructure. On the other hand, local sources report that up to 80 percent of this newly created computing capacity remains unused. The utilization rates of many smart data centers languish at 20 to 30 percent. Facilities that cost billions are largely idle, while their operators desperately search for customers and the ongoing cooling, electricity, and maintenance costs strain their balance sheets.

This bizarre situation results from a combination of political pressure, speculative excess, and fundamental miscalculations. Following the bursting of the housing bubble and the COVID-induced economic downturn, local governments desperately sought new drivers of growth. The enthusiasm surrounding ChatGPT in late 2022 made AI appear as the ideal candidate. By 2023, over 500 data center projects were proposed nationwide. Local authorities aggressively promoted these initiatives, hoping to boost their regional economies. State-owned enterprises, government-affiliated investment funds, as well as private firms and investors enthusiastically embraced the supposedly golden future.

However, as is typical with rushed projects, realistic planning was often lacking. Many facilities were built without regard for actual demand or technical standards. Engineers with relevant experience were scarce, and numerous executives relied on intermediaries who inflated forecasts or exploited procurement processes to secure subsidies. As a consequence, many new data centers fell short of expectations, being expensive to operate, difficult to fill, and technically irrelevant for modern AI workloads.

Suitable for:

• AI boom in China or is the AI ​​bubble bursting now? Hundreds of new data centers are empty

A key problem lies in the type of infrastructure built. Many data centers were designed for training large language models and accordingly located in the western provinces with their cheaper energy. This aligned with the Eastern Data Western Computing Initiative, which aimed to shift data processing from the congested metropolitan areas in the east to the resource-rich regions in the west. However, when demand shifted from pure model training to inference—the practical application of trained models—many western facilities proved to be poorly positioned. Inference typically requires different hardware configurations—faster, more responsive chips that prioritize low latency and efficiency over sheer computing power. Furthermore, inference needs to take place close to end users, i.e., in the large cities in the east. Thus, western data centers are often built for the wrong tasks and located in the wrong places.

In response, Beijing announced the construction of an inference-focused data center in Wuhu, a southeastern prefecture, to serve major urban markets like Shanghai, Hangzhou, and Nanjing. But this is just a drop in the ocean. The misallocation of resources to unsuitable infrastructure has tied up billions in capital that could have been used more productively elsewhere. Some projects apparently never intended to generate profit through actual computing power. Several reports and insiders confirm that some companies used AI data centers to qualify for government-subsidized green energy or land deals. In some cases, earmarked electricity was sold back to the grid while the buildings remained unused. By the end of 2024, most players in the business were aiming to benefit from policy incentives rather than from genuine AI work.

Hardware shortages are further exacerbating the situation. Despite massive government support for domestic chip development, Chinese AI companies remain heavily reliant on foreign technology. The US controls over 70 percent of global computing power and uses export controls to restrict China's access to advanced chips like Nvidia's H100 and critical packaging technologies. China's AI chip supply gap is projected to exceed $10 billion by 2025. Domestic alternatives like Huawei's Ascend 910B lag behind in performance for training large language models. Moreover, advanced AI clusters require not just chips, but highly engineered interconnects spanning tens of thousands of processors. US firms continue to lead in system-level design.

Chinese companies purchased nearly one million Nvidia HGX H20 processors in 2024 alone. This dependency persists because Nvidia's supply scale and mature CUDA software stack create a chicken-and-egg problem for China's AI industry. Domestic hardware lacks both volume and developer support. DeepSeek attempted to train its R2 model on Huawei's Ascend chips but had to resort to Nvidia hardware due to performance instability, weaker interconnects, and CANN's immaturity. Even if Chinese manufacturers could flood the market with Ascend NPUs or Moore Threads GPUs, a weak software stack makes them unattractive to developers.

The software ecosystem for Chinese AI chips is significantly weaker than its Western counterpart. Nvidia's CUDA benefits from over fifteen years of documentation and refinement, a large user base, and robust integration with popular machine learning frameworks like PyTorch and TensorFlow. Huawei's CANN framework was only introduced in 2019, twelve years after CUDA. Developers often describe it as buggy, unstable, and poorly documented, with frequent runtime crashes and limited third-party integration. These issues don't make large-scale training runs on Chinese hardware impossible, but they do make them considerably more expensive.

The lack of common standards among various Chinese chip vendors further fragments the market. Each vendor has its own incompatible low-level software stack. Mainstream AI frameworks primarily support Nvidia chips. Domestic AI chips must adapt to multiple frameworks, and each framework upgrade requires repeated adaptation. This leads to missing operators and optimizations for large models, preventing models from running or rendering them inefficient, precision discrepancies due to architectural and software implementation differences, and high porting costs to enable large-scale model training on domestic chips.

The Model-Chip Ecosystem Innovation Alliance, founded in the summer of 2025, is attempting to address this problem. It unites Huawei, Biren Technologies, Enflame, Moore Threads, and others with the goal of building a fully localized AI stack that connects hardware, models, and infrastructure. Success hinges on achieving interoperability through shared protocols and frameworks and reducing ecosystem fragmentation. While unifying low-level software may be challenging due to differing architectures, mid-level standardization appears more realistic. By focusing on common APIs and model formats, the group hopes to make models portable across domestic platforms. Developers could write code once and run it on any Chinese accelerator. However, until these standards truly exist, fragmentation means that each company must tackle multiple problems simultaneously on multiple fronts in a saturated market.

Huawei made CANN open source in early August 2025, possibly as part of its commitment to the new alliance or as a general attempt to make its Ascend 910 series the platform of choice among Chinese-based companies. Until then, Huawei's AI toolkit for Ascend NPUs was distributed in a limited form. CANN's maturity lags behind CUDA, primarily because there was no broad, stable installed base of Ascend processors outside of Huawei's own projects. Developers follow scale, and CUDA became dominant because millions of Nvidia GPUs had shipped and were widely available, justifying investments in tuning, libraries, and community support. Huawei and other Chinese developers cannot ship millions of Ascend NPUs or Biren GPUs due to US sanctions.

The energy infrastructure presents a mixed picture. China has expanded its grid eighty times faster than the US and is a world leader in solar, wind, and hydropower capacity. These massive investments in renewable energies are intended to make AI scaling sustainable. The Eastern Data Western Computing Initiative is shifting data processing to energy-rich and land-rich western regions, powered by wind and solar energy. The goal is not only to reduce costs but also to create a more robust and sustainable infrastructure. Millions of IT racks are expected to be installed by the end of the fourteenth Five-Year Plan in 2025.

While western regions offer abundant wind and solar resources and lower electricity prices, they often lag behind in infrastructure development. The challenge lies in efficiently combining the abundant green energy resources in the less developed western regions with the growing data processing needs in the east. Computing needs are concentrated in the eastern regions, where renewable energy self-sufficiency is below 40 percent, while the west boasts 70 percent of China's installed renewable energy capacity. Tencent plans to locate its largest smart data center in western China in Ningxia, partly due to the lower electricity prices. Companies tend to train their large-scale language models in western provinces because of lower electricity costs, but base their application-oriented data centers in the east, where a larger customer base allows for faster feedback on their applications.

While Western regions offer low electricity costs, deficiencies in transportation, communication, and talent support systems make it difficult to attract and retain high-tech personnel. Many Western data centers remain idle while awaiting a boom in downstream applications. A cloud vendor employee confirmed that the utilization rate of Chinese smart data centers is below 30 percent.

Our China expertise in business development, sales and marketing - Image: Xpert.Digital

Industry focus: B2B, digitalization (from AI to XR), mechanical engineering, logistics, renewable energies and industry

More about it here:

• Xpert Business Hub

A topic hub with insights and expertise:

• Knowledge platform on the global and regional economy, innovation and industry-specific trends
• Collection of analyses, impulses and background information from our focus areas
• A place for expertise and information on current developments in business and technology
• Topic hub for companies that want to learn about markets, digitalization and industry innovations

Regional division exacerbates the divide

The geographical disparities in China's AI development replicate and exacerbate existing economic inequalities. East Coast provinces such as Guangdong, Jiangsu, Zhejiang, and Shanghai have long held leading positions, with Guangdong demonstrating particularly strong development momentum. Shanghai and Beijing have maintained a high concentration of AI activities, thanks to political support and technological research and development capabilities. Central regions such as Hubei, Henan, and Shandong have gradually shifted to a middle range, indicating steady improvement. However, western provinces such as Qinghai, Tibet, and Gansu remain at a low level overall. Despite some improvements, the gap with the eastern region is still evident, and the problem of unbalanced regional development persists.

From 2014 to 2022, the level of AI in China showed a significant trend of improvement and regional expansion over time. In 2014, the overall level of AI development in the country was low, with only the eastern coastal provinces showing outstanding performance and demonstrating these regions' early advantages in AI. Meanwhile, the central and western regions had a late start overall, and their level of development was generally low. By 2022, the country's AI level had improved considerably, with the Yangtze River Delta and the Bohai Margin becoming the core drivers of growth. Beijing, Tianjin, and Hebei showed strong development momentum, while the western region, although at a lower level of development, exhibited a clear upward trend.

A study on income inequality due to AI found that the impact of AI on income inequality is strongest in the northeastern region, followed by the western region, while the effects are relatively smaller in the central and eastern regions. AI significantly exacerbates the income gap through industrial structural improvements and technological innovation. The regional heterogeneity shows that AI does not act as an equalizer but rather amplifies existing advantages. Provinces with strong digital infrastructure, access to capital, and talent pools benefit disproportionately, while underdeveloped regions fall further behind.

The urban-rural digital divide further exacerbates these disparities. Despite recent government efforts to accelerate the development of digital rural infrastructure within the context of pursuing rural revitalization in China, based on successes in poverty reduction, the problem of the digital divide persists. In terms of financial investment, the funds allocated to rural digital infrastructure lag significantly behind those allocated to urban areas. According to data, China's fiscal and social investments in agricultural and rural informatization at the county level amount to only thirteen million yuan and thirty million yuan, respectively, resulting in an overall informatization development level of just thirty-seven point nine percent.

There is a significant disparity in hardware deployment between rural and urban areas, encompassing variations in digital resources, infrastructure, network equipment, and base stations. In 2022, China reached a milestone of 2.3 million 5G base stations nationwide. However, the number of rural 5G base stations lags considerably behind the national average, further widening the digital divide. At the same time, the goal of providing equivalent network coverage and speed in both rural and urban areas has not yet been fully achieved.

During the COVID-19 pandemic, the disparity in hardware infrastructure development became even more pronounced. A striking example involves a Tibetan college student living in Linzhou, in the Tibet Autonomous Region, who was forced to ride a motorbike for twenty minutes to the foot of a mountain and then climb to the summit in freezing temperatures to attend online classes. This anecdote highlights the stark imbalance in digital hardware development between rural and urban areas.

The lack of data centers at the county and municipal levels, which are essential for maintaining efficient digital application systems, hinders the progress of generative AI technologies in rural areas. This situation is akin to the proverb, "Even the most skilled housewife cannot cook without rice," highlighting the fundamental need for these data centers to advance rural digital development.

From the perspective of software organizations that constitute the “soft power” of rural digital development, rural digital software suffers from deficiencies in digital competence, talent acquisition, and governance compared to urban areas. On the one hand, influenced by traditional, self-interested mindsets prevalent in smallholder farming communities and exacerbated by the inherent lag in rural digital progress, there is a notable lack of enthusiasm among the rural population to actively engage with generative AI services for the revitalization of rural China. Furthermore, the substantial migration of the rural workforce, which results in the elderly, vulnerable individuals, women, and children forming the primary labor force in rural areas, intensifies the phenomena of rural depopulation, depopulation, and population aging, impacting the rural population, economy, society, and overall development.

A survey conducted in rural areas that have not yet implemented electronic governance of village affairs revealed that 84.13 percent of village officials cited “the high proportion of elderly villagers, which hinders technology adoption” as the primary obstacle. These combined factors significantly impede the adoption and promotion of generative AI technologies in rural regions.

Regional disparities are also evident in the AI ​​index. A recent study developed a comprehensive artificial intelligence index with seven primary dimensions, designed for provincial-level and industry-specific analysis. The China-US comparison shows that, under a unified framework, the US composite score exceeds the Chinese score of 59.4 by 68.1. Breaking down China into seven main areas to create a sub-national index reveals stark regional disparities in China's AI development: the northern, eastern, and southern regions lead in composite scores, while central and western regions lag significantly behind, highlighting the effects of regional concentration of innovation and industrial resources.

This geographical fragmentation has far-reaching consequences. It creates different speeds of economic transformation, with leading regions rapidly advancing to knowledge-based economies, while lagging regions remain stuck in traditional manufacturing and agriculture. It exacerbates social tensions as income disparities between regions widen. It complicates national coordination, as different provinces have varying levels of development and priorities. And it creates inefficient resource allocation, with state-of-the-art data centers sitting idle in remote western provinces while eastern metropolises struggle for capacity.

Suitable for:

• China's economy in the crisis? Structural challenges of a growth nation

The overcapacity crisis and the pressure to consolidate

The enthusiastic construction boom of 2023 and 2024 has confronted China with a dramatic overcapacity crisis. Over 500 data center projects were proposed in 2023 alone, with at least 150 expected to be operational by the end of 2024. This development reflects a familiar pattern in China's economic development. When the central government prioritizes a sector as strategic, local authorities and companies rush into it with excessive zeal, often disregarding actual need or rational planning. The result is regularly overinvestment, overcapacity, and a painful consolidation phase.

The automotive industry offers an instructive parallel project. Around 140 companies compete in this sector, with only a few profitable, and a third experiencing capacity utilization rates below 20 percent. To prevent local job losses, regional governments nevertheless help even struggling suppliers stay afloat through subsidies and other forms of support. Market consolidation has therefore slowed, price wars have erupted, and producers are under pressure to increase exports to more profitable markets. Meanwhile, the era of easily accessible export markets is fading. The US banned almost all Chinese vehicle imports on national security grounds under the Biden administration, and the EU imposed tariffs on Chinese electric vehicles last year.

AI infrastructure follows a similar trajectory. The National Development and Reform Commission intervened with stricter regulations. New projects must now meet specific utilization criteria and submit purchase agreements before receiving approval. Additionally, local authorities are prohibited from initiating small-scale computing infrastructure unless they can provide a clear economic justification. Government procurement reached 24.5 billion yuan, roughly 3.4 billion dollars, in 2024 alone, with an additional 12.4 billion yuan earmarked for 2025. Yet, despite robust government investment, reported utilization rates remain between 20 and 30 percent, compromising both economic viability and energy efficiency.

Over the past eighteen months, more than 100 projects have been abandoned, a significant increase compared to just 11 in 2023. This dramatic rise in canceled projects signals a reality check. Investors and operators are realizing that many of these facilities will never become profitable. The initial crisis, fueled by the hype surrounding generative AI following ChatGPT's launch in late 2022, has morphed into a profitability crisis. GPU leasing markets have collapsed. Facilities that cost billions of dollars now sit underutilized, yields are plummeting, and many facilities have become obsolete before they were even fully operational due to changing market conditions.

In July 2025, President Xi Jinping explicitly warned against overinvestment in AI, reiterating his earlier concerns about excessive local government investment. The comments underscore policymakers' desire to avoid a repeat of the overcapacity seen in other emerging industries, such as electric vehicles, which contributed to deflationary pressures. While the state planner did not specify which part of the sector requires restraint, investment has been particularly pronounced globally in the construction of data centers that underpin AI development. A slowdown in this expansion would impact suppliers of chips, networking equipment, and other essential server components, from Cambricon Technologies Corp. to Lenovo Group Ltd. and Huawei Technologies Co.

On August 29, 2025, the State Council emphasized the need to ensure “the orderly flow of talent, capital, and other resources.” Zhang Kailin, an official with the National Development and Reform Commission, told reporters at a briefing that the government would encourage provinces to develop AI in a coordinated and complementary manner. The goal is to leverage their unique strengths to promote growth without duplicating efforts. “We will decisively avoid disorderly competition or a ‘follow-the-crowd’ approach,” Zhang said. Development should be based on local advantages, resources, and industrial foundations.

The software market reflects similar consolidation dynamics. The Cyberspace Administration of China approved a list of over 180 major language models for general use by August 2024, illustrating the wide range of Chinese tech companies vying for domestic market share. These firms are competing not only for a piece of the market but also for funding amid an economic slowdown and a downturn in China's venture capital industry. Workshop participants emphasized that while many Chinese startups have attracted investment from large tech companies like Alibaba and Tencent, many investors remain skeptical about the ability of AI startups to generate revenue in the short term. In their search for economically productive investments, many Chinese venture capital firms are looking to diversify their risk through resource pooling, suggesting a more dispersed funding environment.

Given both funding and hardware constraints for Chinese AI developers, participants suggested that China could succeed in advancing a few firms or AI labs through resource pooling, but these efforts must be selective and targeted, reducing the likelihood of substantial returns. Ultimately, participants suggested that this environment is likely to lead to increased industry consolidation in China's AI market.

Du Hai, a senior manager at Baidu's cloud division, predicted that this will drive market consolidation. The dozen or so domestic AI chip companies currently active will likely dwindle to three or four distinct camps. “The winners will be those whose chips can support the broadest range of models—or enable a killer app that becomes the de facto standard.”

Gartner predicts that by 2029, the GenAI technology landscape will consolidate into 75 percent fewer players as hyperscalers and SaaS platform providers expand and hybrid cloud providers absorb. This is not market speculation, but the inevitable consequence of economic forces already reshaping the industry. The parallels to historical infrastructure developments are striking. Gartner identifies that we are moving from a period of vendor fragmentation to consolidation through acquisitions and market disruptions. Just as the electricity industry evolved from thousands of local generators to a handful of large utilities, AI is following the same path.

Venture capital funding for Chinese AI startups fell by nearly 50 percent year-over-year in early 2025, reflecting broader investor caution amid sluggish growth, regulatory uncertainties, and geopolitical tensions. In the second quarter alone, funding plummeted to just $4.7 billion, its lowest level in a decade. This investor fear was partly fueled by the Chinese government's demonstrated willingness to stifle frontier innovation in the name of redoubling measures to preserve ideological purity.

The rest of the Chinese market, while offering some mixed signals, provides further cause for pessimism. The real estate sector has collapsed, its youth unemployment rate exceeds 17 percent, and consumer confidence is declining. The geopolitical situation doesn't help either, with export controls still impacting China's tech sector, tariffs threatening the broader economy, and ideologically driven, control-focused policies deterring most investors. This funding crisis poses a particular problem for AI deployment. Without patient capital willing to fund these multi-year development cycles, most AI-plus projects will stall before addressing core implementation issues.

Benefit from Xpert.Digital's extensive, fivefold expertise in a comprehensive service package | R&D, XR, PR & Digital Visibility Optimization - Image: Xpert.Digital

Xpert.Digital has in-depth knowledge of various industries. This allows us to develop tailor-made strategies that are tailored precisely to the requirements and challenges of your specific market segment. By continually analyzing market trends and following industry developments, we can act with foresight and offer innovative solutions. Through the combination of experience and knowledge, we generate added value and give our customers a decisive competitive advantage.

More about it here:

• Use the 5x expertise of Xpert.Digital in one package - starting at just €500/month

Future scenarios between euphoria and disillusionment

The range of future projections for China's AI industry could hardly be wider. Optimistic voices like Morgan Stanley predict that Chinese AI investments could break even by 2028 and generate a 52 percent return on invested capital by 2030. The core AI industry could become a market worth 140 billion dollars by 2030. This estimate jumps to 1.4 trillion dollars when related sectors such as infrastructure and component suppliers are included. AI could provide an additional boost to China's long-term GDP growth, offsetting factors such as its aging workforce and slowing productivity growth. Over the next two to three years, AI could add an extra 0.2 to 0.3 percentage points to China's annual growth.

The global market for humanoid robots could reach five trillion dollars by 2050, with one billion units in use, and thirty percent of those in China. China's efficiency-driven and low-cost approach creates a different path to return on investment. The cost advantage demonstrated by companies like DeepSeek—developing influential models for as little as five point six million dollars—could enable Chinese firms to penetrate global markets that cannot afford or are unwilling to adopt Western solutions.

The next six to twelve months will be a critical period for Chinese AI companies, as an increasing number of enterprise implementations attempting to solve real-world problems will begin to demonstrate productivity gains. In the long term, humanoids, or human-like robots powered by AI, could be widely used for industrial, commercial, and household purposes. Over the longer term, the AI ​​revolution will translate into a productivity boost by increasing efficiency, streamlining production processes, and unlocking new products, services, and jobs.

The Asia-Pacific region will account for 33 percent of AI software revenues in 2025, but as China ramps up its involvement in the AI ​​race with the United States, analysts expect the region to represent 47 percent of the market by 2030. Forecasts indicate that China alone will account for two-thirds of total AI software revenues in the Asia-Pacific region, amounting to 149.5 billion dollars, by 2030. This significant growth projection for the AI ​​market is driven by the following industry-shaping trends.

But these optimistic projections stand alongside dire warnings. Capital Economics predicts that the AI-driven stock market bubble will burst in 2026. The research firm said that rising interest rates and higher inflation will push stock valuations down. From 2026 onward, these stock market gains should unwind predictively, as higher interest rates and increased inflation begin to drive down stock valuations. Ultimately, they anticipate that returns from stocks will be poorer over the next decade than over the previous one. And they think that the long-standing outperformance of the US stock market may be coming to an end.

The International Monetary Fund noted that while a downturn is plausible, it is unlikely to develop into a systemic crisis that would devastate the US or global economy. Gourinchas observed that, similar to past trends, the hype surrounding a groundbreaking technology may not meet market expectations in the short term, potentially leading to a decline in stock prices. However, he noted that, unlike in 1999, the current investment landscape is characterized by cash-rich technology companies rather than debt-driven ones.

Forrester predicts that by 2026, AI will lose its luster, trading its tiara for a hard hat. Enterprise ROI concerns will outweigh vendor hyperbole. With this market correction, companies will prioritize function over flair. CFOs will be drawn into more AI deals. Firms will spread their bets across agent ecosystems and reallocate talent as AI agents take over grunt work. Smart companies will invest in AI governance and AI fluency training to mitigate risk and slowly map their AI journey.

A Bain report estimates that by 2030, global capital expenditures for AI data centers will reach $500 billion annually, requiring 200 GW of additional power capacity—half of it in the US. But the AI ​​sector needs to generate $2 trillion in annual revenue to justify the outlay. Currently, there's an $800 billion gap. One executive said China's AI chip sector still faces hurdles in demand and foundry capacity. The market needs real-world applications to scale. It's application demand that will determine everything. The American style of desperately expanding computing power is not the choice for Chinese companies.

China's AI infrastructure boom is faltering, as the country built hundreds of data centers to support its AI ambitions, according to MIT Technology Review, but many now sit idle. Billions were invested by both state-owned and private entities in 2023 and 2024, with the expectation that demand for GPU leases would continue to grow, but adoption has actually declined, and as a result, many operators are now struggling to survive. Local publications report that up to 80 percent of this new computing capacity remains idle.

These diverging future scenarios reflect fundamental uncertainties. Will China overcome its software ecosystem fragmentation? Can domestic chip manufacturers close technological gaps quickly enough? Will US export controls tighten, loosen, or remain at their current levels? Will the Chinese government intensify its ideological control, thereby discouraging innovators, or will it pursue a more pragmatic policy? Will global demand for low-cost AI solutions favor Chinese efficiency-focused approaches, or will concerns about quality and trust favor Western solutions?

The answers to these questions will not only determine China's fate but also shape the global AI landscape. Three possible scenarios are emerging. The first scenario sees the US maintain its dominance. With control over advanced chips and the world's leading AI companies, Washington retains its technological leadership, while China struggles with computing limitations and has restricted access to key markets. The second scenario depicts a split AI development into two competing ecosystems. One is led by the US and its allies, prioritizing transparency and ethical standards, while the other is dominated by China, where state-controlled AI serves as a tool for digital surveillance. Countries will be forced to align themselves with one of these models, creating a fragmented digital landscape.

The third scenario sees China dominating consumer AI but falling behind in high-end applications. US chip restrictions hamper China's ability to develop cutting-edge AI for defense and scientific research, yet Beijing excels in mass-market AI, offering affordable platforms like DeepSeek to global users. However, this balance could shift dramatically if China were to pursue its ambitions in Taiwan, home to TSMC, which manufactures roughly 90 percent of the world's most advanced chips.

Ultimately, the race for AI supremacy is reshaping global power dynamics. While the US currently leads in advanced AI research, China's strategic focus and state-driven investment have made it a formidable competitor. Although Beijing faces hurdles such as Western restrictions and market skepticism, its progress in consumer AI and influence in emerging markets keep the race unpredictable. Whether this competition leads to continued US dominance, a divided digital landscape, or China's rise in critical sectors, one thing is clear: AI will profoundly shape the global economy, national security policies, and interpolitical alliances in the coming years.

Suitable for:

• Beijing's new five-year plan and massive investment program: How China is challenging the global economic order

Implementation problems and governance deficits

Beyond hardware and personnel issues, China grapples with fundamental implementation challenges that are often overlooked. The adoption of AI in businesses remains fragmented and experimental. While China is a leader in the adoption of generative AI, Chinese organizations have not yet implemented it as fully as they could. When SAS surveyed Düber about the extent to which their organizations are using generative AI, nineteen percent of Chinese organizations said they “use and have fully implemented generative AI,” which is ahead of the global average of eleven percent but lags behind the world leader in full implementation, the US, at twenty-four percent.

Meanwhile, 64 percent of respondents from China said their organization “uses generative AI but has not yet fully implemented it,” which is well above the global average of 43 percent. Given China’s emphasis on careful regulation and authorized approval of generative AI, it makes sense that many organizations are conducting initial tests before fully integrating generative AI into their processes. It is clear that China is fully committed to generative AI, but Chinese organizations are proceeding cautiously, even as they collectively embrace this new technology.

When asked about implementation challenges, Chinese respondents were far less likely than the global average to cite a lack of internal expertise or adequate tools: only 31 percent said they lacked the right tools to implement generative AI, compared to 47 percent globally, while only 21 percent said they lacked internal expertise, compared to 39 percent globally. These figures stand in stark contrast to the previously discussed talent gaps and suggest a discrepancy between self-perception and reality, or differing standards for what constitutes “adequate expertise.”

Data privacy and data security ranked as the top two concerns among all survey respondents regarding the implementation of generative AI, cited by 76 and 75 percent respectively. However, more than half of the respondents (51 percent) expressed concerns about the need for internal talent and skills. Governance and monitoring training was found to be particularly inadequate. According to SAS, less than one in ten respondents (7 percent) reported a “high” level of governance and monitoring training for generative AI. Thirty-two percent reported an “adequate” level, while 58 percent—a clear majority—said their governance and monitoring training was “minimal.”

When asked about their organizational governance frameworks for generative AI, only five percent of respondents said they had a “well-established and comprehensive” governance framework. More than 55 percent said their governance framework was “under development,” while 28 percent described it as “ad hoc or informal.” Approximately one in 11 percent said their generative AI governance framework was “non-existent.” These governance gaps create substantial risks for implementations, particularly in regulated industries or with sensitive applications.

Fragmented data flows across industries hinder the ability to consolidate data into a coherent, accessible resource pool for AI applications. These data silos prevent effective AI model training and limit insights across sectors. Government agents and businesses are working to improve data interoperability and promote cross-industry data sharing and structured, cross-border data circulation under under-regulated frameworks to unlock the full value of China's data ecosystem. By addressing these data-related challenges, China can further strengthen its AI ecosystem while contributing to a more coherent and innovative global data landscape.

The implementation of generative AI is also insufficiently integrated with rural governance. As a leading force in emerging technologies, generative AI will further complicate the existing diverse interest structure in empowering rural revitalization in China. For the government, which holds a prominent position, the digital divide stemming from urban-rural economic disparities requires substantial investments in labor, resources, and finance to bridge this gap. This process is characterized by an extended return on investment timeline. Unlike the market, which prioritizes economic factors alone, government-led rural governance involves a holistic evaluation of multifaceted governance costs.

Technology developers and suppliers primarily interact with government departments. Consequently, their offerings are largely tailored to meet government requirements, potentially neglecting the genuine development needs of rural areas and their residents. This exacerbates the fluid nature of digital governance. At the national level, despite the issuance of legal documents such as the Action Plan for the Development of Digital Villages 2022-2025 and the Interim Measures for the Management of Generative Artificial Intelligence Services, the involvement of numerous departments can lead to blurred lines of responsibility, causing delays and reducing governance effectiveness. Unless these issues are addressed swiftly, they will not only hinder the activation of rural residents' intrinsic motivation to actively participate in generative AI-driven rural revitalization in China, but could also generate new digital conflicts.

The great AI consolidation: Only a few Chinese models will survive.

China's pursuit of AI leadership by 2030 faces a complex mix of structural challenges that extend far beyond the oft-cited chip export restrictions. The talent gap of over five million skilled workers, the fragmented infrastructure with dramatically unused capacity, the massive regional disparities between urban centers and rural peripheries, and the looming market consolidation after years of speculative overinvestment paint a picture that is considerably more sobering than official pronouncements suggest.

This paradoxical situation is particularly evident in data centers: While Frankfurt cannot build new facilities due to a lack of electricity, state-of-the-art facilities in China's western provinces stand largely empty because of a lack of downstream infrastructure, human capital, and practical demand. In both cases, it becomes clear that gigantic investments in individual components are wasted if the overall system is not developed consistently.

The next 18 to 36 months will be crucial. Either China succeeds in overcoming fragmentation through initiatives like the Model-Chip Ecosystem Innovation Alliance, closing the talent gap through massive investments in education, and intelligently utilizing existing but underutilized capacity. Or the nation watches as investments migrate, top talent leaves, and digital value creation moves elsewhere. The coming market consolidation will be brutal. Of the more than 180 major language models currently approved, perhaps only three or four will survive. Hundreds of data centers will have to close or be repurposed. Venture capital funding remains at its lowest level in a decade.

But it would be premature to dismiss China's ambitions. Its efficiency-focused strategy, deployment-first approach, and the cost advantages of solutions like DeepSeek could capture significant market share in global markets that cannot afford high-end Western solutions. Government support remains robust, even if it needs to become more coordinated and less wasteful. And demographic challenges—an aging population and a shrinking working-age population—make AI-driven productivity gains not optional, but essential.

Global observers should neither underestimate China nor take its official pronouncements at face value. As is so often the case, reality lies somewhere between these extremes. China will neither rise to become an unassailable AI hegemon nor sink into technological insignificance. Instead, a complex, fragmented picture is emerging: regionally concentrated clusters of excellence on the east coast, experimental implementations in thousands of companies, spectacular failures in overambitious infrastructure projects, innovative efficiency solutions for specific use cases, and continued dependence on foreign technology coupled with accelerated efforts toward self-sufficiency.

When the final assessment is made in 2030, it is likely that neither the most optimistic nor the most pessimistic predictions will have come true. China will have made significant progress, but will not have achieved the dominant position that Beijing seeks. The US will continue to lead in frontier research, but Chinese solutions will be ubiquitous in emerging economies. And the world will have to operate with two partly separate, partly intertwined AI ecosystems, whose coexistence, competition, and occasional cooperation will shape the geopolitical landscape of the twenty-first century.

☑️ Our business language is English or German

☑️ NEW: Correspondence in your national language!

Konrad Wolfenstein

I would be happy to serve you and my team as a personal advisor.

You can contact me by filling out the contact form or simply call me on +49 89 89 674 804 (Munich) . My email address is: wolfenstein ∂ xpert.digital

I'm looking forward to our joint project.

☑️ SME support in strategy, consulting, planning and implementation

☑️ Creation or realignment of the digital strategy and digitalization

☑️ Expansion and optimization of international sales processes

☑️ Global & Digital B2B trading platforms

☑️ Pioneer Business Development / Marketing / PR / Trade Fairs

A new dimension of digital transformation with 'Managed AI' (Artificial Intelligence) – Platform & B2B Solution | Xpert Consulting - Image: Xpert.Digital

Here you will learn how your company can implement customized AI solutions quickly, securely, and without high entry barriers.

A Managed AI Platform is your all-round, worry-free package for artificial intelligence. Instead of dealing with complex technology, expensive infrastructure, and lengthy development processes, you receive a turnkey solution tailored to your needs from a specialized partner – often within a few days.

The key benefits at a glance:

⚡ Fast implementation: From idea to operational application in days, not months. We deliver practical solutions that create immediate value.

🔒 Maximum data security: Your sensitive data remains with you. We guarantee secure and compliant processing without sharing data with third parties.

💸 No financial risk: You only pay for results. High upfront investments in hardware, software, or personnel are completely eliminated.

🎯 Focus on your core business: Concentrate on what you do best. We handle the entire technical implementation, operation, and maintenance of your AI solution.

📈 Future-proof & Scalable: Your AI grows with you. We ensure ongoing optimization and scalability, and flexibly adapt the models to new requirements.

More about it here:

• The Managed AI Solution - Industrial AI Services: The key to competitiveness in the services, industrial and mechanical engineering sectors