China may not overtake the United States across the full artificial-intelligence stack within the next five years, but the claim that it is simply “behind” is no longer accurate. Lin Junyang, technical lead of Alibaba’s Qwen team, reportedly estimated that Chinese companies had less than a 20% chance of surpassing leading U.S. firms such as OpenAI and Google DeepMind within three to five years. His assessment centered on the much larger computing resources available to U.S. companies.
That is one expert’s probability estimate—not a scientific consensus. Evidence available by March 2026 shows that Chinese and U.S. frontier models had repeatedly exchanged the lead, while the two countries remained sharply different in compute, capital, chips, research influence, and industrial deployment.
As an Amazon Associate I earn from qualifying purchases.
What the Alibaba scientist actually predicted
The forecast reported by the South China Morning Post came from Lin Junyang, a technical leader associated with Alibaba’s Qwen artificial-intelligence team.
Free tools Windows power users keep installed
One-click scans. No signup required.
Lin reportedly put the probability below 20% that a Chinese company would surpass leading U.S. companies within three to five years. Because the statement was reported in late 2025, that window points roughly to 2028–2030, rather than an indefinite prediction about China’s technological future.
#1 Best Overall
The forecast concerns Chinese companies exceeding frontier U.S. laboratories. It does not prove that China cannot lead in industrial AI, robotics, military applications, low-cost models, or national AI deployment. Nor does it establish that scientists generally agree with Lin.
Lin’s central rationale was that U.S. computational resources were one to two orders of magnitude larger. That figure should be treated as his assessment, not as an independently verified global census. It also does not mean that every U.S. laboratory has ten to 100 times the compute of every Chinese laboratory. It refers to the scale of resources available at the frontier or in aggregate.
“AI dominance” is not one thing
A country can lead in benchmark performance while another leads in manufacturing, robotics, patents, cost-efficient deployment, or international adoption. A useful comparison separates at least these categories:
Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Repair Windows errors before they cause bigger problemsFix Now →| Measure | What it asks |
|---|---|
| Frontier models | Which companies produce the strongest general-purpose systems? |
| Compute | Which country can train and operate the largest models at scale? |
| Semiconductors | Who controls advanced chips, manufacturing equipment, design tools, and supply chains? |
| Research and talent | Who produces influential research and attracts leading scientists? |
| Commercialization | Who converts AI into revenue, productivity, cloud services, and business adoption? |
| Industrial deployment | Who embeds AI most effectively in factories, vehicles, logistics, and infrastructure? |
| Global influence | Whose models, chips, standards, and cloud platforms are adopted internationally? |
These measures can point in different directions. A narrow U.S. lead in frontier-model quality would not automatically mean U.S. dominance in every application. Likewise, Chinese leadership in industrial deployment would not necessarily mean that China had built the world’s strongest general-purpose model.
The latest model evidence narrows the gap
The original forecast is more difficult to treat as a simple “China is behind” story after developments recorded in Stanford’s 2026 AI Index.
Stanford reported that U.S. and Chinese frontier models had traded the lead multiple times since early 2025. As of March 2026, Anthropic’s leading model held only a 2.7% advantage in the report’s composite comparison. Stanford still found that the United States produced more top-tier models, but China’s progress made the model-performance gap effectively closed by that measure.
The result is an important distinction: China may be behind in the quantity of frontier systems and the resources supporting them while being close to parity in the capabilities visible to users. Efficiency improvements, distillation, quantization, sparse architectures, better data, and training advances can allow a company to obtain more capability from constrained hardware.
Rank #2
DeepSeek’s rise is relevant for this reason. It demonstrated why raw compute totals do not necessarily settle the competition. It does not prove that compute no longer matters, or that its economics permanently transformed AI. It does show that efficiency can weaken the assumption that the largest budget always produces the decisive lead.
Why the United States may retain the advantage
Capital and infrastructure
Frontier AI requires spending on accelerators, data centers, electricity, cooling, networking, specialized employees, and software. Stanford reported U.S. private AI investment of $285.9 billion in 2025, compared with $12.4 billion in China. That is a substantial difference, although it is not a complete comparison because Chinese state-guidance funds, local-government financing, subsidized infrastructure, and strategic procurement are harder to capture in private-investment figures.
The United States also hosted 5,427 data centers, according to Stanford’s 2026 report—more than ten times the number in any other country. Data-center counts are not the same as AI-training capacity: they do not show how many facilities contain advanced accelerators, how much power is available, or how heavily the systems are used. They nevertheless indicate the scale of the U.S. infrastructure base.
A concentrated frontier ecosystem
The U.S. combines frontier laboratories such as OpenAI, Anthropic, Google DeepMind, Meta, and xAI with hyperscalers including Microsoft Azure, Amazon Web Services, and Google Cloud. It also has major chip-design and infrastructure companies, universities, venture-capital networks, and a large enterprise-software market.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Scan for outdated or missing drivers - takes under a minute3Repair Windows errors before they cause bigger problemsThese companies are not interchangeable. One may lead in model development, another in cloud distribution, another in chips, and another in consumer or enterprise adoption. Their combined ecosystem is more important than any single company’s ranking.
Access to the advanced-chip supply chain
Frontier AI depends on a multinational semiconductor system. U.S. firms are important in chip design and software; Taiwanese companies provide much of the leading-edge fabrication; Dutch and Japanese companies supply critical equipment and materials. This gives the United States and its partners considerable leverage, but it also means that U.S. strength is partly coalition-based rather than entirely domestic.
U.S. export controls have targeted China’s access to advanced computing chips and semiconductor-manufacturing technology. The Congressional Research Service describes these controls as a central policy tool while noting that their scope and implementation have changed over time. Controls may constrain China’s access, but they can also encourage domestic chip development, software substitution, and more efficient algorithms.
Talent and research networks
The United States continues to benefit from leading universities, private research laboratories, international recruitment, and concentrated frontier-industry teams. That advantage is not guaranteed. Immigration restrictions, research-security rules, and geopolitical tensions could make the country less attractive to scientists and engineers.
China has substantial research strength. Stanford reported that China’s share of the 100 most-cited AI papers increased from 33 in 2021 to 41 in 2024. That measures research influence, not necessarily commercial frontier capability, but it shows why publication and talent metrics cannot be dismissed.
Where China may already have an advantage
China’s strongest position may be in the conversion of AI into physical and industrial systems. Its manufacturing base, large domestic market, engineering workforce, state-directed financing, and ability to coordinate infrastructure can accelerate deployment.
Stanford says China leads the United States in publication volume, citations, patent output, and industrial-robot installations. Those indicators do not prove Chinese superiority in general-purpose models. They do show that a country can gain strategic value from AI without producing the single best chatbot.
Potential areas of Chinese strength include:
- industrial robotics and smart manufacturing;
- logistics and supply-chain optimization;
- electric vehicles and autonomous systems;
- drones and other machine-enabled platforms;
- public-sector and surveillance applications;
- low-cost or open-weight model distribution; and
- AI services designed for developing markets.
China may therefore become highly influential even if U.S. companies retain a narrow lead on the most compute-intensive general-purpose models.
Why open and inexpensive models matter
A proprietary model can be slightly better on a benchmark and still lose commercial influence if a competing model is much cheaper, easier to run locally, or available under more flexible terms.
For global users, the relevant questions may include:
- Which model delivers the lowest cost for a specific task?
- Can it run on locally controlled hardware?
- Where is user data stored?
- Does it work with local languages and regulations?
- Which company supplies the surrounding cloud and infrastructure?
- Will the model remain available if export controls or geopolitical tensions change?
Chinese companies could gain international reach through inexpensive models and infrastructure without first winning the absolute frontier. Conversely, U.S. companies could retain the model lead while losing portions of the global market if their services are expensive, restricted, or difficult to deploy.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.How to judge the forecast through 2030 and 2031
Frontier capability
Watch whether U.S. models remain ahead on broad, contamination-resistant evaluations, scientific work, coding, reasoning, and real-world tasks. Also compare the cost of achieving similar results. A small performance lead may matter less if Chinese systems deliver comparable results at substantially lower cost.
Compute and chips
The key questions are whether China can secure or manufacture enough advanced accelerators, whether domestic chips can approach leading performance and software compatibility, and whether efficiency improvements reduce the importance of raw hardware.
Infrastructure
Both countries must add data centers, electricity, cooling, and networking. China’s ability to build industrial infrastructure quickly may offset some chip constraints. U.S. permitting, grid, and power bottlenecks could slow the expansion of its theoretical advantage.
Capital
The U.S. lead in private financing matters, but China can compensate through state funding, strategic procurement, subsidized energy, and coordinated industrial policy. The more important question is not simply who spends more, but who turns spending into reliable capability and revenue.
Industrial diffusion
AI’s economic value will depend on adoption across factories, vehicles, logistics, health care, finance, defense, and government. A country that deploys slightly less capable systems everywhere may achieve more economic impact than one that develops the best model but struggles to integrate it.
The Tool Desk
Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →International reach
Global leadership will depend on exports of models, chips, cloud services, standards, and infrastructure. Export controls could slow China’s access to advanced technology, but they could also encourage other countries to build independent supply chains and split the global AI ecosystem.
Best Value
Four plausible outcomes
1. Competitive parity
The United States keeps a narrow frontier lead, while Chinese companies match or exceed U.S. offerings in cost, open models, industrial AI, and selected applications. This is consistent with the current evidence that the model gap is small but the broader ecosystem remains uneven.
2. A durable U.S. lead
The United States maintains its capital and compute advantages, continues attracting international talent, expands data-center capacity, and converts frontier models into dominant cloud and software platforms.
3. Chinese overtaking in strategic impact
China achieves breakthroughs in chips or algorithmic efficiency, deploys AI more effectively across manufacturing and robotics, and wins international adoption through cheaper models and infrastructure. It could become strategically dominant in important sectors without leading every model benchmark.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
4. Fragmentation
Neither country clearly dominates. U.S.-aligned and China-centered ecosystems develop different chips, models, cloud platforms, standards, and regulations. Companies and governments then choose between systems based on cost, security, political alignment, and data sovereignty.
The bottom line on China versus U.S. AI leadership
Lin Junyang’s less-than-20% estimate is a credible warning about the importance of computing resources, but it is not proof that China cannot overtake the United States. By March 2026, Stanford’s evidence showed near parity in frontier-model performance, even as the U.S. retained major advantages in private capital, data-center scale, top-tier model production, and the advanced-chip ecosystem.
The most defensible forecast is that the United States may retain the strongest overall frontier-AI position through 2030, while China remains capable of leading in industrial deployment, robotics, cost-efficient models, patents, and selected strategic applications. Whether that becomes “AI dominance” depends on the metric—and on how quickly chips, compute, capital, talent, and global distribution change.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
Recommended Free Tools
