A "Declaration from Australia" by McKinsey
In April 2026, McKinsey published a research report titled "Australia's AI moment: Building Asia–Pacific's compute hub," systematically analyzing Australia's strategic opportunity in the Asia-Pacific AI computing power race.
The report opens with a striking statement: global computing demand will grow at least 3.5 times by 2030 (McKinsey, 2026). Against this backdrop, Australia has the potential to become the digital infrastructure hub of the Asia-Pacific region — but this requires up to A$190 billion in new capital investment to scale its compute capacity from 1.5 GW to 5.0 GW.
Although the report was written for Australian government and businesses, it reveals an important trend for China, which is also navigating the computing power race: the Asia-Pacific compute landscape is being redrawn, and each economy is searching for its own niche.
Insight 1: The "Spillover Effect" in Asia-Pacific Compute — Who Is Capturing U.S. Demand?
The report reveals an interesting dynamic: major global data center markets (the U.S., Europe, Singapore) are facing bottlenecks in power supply, land availability, and approval timelines, causing compute demand to "spill over" to neighboring regions.
The spillover capture landscape in Asia-Pacific has already taken shape:
- Malaysia: Received over US$20 billion in North American hyperscaler investments in the first 10 months of 2024 (Johor region especially benefiting)
- Thailand: Approved approximately US$10 billion in data center investment applications in the first half of 2025
- Indonesia: Benefiting from spillover demand from Singapore
- India, Malaysia: Have already launched targeted fiscal and regional incentive measures
Australia is competing with these faster-moving, lower-cost regional rivals.
💡 Implications for China: The Asia-Pacific compute "spillover effect" is essentially a slice of the global tech supply chain restructuring. When North American compute capacity becomes constrained, Asia-Pacific countries are all scrambling for this "spillover pie." China's own compute infrastructure faces similar constraints in energy, land, and approval speed — this isn't just Australia's problem, it's China's too.
Insight 2: Australia's "Comparative Advantage" — How Is It Different from China?
The report lists Australia's competitive advantages: ample land, political stability, abundant renewable energy potential, and a geopolitical position as a "trusted technology partner."
But Australia also faces significant disadvantages:
This comparison shows that: China has policy concentration and cost advantages in computing infrastructure, but lags in geopolitical trust and international connectivity.
Insight 3: Compute Landscape Determines AI Competitive Landscape
The report projects that by 2030, over half of global data center workloads will be AI-related — whether training or inference.
This reflects a changing demand dynamic for AI computing power:
- AI training: Requires extremely high-density compute, with the most demanding energy and cooling requirements; can be "offshored" to the most cost-effective regions
- AI inference: Requires low latency and must be deployed close to users
This means the AI compute market will naturally stratify: training data centers chase "cost arbitrage," while inference data centers chase "market proximity."
The implications for China are twofold:
1. On the training side: China's domestic AI chip ecosystem (Ascend, Cambricon, etc.) is still catching up with NVIDIA. If export controls continue, the "relative cost" of training compute may be higher than in other Asia-Pacific regions
2. On the inference side: China's massive domestic market of 1.4 billion people means enormous demand for localized inference compute — an advantage few other countries can replicate
Insight 4: Australia's Computing Dilemma Is Also a Wake-Up Call for China
The report highlights a critical data point: Australia's GDP growth has been declining since 2016, with per capita GDP growing at an average of just 0.6% annually from 2016 to 2024, and negative growth since 2020. The A$80 billion annual GDP boost and 100,000 jobs from data center construction are seen as a key lever to "repair productivity growth."
China faces similar structural challenges:
- After the real estate engine stalled, a new growth pole hasn't fully taken shape yet
- AI and computing infrastructure are seen as the core drivers of "new infrastructure"
- But like Australia, China faces a situation where "investment has gone down, but productivity data hasn't shown significant improvement yet"
This is the same global phenomenon behind the Fortune report's CEOs complaining that "AI isn't working." Computing power is infrastructure, not productivity itself. Only when computing power + organizational change + talent development all come together can the true productivity dividend of AI be unlocked.
Implications for Chinese Enterprises
1. Southeast Asia is not a "cost arbitrage" zone but a "compute battlefield": Southeast Asian countries are rolling out data center incentive policies one after another. This means for Chinese companies going global — don't view Southeast Asia as just a cheap manufacturing base, but as a growing, high-value digital infrastructure market
2. The strategic value of "East Data, West Compute" is globally validated: One of the core conclusions of the Australia compute report — location determines compute competitiveness — is precisely the underlying logic of China's "East Data, West Compute" strategy. Over three years, western compute hubs in Inner Mongolia, Guizhou, and Gansu are showing initial results
3. Localized compute deployment is a must for global companies: Chinese tech companies going overseas (TikTok, Shein, miHoYo, etc.) need to deploy inference compute abroad to serve local users. Asia-Pacific countries are competing not just for data center investment, but for the "digital foundation" of these enterprise ecosystems
Summary
McKinsey's report tells us: The Asia-Pacific computing race has begun, and every country is finding its place.
>
For China, the greatest advantages are scale (inference demand from 1.4 billion people), policy concentration (infrastructure efficiency through a unified large market), and a complete manufacturing ecosystem (data center hardware supply chain).
>
The greatest challenges are: maintaining the growth rate of training compute competitiveness under export controls, and translating compute investment into real AI productivity — rather than becoming another "impressive but idle" infrastructure project.
References
- McKinsey (2026.04). Australia's AI moment: Building Asia–Pacific's compute hub. Infrastructure Practice & TMT Practice.
- McKinsey (2026). Australia's productivity challenge: Restoring private sector investment.
- Stanford HAI (2026). AI Index 2026 Report.
Some comparative analysis is based on comprehensive assessment. [Further verification needed for specific comparative data on data center approval efficiency between China and Australia.]