Japan could gain one of its largest AI data centers under a project in Akita now expected to cost more than ¥2 trillion, or roughly $12.6 billion, with Abu Dhabi sovereign investor Mubadala considering an investment.

If completed, the 300–500 MW project would add substantial AI compute capacity outside Japan’s established data center hubs.

For enterprises planning future Japan-based workloads, however, the useful capacity will depend on whether organizers can secure financing, grid connections, and enough deliverable power to support the proposed scale.

A ¥2 trillion project could expand Japan’s AI capacity map

Japan Today reported that expected construction costs now exceed ¥2 trillion and that Mubadala Investment Co. is considering backing the project. No final investment commitment has been announced.

The financial scope being discussed is considerably larger than earlier figures. JETRO reported in November 2025 that planned investment could reach ¥400 billion.

The figures are not directly comparable because the earlier number described expected investment while the new figure is reported as construction cost. Still, the latest estimate establishes Akita as a much larger infrastructure project than its initial public outline suggested.

Akita City signed a cooperation agreement with S2 and Bitgrit in October 2025. Organizers say they have secured 50 hectares in a renewable-energy industrial park in northern Akita City and are targeting full operation around 2033.

That scale could broaden where enterprises source Japan-based AI infrastructure. Most current capacity remains concentrated around established markets, although providers are expanding, including Alibaba Cloud with its fifth Japan data center.

Akita would offer a different proposition: large-scale AI infrastructure built around access to regional renewable generation rather than proximity to an existing metropolitan data center cluster.

Power and financing will decide how much capacity arrives

The project’s official site describes 300–500 MW of renewable-energy potential, including offshore wind. That does not mean the data center has already secured 500 MW of usable IT load or the grid connections required to deliver it.

That distinction directly affects future customers. A planned 500 MW campus does not create 500 MW of bookable compute capacity if electricity connections, substations, or generation cannot arrive on the same timetable.

Similar constraints are already affecting data center expansion elsewhere. US grid capacity constraints are slowing some proposed projects, while Australia is preparing for additional pressure from AI data center demand.

Akita therefore remains a long-term capacity prospect rather than infrastructure enterprises can include in near-term deployment plans. A Mubadala investment could strengthen the financing side, but customers will also need evidence of grid interconnection, construction milestones, operating partners, and committed capacity before the project becomes a realistic sourcing option.

Also read: Uptime Institute found 57% of major data center outages now cost more than $100,000, even as overall outage frequency declines.

Humanoid robots are no longer just dancing for the cameras; they are starting to clock in for work.

Smart Analytics Global (SAG) estimates that global humanoid robot shipments reached about 19,100 units in the first half of 2026, up 272% from a year earlier.

The growth is also reshaping the market’s leadership. Shanghai-based AGIBOT shipped about 8,400 robots, giving it 44% of global shipments and pushing it ahead of Unitree Robotics, which shipped about 5,900 units for a 31% share. Together, the two companies accounted for roughly three-quarters of worldwide shipments.

China’s lead extends beyond its biggest manufacturers. Chinese vendors accounted for more than 97% of global humanoid robot shipments during the period, while Chinese customers represented more than 85% of global demand, SAG said.

SAG expects worldwide shipments to approach 60,000 units for all of 2026, with industry revenue reaching about $1.6 billion. It forecasts shipments could reach 500,000 units by 2030.

Robots are moving into the workplace

The most important change may not be the number of robots sold, but where they are being used.

Industrial and commercial applications represented more than 70% of shipments in the first half, compared with about 50% a year earlier, SAG said. Manufacturing, logistics, warehousing and other controlled environments are emerging as the industry’s main route to commercial scale.

“Global humanoid robot shipments nearly quadrupled in 1H 2026, but an equally important shift is taking place in how these robots are being deployed,” Linda Sui, founder and principal at SAG, said in the report.

AGIBOT’s broader product lineup has helped it target industrial and commercial customers, while research, education and performance applications continue to represent a meaningful portion of Unitree’s deployments, according to SAG. Other Chinese manufacturers are also pursuing wheeled designs, which may prioritize stability and cost in structured workplaces.

China’s manufacturing advantage

China’s dominance reflects more than government support. The country has a large manufacturing base, domestic suppliers and a huge pool of industrial environments where robots can be tested and refined.

That combination could create a powerful feedback loop: more robots deployed in factories and warehouses can generate more operational experience, while domestic manufacturing can help companies improve designs and lower costs. But shipment numbers alone do not prove that humanoid robots are ready for widespread use. SAG said challenges remain around training data, AI models, reliability, manipulation accuracy, safety and cost.

The volume race has limits

The industry’s biggest question is whether rapid hardware production can keep pace with improvements in the software and AI controlling these machines.

SAG does not expect general-purpose humanoid robots for household tasks to reach meaningful mass-market scale within the next five years. More human-like robots designed for interaction and companionship also face questions around cost, consumer acceptance and regulation.

Geopolitics adds another complication. Recent US restrictions generally prevent new foreign-produced advanced robotic devices from receiving the FCC authorization needed for importation or sale. Previously authorized models remain unaffected, and conditional approvals are possible, but the rules could still make U.S. expansion harder for Chinese manufacturers.

For now, China’s shipment lead is undeniable. But the next stage of competition will depend less on how many robots leave factories and more on whether those machines can reliably perform useful work at an economically viable cost.

Read more: Chinese EV makers are racing Tesla to mass-produce humanoid robots, further expanding China’s lead in the global robotics industry.

Singapore has raised its 2026 economic growth forecast to 4.5%–5.5% after stronger-than-expected global AI investment helped lift semiconductor demand, exports, and manufacturing output.

The upgrade turns the AI spending boom into something more concrete than rising tech valuations. For hardware manufacturers and suppliers across APAC, Singapore is showing how global AI capital spending can feed through into national growth. It also leaves those businesses more exposed if that spending cools.

AI demand is lifting Singapore’s exports and manufacturing

The new forecast is Singapore’s second upgrade this year. The Ministry of Trade and Industry raised its outlook from 1%–3% to 2%–4% in February before moving it to 4.5%–5.5% in August.

Singapore’s economy grew 5.9% year over year in the second quarter, pushing first-half growth to 6.1%. MTI said global AI investment had been stronger than expected, supporting producers and exporters of AI-related products such as semiconductors.

That connection is already visible in trade data. Singapore’s non-oil domestic exports grew 27.4% year over year in the second quarter, after rising 9.6% in the first.

Electronics exports grew particularly quickly as demand increased for integrated circuits, storage products, and other hardware used in AI systems. The same demand pressure is driving investment elsewhere in the supply chain, including SK hynix’s $38 billion memory expansion.

Singapore also sits close to the infrastructure side of the spending cycle. APAC operators are already dealing with larger AI data center fleets and tighter power constraints, while Singapore-based DayOne raised $2 billion earlier this year for its AI infrastructure expansion.

For technology suppliers, continued AI investment supports demand for chips, storage, semiconductor equipment, and the trade services surrounding them. Singapore’s revised forecast provides a measurable indication of how far that spending is reaching beyond the companies building AI models and data centers.

Singapore’s gains also reveal an AI concentration risk

The stronger AI contribution also increases the downside if investment slows.

The ASEAN+3 Macroeconomic Research Office estimates that roughly half of global AI-related trade passes through ASEAN+3 and that AI-linked exports generated around two-thirds of the region’s export growth in the first quarter.

AMRO raised its 2026 regional growth forecast to 4.1%, citing stronger AI-related demand. But it also modeled what happens if the cycle weakens: if global AI investment growth falls back to its 2024 pace, ASEAN+3 growth could slow to 2.5% in 2027, its weakest rate outside the pandemic years since the Asian Financial Crisis.

For manufacturers, semiconductor suppliers, and enterprise procurement teams, that makes global AI capital expenditure a useful signal alongside Singapore’s headline GDP numbers. Capacity plans built around today’s electronics growth still depend heavily on continued investment in the infrastructure and hardware supporting AI workloads.

Singapore’s upgraded forecast captures both sides of the cycle: AI investment is already strong enough to lift exports, manufacturing, and national growth, while greater dependence on that demand increases the exposure if spending changes direction.

Also read: Japan’s planned Akita AI data center is now estimated at $12.6 billion, with Mubadala considering an investment in the renewable-powered project.