Japan builds world’s first national AI factory with Nvidia, backed by $6 billion government investment

Japan builds world’s first national AI factory with Nvidia, backed by $6 billion government investment

A consortium of SoftBank, Sony, Honda, and NEC is deploying 27,500 Nvidia Rubin GPUs to power Japan's sovereign AI ambitions and chase a $133 billion robotics market.

Japan just did something no other country has done: commit to building a national-scale AI factory from scratch, powered entirely by Nvidia’s latest silicon. The facility, announced on July 16, will pack 13,750 Vera CPUs and 27,500 Rubin GPUs into roughly 382 NVL72 rack configurations, drawing 140 megawatts of data center capacity.

The project is backed by 1 trillion yen, approximately $6 billion, in Japanese government funding over five years.

Inside the factory and the consortium behind it

The entity building this thing is Noetra Corp., a company established in January 2026 by SoftBank, Sony, Honda, and NEC.

Noetra will operate the facility as part of Japan’s METI-backed FRONTia Project. The initiative’s goal is to develop open multimodal foundation models, the kind of AI systems that can process text, images, video, and sensor data simultaneously, and then share those pretrained models with domestic developers across manufacturing, logistics, and healthcare.

Advertisement

Construction is expected to begin in 2027, with operations targeting a June 2028 start date.

Nvidia CEO Jensen Huang framed the partnership around Japan’s deep manufacturing heritage. The argument is straightforward: a country that already dominates precision manufacturing and robotics hardware should also own the AI software layer that makes those machines intelligent.

Why sovereign AI infrastructure matters for markets

The strategic context is Japan’s ambition to capture over 30% of the global AI robotics market by 2040, a market estimated at $133 billion.

For Nvidia, the Vera Rubin platform, which combines the company’s newest CPU and GPU architectures, gets its flagship deployment not from a Silicon Valley cloud provider but from a national government.

The competitive landscape and what investors should watch

For investors in the semiconductor supply chain, a single order for 27,500 Rubin GPUs and 13,750 Vera CPUs represents significant demand for TSMC’s advanced packaging, for high-bandwidth memory from SK Hynix and Samsung, and for the power infrastructure companies building out 140-megawatt facilities.

The consortium structure is also worth watching. SoftBank brings capital and telecom infrastructure. Sony brings sensor technology and entertainment AI applications. Honda brings autonomous driving and robotics expertise. NEC brings enterprise IT and government contracting relationships.

The risk, as always with government-backed megaprojects, is execution. A June 2028 operational target means the facility needs to be online before the Rubin architecture is even fully mature in the market. Supply chain delays, construction timelines, and the sheer complexity of deploying 382 NVL72 racks in a greenfield facility could push that date.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.

Japan builds world’s first national AI factory with Nvidia, backed by $6 billion government investment

Japan builds world’s first national AI factory with Nvidia, backed by $6 billion government investment

A consortium of SoftBank, Sony, Honda, and NEC is deploying 27,500 Nvidia Rubin GPUs to power Japan's sovereign AI ambitions and chase a $133 billion robotics market.

Japan just did something no other country has done: commit to building a national-scale AI factory from scratch, powered entirely by Nvidia’s latest silicon. The facility, announced on July 16, will pack 13,750 Vera CPUs and 27,500 Rubin GPUs into roughly 382 NVL72 rack configurations, drawing 140 megawatts of data center capacity.

The project is backed by 1 trillion yen, approximately $6 billion, in Japanese government funding over five years.

Inside the factory and the consortium behind it

The entity building this thing is Noetra Corp., a company established in January 2026 by SoftBank, Sony, Honda, and NEC.

Noetra will operate the facility as part of Japan’s METI-backed FRONTia Project. The initiative’s goal is to develop open multimodal foundation models, the kind of AI systems that can process text, images, video, and sensor data simultaneously, and then share those pretrained models with domestic developers across manufacturing, logistics, and healthcare.

Advertisement

Construction is expected to begin in 2027, with operations targeting a June 2028 start date.

Nvidia CEO Jensen Huang framed the partnership around Japan’s deep manufacturing heritage. The argument is straightforward: a country that already dominates precision manufacturing and robotics hardware should also own the AI software layer that makes those machines intelligent.

Why sovereign AI infrastructure matters for markets

The strategic context is Japan’s ambition to capture over 30% of the global AI robotics market by 2040, a market estimated at $133 billion.

For Nvidia, the Vera Rubin platform, which combines the company’s newest CPU and GPU architectures, gets its flagship deployment not from a Silicon Valley cloud provider but from a national government.

The competitive landscape and what investors should watch

For investors in the semiconductor supply chain, a single order for 27,500 Rubin GPUs and 13,750 Vera CPUs represents significant demand for TSMC’s advanced packaging, for high-bandwidth memory from SK Hynix and Samsung, and for the power infrastructure companies building out 140-megawatt facilities.

The consortium structure is also worth watching. SoftBank brings capital and telecom infrastructure. Sony brings sensor technology and entertainment AI applications. Honda brings autonomous driving and robotics expertise. NEC brings enterprise IT and government contracting relationships.

The risk, as always with government-backed megaprojects, is execution. A June 2028 operational target means the facility needs to be online before the Rubin architecture is even fully mature in the market. Supply chain delays, construction timelines, and the sheer complexity of deploying 382 NVL72 racks in a greenfield facility could push that date.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.