Nvidia Spends $150 Billion a Year in Taiwan as Jensen Huang Calls It the “Epicentre of the AI Revolution” 

Sunita Somvanshi

Jensen Huang speaking at the Nvidia Constellation campus launch celebration at Beitou-Shilin Technology Park in Taipei, May 27, 2026
Giganectar · AI Infrastructure · May 27, 2026

Nvidia plants its flag in Taiwan —
$150 billion a year, one campus, the world’s AI supply chain

From chips and packaging to full AI supercomputers, Taiwan is now Nvidia’s largest operational hub. CEO Jensen Huang made it official on May 27, 2026, at the launch of Nvidia Constellation in Taipei.

On May 27, 2026, Nvidia CEO Jensen Huang stood at the T17 and T18 lots in Taipei’s Beitou-Shilin Technology Park and formally launched Nvidia Constellation — the company’s first overseas headquarters. Construction breaks ground this year. The campus is expected to be operational by 2030. Around 1,000 employees attended, along with Taipei Mayor Chiang Wan-an and Huang’s family. The mayor presented Huang with a key to the city and a handmade calligraphy scroll.

Huang, who was born in Tainan in 1963 and emigrated to the United States at age nine, has described Taiwan as the centre of Nvidia’s supply chain. The company’s annual Taiwan spend was around $10–15 billion four to five years ago. Huang said on May 27 it is heading toward $150 billion per year. As Nvidia has worked through tightening US-China chip trade rules, Taiwan’s manufacturing ecosystem has become more central to its operations. Nvidia became the first company to reach a $5 trillion market valuation in late 2025.

$150B Annual Taiwan spend (projected)
4,000 Employees planned at Constellation
~4 ha Campus footprint, Beitou-Shilin
2030 Target operational date

A closer look at the Taipei headquarters

📍

Location

T17 and T18 lots at Beitou-Shilin Technology Park, northern Taipei. The two parcels were combined by removing a 15-metre dividing road. Nvidia Iconic Taiwan Co. holds the 50-year lease, extendable by 20 years.

🏛️

Design & Scale

The design echoes Nvidia’s Santa Clara headquarters. The campus spans nearly 4 hectares and is planned to house roughly 4,000 employees — making it one of the largest AI R&D hubs in the APAC region once open.

🤝

Partner Ecosystem

TSMC, the world’s largest contract chipmaker, is Nvidia’s primary semiconductor manufacturer and a near neighbour. Key AI server partners — amid surging global AI infrastructure demand — include Foxconn, Wistron, and Quanta Computer.

Power Supply

Power remains an open question. Taiwan Power Co. is planning a new substation at the park. Taipei’s city government is in active talks with the utility over the energy policy needed to support AI workloads at this scale.

COMPUTEX 2026 · Best Choice Awards

Nvidia swept COMPUTEX 2026

At this year’s COMPUTEX Best Choice Awards, Nvidia won across four categories. Select a platform below to see what it does and how it performs.

Award
Golden Award + Sustainable Tech Special Award
Configuration
36 Vera CPUs + 72 Rubin GPUs per rack
Inference performance / watt
10× higher vs Blackwell generation
Cost per token
10× lower than previous generation
With Groq 3 LPX (trillion-param models)
35× higher throughput per watt
Cooling
100% liquid-cooled · 45°C operating temp
Assembly time per compute tray
2 hours → 5 minutes (cable-free modular tray design)
Power resilience
6× more onboard energy storage for power smoothing
Networking
6th-gen NVLink Switch (scale-up) + ConnectX-9 SuperNICs + Spectrum-X Ethernet Photonics switches (scale-out) + BlueField-4 DPUs for storage and security
Award
Golden Award — Edge AI & Robotics
GPU Architecture
NVIDIA Blackwell
AI Performance
2,070 FP4 teraflops
vs Jetson Orin
7.5× compute · 3.5× energy efficiency
Power range
Configurable 40–130 watts
Built for
Smart robots, industrial automation, medical devices, autonomous machines — already in production across hundreds of applications
Award
Vehicle Technology & Smart Cockpit Category Award
Platform type
Open platform for autonomous vehicle development
Models
Alpamayo 1 & 1.5 — 10-billion-parameter chain-of-thought reasoning vision-language-action models
Simulation tool
AlpaSim — open-source end-to-end AV simulation framework
Training data
NVIDIA Physical AI Open Datasets: 1,700+ hours across geographies and conditions
Key focus: long-tail AV scenarios
Ambiguous pedestrian hand signals · conflicting traffic lights and road markings · emergency vehicles parked partially in lane — cases outside standard training distributions

Performance gains at a glance

Relative improvements as reported by Nvidia. Bars show scale — not absolute values.

Inference performance per watt (Vera Rubin vs Blackwell) 10× improvement
Cost per token reduction (Vera Rubin vs Blackwell) 10× lower
Throughput per watt with Groq 3 LPX — trillion-param models 35× improvement
Jetson Thor compute vs Jetson Orin 7.5× higher
Jetson Thor energy efficiency vs Jetson Orin 3.5× better

Taiwan’s AI manufacturing cluster

The key partners in Nvidia’s Taiwan ecosystem are geographically concentrated — from TSMC’s Hsinchu fabs to the planned Constellation campus in northern Taipei. Click any pin for details.

How the campus came to be

Key milestones from land deal to launch celebration.

Late 2025
Land secured
Taipei City Government terminated the existing surface-rights contract for T17 and T18 lots at Beitou-Shilin Technology Park, freeing them for Nvidia’s use.
January 2026
Investment approved · Lease signed
Taiwan’s Ministry of Economic Affairs approved Nvidia’s NT$3.3 billion (approx. US$105 million) capital injection into Nvidia Iconic Taiwan Co. The city government signed a 50-year lease (extendable by 20 years) and merged the two lots by removing a 15-metre dividing road.
May 21, 2026
COMPUTEX Best Choice Awards
Nvidia received four COMPUTEX 2026 Best Choice Awards — for Vera Rubin NVL72 (two awards), Jetson Thor, and Alpamayo — ahead of Jensen Huang’s arrival in Taipei.
May 23, 2026
Jensen Huang arrives in Taipei
Huang landed at Songshan Airport and immediately visited the “Meet-a-Claw” developer event, where he demonstrated OpenClaw, an open-source AI agent platform. “The era of useful AI has arrived,” he told attendees.
May 26, 2026
TSMC dinner with C.C. Wei
Huang and TSMC CEO C.C. Wei dined together in Taipei — a visible sign of the decades-long partnership now powering the world’s AI factories.
May 27, 2026
Constellation launch celebration
~1,000 employees, Taipei Mayor Chiang Wan-an, and Huang’s family gathered at the T17/T18 site. The mayor gifted Huang a key to the city and a handmade calligraphy scroll. Huang announced $150 billion in annual Taiwan spend and the 4,000-employee campus plan.
June 1–5, 2026
GTC Taipei keynote at COMPUTEX
Huang scheduled to deliver the opening keynote at Taipei Music Center on June 1. GTC Taipei at COMPUTEX runs June 1–4, covering AI factories, agentic AI, physical AI, and robotics.
2026 → 2030
Construction through operations
Construction expected to begin mid-2026, pending a construction permit and Taipei’s urban design review. Target: operational by 2030.

Direct from the source

“Taiwan is the epicentre of the AI revolution. This is where the chips come, packaging comes, this is where the systems are made, this is where AI supercomputers were created.”

Jensen Huang — Constellation launch, Taipei, May 27, 2026

“Vera Rubin is the largest product launch, probably in the history of Taiwan. Each one of the Vera Rubin systems consists of almost 2 million parts, and it includes 150 different ecosystem partners here in Taiwan to build it.”

Jensen Huang — media remarks, Taipei Songshan Airport, May 23, 2026

“The world is watching NVIDIA shape the future of AI.”

Taipei Mayor Chiang Wan-an — Constellation launch celebration, May 27, 2026

“It’s become a really, very powerful assistant. The era of useful AI has arrived. That’s what this event is about, to show you what open source agents can do and then you can go create your own.”

Jensen Huang — Meet-a-Claw developer event, Taipei, May 23, 2026

AMD separately announced plans to invest more than $10 billion in Taiwan’s AI sector in the week before the Constellation launch, deepening its own partnerships with local manufacturers. Both announcements reflect the pace of global tech infrastructure investment — and Taiwan’s central role in it. Nvidia’s Vera Rubin NVL72 products are expected to ship in the second half of 2026 from providers including AWS, Google Cloud, Microsoft, and CoreWeave.

Last week, Nvidia told investors it had forecast over $1 trillion in sales for its flagship AI chips. Huang said on May 27 the company would be worth even more in three to five years. Consumer device makers — from audio hardware to smartphones — continue integrating AI advances at the chip and software level. The broader consumer technology industry continues absorbing AI hardware advances across devices and services. Nvidia’s Constellation campus, the Vera Rubin NVL72, Jetson Thor, and Alpamayo were covered here as part of the company’s May 2026 Taipei announcements. GTC Taipei at COMPUTEX runs June 1–4, 2026.

Nvidia Taiwan AI Infrastructure COMPUTEX 2026 Vera Rubin NVL72 Jensen Huang TSMC Jetson Thor

Leave a comment