NVIDIA's Huang: AI Buildout Still in Its Early Stages, $3 Trillion to $4 Trillion Spending by 2030
NVIDIA's CEO Jensen Huang emphasized that the AI buildout is still in its early stages during the Goldman Sachs Communacopia + Technology Conference 2026. He stated that the market is moving toward a new computing model, but acknowledged supply-chain limits, land shortages, and power constraints that could slow deployment.
Huang also reiterated his view that AI infrastructure spending could reach $3 trillion to $4 trillion by 2030, driven by generative computing and the end of Moore's Law. He highlighted NVIDIA's strong growth in its systems business, with month-over-month growth of 27% in Grace, Blackwell, and NVLink systems.
NVIDIA is seeing higher pricing across its systems, from $18,000 per Hopper GPU system to $40,000 for Vera Rubin, while some integrated systems now sell for $8.5 million. Huang described NVIDIA as both a growth and value stock, arguing that its compute systems are becoming asset-backed investments rather than simple technology purchases.
The company's customer base includes cloud service providers, regional cloud providers, enterprises, and AI-native companies, with partnerships in cybersecurity with CrowdStrike, Cisco, and Palantir. Huang also mentioned the next major growth areas for AI as cybersecurity and physical AI, including autonomous vehicles, robotics, industrial automation, and sensing in the physical world.
NVIDIA is expanding its presence globally, with a 2 gigawatt data center commitment in Australia representing $80 billion in infrastructure investment. Huang said that the company's long-lived architectures support its view as both a growth stock and a value stock, noting that Volta is still rented 10 years later and that all Ampere units are rented out globally.