NVIDIA
NVIDIA is the "utility" of global AI compute — GPUs are the hardware foundation for all AI training and inference, from LLMs to CV. As a Southeast Asian hub and financial-routing centre, Singapore **contributes about 15% of NVIDIA's global revenue (~USD 2.7 billion per quarter)** — a number that gives Singapore a special place in NVIDIA's global strategy.
📖 What it is
NVIDIA in Singapore:
- Compute supply: ships AI GPUs such as H100 / B200 via OEM channels, cloud providers (AWS, Google Cloud, Azure), and direct sales
- NSCC collaboration: parts of Singapore's National Supercomputing Centre's GPU clusters are NVIDIA hardware
- Enterprise market: AI deployments in finance, telecoms, and government rely heavily on NVIDIA GPUs
- DGX SuperPOD: several local enterprises have deployed DGX clusters
Why Singapore accounts for 15% of NVIDIA revenue: partly real Singaporean demand, and partly Singapore's role as a Southeast Asian routing / billing centre — invoiced here on paper, but the GPUs end up in other Southeast Asian countries. NVIDIA's reporting convention makes the Singapore figure look enormous.
🤖 Relation to AI
NVIDIA does not do AI model research directly, but in the AI era it is the absolute monopolist of compute — no serious AI training or inference happens without NVIDIA GPUs.
Singapore's AI ecosystem depends on NVIDIA across:
- SEA-LION training: H100 clusters
- Enterprise AI inference: LLM deployments in local finance and telecoms
- National compute: NSCC upgrades cannot avoid NVIDIA
This dependence has become sensitive in the US–China AI competition — US export controls on China (the A100 / H100 bans) have turned "how to obtain NVIDIA compute" into a geopolitical question. As both a US ally and a Southeast Asian hub, Singapore can currently buy top-end NVIDIA GPUs freely, while simultaneously being suspected as a "China rerouting channel". This is the most sensitive part of the 2024–2026 NVIDIA-in-Singapore narrative.
🇸🇬 Relation to Singapore
In Singapore's AI strategy, NVIDIA is the "compute chokepoint" — both an enabler and a geopolitical risk.
In the "seven transmission levers" framework:
- Lever 1 (infrastructure): the physical foundation of compute
- Lever 6 (international): how to position itself amid US–China compute controls
A take: Singapore's "15% of NVIDIA revenue" figure is a double-edged sword — on one hand it proves Singapore is an APAC AI hub; on the other it sharpens US suspicions of "GPU rerouting" via Singapore. Starting in 2024 the US has investigated whether Singapore is reselling controlled GPUs to Chinese entities — a real geopolitical risk for Singapore's AI strategy.
Worth watching: whether US GPU export controls on Singapore will tighten, whether local projects like SEA-LION can secure stable GPU supply, and the hardware procurement strategy of the national compute centre.
🗓️ Key Milestones
- 2023Singapore-billed revenue reaches ~15% of NVIDIA global revenue
- 2024US investigates suspected GPU rerouting via Singapore