Everyone Is Watching China. Singapore and India Are in Just as Many AI Deals.

We counted the AI deals we tracked in 2026 so far: 379 of them, 374 with at least one company we could place in a country. Then we counted which countries actually show up.

China appears in 38. Singapore appears in 37. India, 33. Japan, 29. South Korea, 22.

If your mental model of Asian AI is "China, and then a gap", the data does not support it. Asia shows up in 144 of those 374 deals, roughly 38 percent, and China is barely a quarter of that.

## The gravity well is still American

Before the contrarian read runs away with itself: the United States is in 305 of 374 deals, about 82 percent. That is not a close race. Almost every significant AI transaction still has an American company on one side of it, usually holding the capital or the chips.

What changed is the other side of the table. The counterparty is increasingly not in San Francisco.

## Why Singapore's number has an asterisk

Singapore at 37 deserves honesty rather than a headline. A large share of that is domicile of convenience. Regional funds, holding companies and Southeast Asian startups incorporate in Singapore for tax and legal reasons, so the flag on the deal is Singaporean while the engineers, customers and revenue sit in Jakarta, Ho Chi Minh City or Bangalore.

That does not make the number meaningless. It makes it a different fact than it first appears. Singapore is functioning as the clearing house for Southeast Asian AI capital, the way Delaware does for American corporate structure. When a number is that high for that reason, the interesting question is what it is a proxy for: a regional market that has grown big enough to need a financial hub.

India at 33 has no such asterisk, and that is the more striking figure.

## What the deals are actually about

The texture matters more than the count. The Asian deals we logged this month are not chatbot rounds. Tata Consultancy Services is opening a Bengaluru industrial AI lab on Nvidia platforms for manufacturing. Nvidia and Fujitsu are pulling Japan's robot makers together for a physical AI push. Mitsubishi Heavy is building cooling and power for large AI data centers. Apple is routing Apple Intelligence in China through Alibaba's Qwen and Baidu.

Factories, robots, power, and distribution into a market you cannot enter alone. Western AI capital is concentrated in models and the compute to train them. Asian AI capital is disproportionately going into the places where AI touches something physical or regulated.

That is a market-structure difference, not a sophistication gap. It reflects who owns the underlying industry. Japan has the robot makers. India has the services firms and the engineers. China has the manufacturing base and a regulator that decides who gets to sell there.

## The mistake this frame causes

The US-versus-China story is a story about frontier models, and on frontier models it is roughly right. But the frontier is one segment. If you allocate attention only there, you will keep being surprised by things that were legible in the deal flow months earlier: that Apple needed a Chinese partner to ship in China, that Japan's robotics incumbents would be an Nvidia distribution channel, that the interesting industrial AI work has a Bengaluru address.

You can browse the underlying records yourself on our deal tracker or in the live deal graph.

One caveat on our own numbers, because you should hold us to the same standard: these are counts of deals, not dollars. We count each country once per deal, so a deal with a US investor and a Japanese recipient counts for both, and the country shares sum to more than 100 percent. This is the set we track, sourced from news coverage, not a census of global AI dealmaking. Monthly slices are small enough to be noisy. The annual shape is the part we would defend.

## What to do with this

**Stop reading "Asia AI" as "China AI".** When you see an Asia number in someone's deck, ask which countries are in it and whether Singapore is carrying the count as a domicile rather than as a market. That single question will separate the analysts who counted from the ones who assumed.

**Watch the industrial counterparties, not the labs.** The next batch of Asian AI deals worth your attention will most likely be signed by a manufacturer, a services firm or a utility, not by a frontier lab. Those are the companies that own the thing AI is being pointed at. When Mitsubishi Heavy shows up in AI infrastructure, that is the signal, and it will not be in the model-release coverage.