Venture and private capital activity can offer a useful view of where investors see momentum building. Looking across the 2026 activities of 15+ Asia Pacific-focused investors, the pattern goes beyond “everyone is investing in AI”. That is already well established. More interesting is where within AI, capital is concentrating.
The same patterns emerge when we widen the lens to major global funds. The trends are not limited to Asia Pacific but demonstrate a broader area of interest for AI investment globally.
1. Physical AI is Attracting Growing Investor Interest
One clear theme across the investor activity is growing interest in AI that operates in the physical world. SoftBank is negotiating a majority stake in humanoid-robotics company 1X Technologies at a valuation of around USD 6 billion and has made more than 20 physical-AI investments. Hillhouse and its early-stage arm GL Ventures backed three Chinese embodied-AI companies – TARS, Tianji Intelligent and Kunlun Robotics – and participated in 18 embodied-AI funding rounds across 14 companies in the first half of 2026. HongShan backed Genesis AI’s robotics foundation models, while Granite Asia funded MindOn, a “robot brain” designed to transfer intelligence across humanoid platforms. Square Peg backed a physics-based foundation model for CNC manufacturing, while AirTree invested in autonomous navigation and robotics data infrastructure. East Ventures also backed Thai physical-AI and robotics startup Amity Robotics.
These deals span multiple investors and markets, including China, Australia, Thailand and Southeast Asia. Investor interest is expanding beyond software-only AI, towards robotics, autonomous systems, physical-world models and the infrastructure needed to support them.
The trend is also visible among major global funds. Tiger Global was an early investor in 1X Technologies, while Index Ventures backed Enigma, a robotics company focused on making robots easier to control. NEA led the USD 50 million Series A for P-1 AI’s Archie, an agentic AI engineer focused on mechanical, electrical, thermal and fluid-systems engineering. Sequoia’s 2026 strategy also identifies robotics as one of its priority categories, alongside agents and AI infrastructure.
2. AI Infrastructure is a Major Investment Category
Some of the largest AI-related investments are going into the infrastructure needed to run AI at scale. Rather than focusing only on foundation models, investors are putting substantial capital into power, data centres, semiconductors and the systems that connect them.
SoftBank-backed SB Energy is developing almost 9GW of contracted data-centre capacity. EQT’s AI Infrastructure strategy, seeded by EdgeConneX, reached USD 9.4 billion in fee-generating assets within three months of launch and plans more than 10GW of additional data-centre capacity across the region. Temasek has also increased its exposure to AI infrastructure, including energy, data centres and semiconductors, and joined the AI Infrastructure Partnership alongside BlackRock, Global Infrastructure Partners and MGX.
Investment is also reaching more specialised infrastructure. Peak XV backed Indian semiconductor startup C2i for its work addressing power-conversion losses in AI data centres, while Vertex Ventures backed Malaysian chip-design startup GreatAsic and Singapore’s Acrab, which is developing AI silicon and inference infrastructure.
| Layer | Example deals |
|---|---|
| Energy | SB Energy (9GW), EQT/EdgeConneX (10GW+) |
| Compute & silicon | GreatAsic, Acrab, C2i |
| Institutional capital | Temasek + AIP, EQT’s USD 15.6B fund |
Investors are looking beyond the model layer to the physical and technical capacity required to support AI growth. For large infrastructure and private-capital investors, energy, data centres and semiconductors offer exposure to the underlying demand for AI without relying on any single model or application winning the market.
The same interest is visible globally. a16z launched a dedicated USD 1.1 billion Machine Age Fund covering chips, memory, networking, storage, data centres and robotics, while Tiger Global, Sequoia and a16z backed AI-chip company Etched in a USD 700 million round.
3. AI Agents Are Emerging as a Distinct Investment Theme
Growing investment is also visible in agentic AI that is designed to take actions and complete tasks, rather than simply respond to prompts.
HongShan backed Genspark, an AI-agent company founded by former Baidu executives and now valued at more than USD 1 billion, and was an early investor in Manus, the general-purpose AI agent later acquired by Meta for more than USD 2 billion. Vertex Ventures backed Acrab, which develops agent orchestration infrastructure, and Singapore’s ReN3, which focuses on enterprise agentic AI with air-gapped and sovereign deployment options for regulated industries. Square Peg backed Octen, an LLM-native search infrastructure layer designed for AI agents and research assistants, as well as Byron, an AI-agent platform for business tax preparation. Peak XV backed Ringg AI, an enterprise voice-AI company in Bengaluru.
The investment pattern is notable because it spans both agents and the infrastructure around them. Capital is going into agent companies themselves, but also into orchestration, search, voice interfaces and other capabilities needed to make agents reliable and useful in enterprise environments.
Investors are beginning to look beyond individual AI applications towards the technology layers required to support more autonomous systems at scale.
4. Sovereign AI Stacks are Becoming an Investment Theme
Investment is also beginning to reflect the national AI strategies reshaping the region, with some investors backing capabilities that strengthen local control over critical parts of the AI stack. Peak XV described its investment in India’s Sarvam AI – part of a USD 234 million first close of a USD 300 million Series B – as strengthening “India’s sovereign AI stack,” with funding for frontier-model development, agentic AI and large-scale compute. Granite Asia’s Galatek Technologies opened an AI-enabled semiconductor-equipment facility in Penang, explicitly linked to Malaysia’s ambition to move further from chip assembly towards chip design, with a scaling plan from USD 2 million initially to USD 100 million within five years. Vertex Ventures has also said it is increasing deployment across India and Southeast Asia, with semiconductors among the deep-tech categories of interest.
These investments span models, compute, semiconductor capabilities and deployment environments, showing investor interest in the infrastructure and capabilities that can support more locally controlled AI stacks. The investment case is extending beyond compliance or market access. For some investors, national AI priorities are creating opportunities to back companies building local capabilities, reducing reliance on individual external suppliers or technology stacks, and addressing requirements such as data control, compute access and air-gapped deployment.
5. Vertical AI is Attracting Broad Investor Interest
A notable theme is the breadth of investment in AI applications built for specific industries and workflows, rather than in foundation models themselves. East Ventures has stated its preference for “AI-first applications rather than foundation models”.Â
The deal activity supports that view across the region. AirTree backed an AI legal-research workspace, Legora, and AI compliance platform Haast. Vertex backed AI-native lending platform Rezolv and Hubble, an AI contract-management platform for Japan, targeting a market where CLM penetration remains below 2%. Granite Asia backed Craif, a bio-AI company using microRNA analysis for early cancer detection. Square Peg backed AI weather-intelligence platform Tomorrow.io and healthcare imaging platform Aidoc, while Peak XV backed M, an AI-powered household-maintenance concierge.
These investments span law, lending, contracts, healthcare, weather intelligence and home services – investors are applying existing AI capabilities to specific markets and workflows. In many of these businesses, differentiation will depend on domain expertise, proprietary data, workflow integration and distribution, rather than competing on foundation-model scale.
Foundation models continue to attract significant capital. Moonshot AI raised USD 2 billion at a USD 20 billion valuation, while StepFun raised more than USD 718 million. However, these large foundation-model financings sit alongside a much broader set of investments in vertical applications and specialised AI capabilities.
Ecosystm Opinion
The investment activity points to a change in where AI’s economic value is expected to accumulate. The model layer remains critical, and frontier-model companies continue to attract substantial capital. But investors are also placing significant bets on the constraints that determine whether those models can be deployed, integrated and monetised: physical systems, compute and power, specialised silicon, agent infrastructure, sovereign capabilities and domain-specific applications.
That has an important implication for the region. Asis’s opportunity in AI may not depend on producing the next frontier model. It can also come from building the infrastructure, industrial capabilities and specialised applications that make advanced AI usable in markets with different regulatory, economic and operational requirements.
For investors, this broadens the addressable AI opportunity beyond a small number of model companies. For enterprises and governments, it points to a more consequential question than which model performs best: which capabilities will determine access to AI, control how it is deployed, and capture value from its use?
