SuperAI 2026 Singapore Global AI Industry Mapping

Field One · Influence Tier: T0 category-definer · T1 first tier · T2 effective participant. The judgment is based on the company's global standing within its specific sub-sector.

Field Two · Company Stage: Mature publicly listed or profitable at scale · Growth shipping at scale, Series B or later, or on an IPO track · Early Angel–Series A · Absent a global leader in its category that does not appear in the official report — flagged with ➕ and added in place to keep the category's competitive landscape complete.

• Data sources are SuperAI's official 2026/2025 event reports and the 2024 recap page. These are marketing materials aimed at sponsors and attendees; their figures (such as "1,500+ AI companies" or "3.4 billion+ reach") are not third-party audited and should be read as organizer disclosures rather than independent statistics.
• This mapping's company sample is drawn from verifiable public evidence of participation in the official reports — main-stage speakers, explicitly named partners, local Road to SuperAI co-hosts, and judges or supporting institutions for the Genesis/NEXT competitions. "Absent — added for completeness" only means the company does not appear in the above record, not that it sent no one to the event — companies attending merely as general audience members would not appear on these lists.
• Company tiering and stage assessments reflect this mapping's own analytical judgment and do not constitute investment advice.

The 2026 report discloses two industry-composition datasets — the most direct entry point for understanding the shape of SuperAI's industry.

Observation one: fintech accounts for 16.5%, tied for third with developer tools — this is SuperAI's most distinctive structural feature. Singapore is Asia-Pacific's treasury, payments and private-banking hub, and MAS's regulatory framework gives financial institutions ample incentive to engage publicly on AI — a density rarely matched at other regional AI conferences.

Observation two: foundation models and Agents account for only 2.6% (8% on the AI-native basis). Companies that actually build models are few, and their number keeps shrinking year over year — capital and talent at the model layer have become highly concentrated in a handful of labs, and the companies that remain, even when present at the conference, mostly show up in the form of "applications built on a foundation model" rather than as independent model trainers.

Observation three: applications account for 51% among AI-native companies, but only 28.7% on the all-companies basis. The gap between the two figures is exactly the footprint of traditional enterprises: finance, consulting, media and consumer brands showed up in force as "demand-side" participants. SuperAI's buyer density exceeds its seller density — the official figures of 5,000+ matchmaking meetings and 21,000+ matches reflect precisely this two-sided matching dynamic, not simple technology showcasing.

Observation four: one official category is worth flagging on its own — "Decentralized Intelligence" accounts for 4% of attendees' industry composition. This is a clear imprint left by the crypto/Web3 community at this AI conference; see Section 9.3 for detail.

Per the official figures, "AI infrastructure, compute, chips and cloud" together account for 20.3%. This layer presents itself differently at a conference than it would at a trade show: there is no booth stacked with chip samples — the currency here is main-stage narrative and developer relationships — Cerebras's chief strategy officer holding up a wafer-scale chip on stage is this layer's standard playbook at a conference.

Per the official figures, this broad category totals 20.3% (combined with chips in the reporting). Asia-Pacific is currently the fastest-growing battleground for global cloud vendors, and this layer's participation density is the direct expression of that.

This layer carries the strongest signal in the entire mapping. Among the top-tier partners named in the 2026 report, at least three are web data collection companies — all doing the same thing: scraping the public web at scale, compliantly, and delivering structured results. Giving this category the same billing as hyperscale cloud is the clearest industry signal of the Agent era: once model capability converges and compute becomes rentable, what data you can get — and how fast — becomes a source of differentiation again.

Echoing this is the appearance of Exa (Agent-specific search), Temporal (durable workflows for long-running tasks), and E2B (Agent code sandboxes) in the same list — together, these four categories form the Agent's "senses and limbs", the fastest-growing slot outside the model layer.

By the official classification, "Frontier Labs & AI Agents" account for only 2.6% (8% under the AI-native classification). Few in number, but their share of main-stage time runs far higher than that — at the conference, this layer shows up as narrative, not booths.

AI applications are the largest category under the official classification (28.7% of all companies, 51% of AI-native companies). Most companies at this layer already have verifiable paying revenue or a global user base.

This is SuperAI's most irreplaceable layer. Singapore's identity as a city — Asia-Pacific's treasury hub, a regulatory sandbox, and a concentration of private banks and family offices — gives financial institutions a far higher participation rate here than at most AI conferences.

This is a relatively thin layer within SuperAI, and thin for structural reasons. Robotics is an asset-heavy, display-heavy category that needs booths, floor space, and full-unit logistics — a conference-format event is inherently a poor fit for carrying it. What SuperAI presents is the "narrative layer" of the robotics track: how founders pitch it and how investors question it, rather than the density of physical product on display.

Investment and finance roles make up 11% of attendees, and the Genesis competition and 1:1 Investor Hours turn matchmaking into a formal part of the program.

This layer is the most direct expression of Singapore as host, and the section of this mapping with the most practical value for practitioners on the ground.

In the official attendee industry breakdown, "Decentralized Intelligence" accounts for 4% — a category that doesn't exist in most conventional AI industry taxonomies. This thread's very existence is a product of SuperAI's entanglement with Singapore's strong crypto-community DNA — Singapore is also one of Asia-Pacific's most important crypto hubs, and the overlap between the two communities surfaces naturally at a conference like this.

The five Road to SuperAI tour cities are, in themselves, a geopolitical map:

Reading one: in the roster's structure, "East meets West" is mainly "East meets East." Four of the five stops are in Asia, and 60% of attendees come from Asia-Pacific and Singapore. Genuine Western participation is concentrated among the executives and investors of a handful of companies, rather than company attendance at scale. It functions first as an Asia-Pacific regional hub, and only secondarily as an East-West connector — a distinction in positioning that shapes who you'll actually meet here.

Reading two: the international-expansion players in attendance share one common trait — all have already established an overseas entity or overseas revenue. Alibaba Cloud International, Tencent Cloud International, Manus (now Singapore-based), StepFun and 01.AI (co-hosts of the Shanghai stop), Dify (open-source international expansion), TiDB/PingCAP (infrastructure-software international expansion), Longbridge (Singapore-licensed), Plaud (supply chain + global subscriptions). Almost none of the companies that make it onto this roster got here on the strength of a single-market position — they got here because they already have a working overseas channel.

Reading three: Genesis ($2.3M prize pool, 40+ countries), 1:1 Investor Hours, and 100+ Singapore AI Week satellite events together form a complete matchmaking mechanism. For teams planning international expansion, the most practical way to frame this conference is as "the first checkpoint for Southeast Asian and international customers, and for USD capital."

The table below aggregates every Absent entry in this mapping: global leaders in their respective tracks who do not appear among the speakers, named partners, or competition rosters in the official SuperAI report.

The logic behind this absentee list is fairly consistent: these companies currently have little need to acquire customers through a regional conference. Leading North American AI companies' sales channels run through cloud marketplaces, developer communities, and direct sales — attending a regional conference typically doesn't pencil out well in their customer-acquisition cost model. Absence doesn't mean indifference to Asia-Pacific — it just reflects a different customer-acquisition path.

① Web data collection is being accorded the same level of attention as hyperscale cloud. At least three companies doing the same thing (Bright Data, Oxylabs, NetNut) are named simultaneously in the official report. The logic chain is clear: once model capability converges and compute becomes rentable, differentiation reverts to "what data you can get, and how fast." This is the single most worth-following thread in this year's roster.

② Storage is moving from a supporting role in AI infrastructure to a leading one. WEKA has participated at a high profile for two consecutive years and has its name on a dedicated stage — a parallel file-system company drawing attention on par with hyperscale cloud. As training and inference clusters scale up, data-pipeline throughput becomes the real bottleneck, and pricing power in this layer rises accordingly. Names worth tracking here include VAST Data (private) and Pure Storage.

③ An agent's "hands and feet" are more worth watching than the agent itself. Exa (search designed for agents), Temporal (durable state for long-running tasks), E2B (code-execution sandboxes), Cequence (API and agent security), Stripe (agent payments) — all five categories are on the list, and none of them are "agent companies." The historical pattern is that revenue certainty at the tooling layer usually shows up before it does at the application layer.

① Official figures are marketing material, not an audit report. Figures like "1,500+ AI companies" and "3.4 billion+ media reach" carry no third-party verification, and the definition of "companies in attendance" includes companies that simply registered as attendees. The most credible data points are the individually verifiable speakers and the explicitly named partners — this mapping's sampling frame is built on these two categories of evidence.

② The C-level share has fallen from 56% to 25%, and the conference's function has changed accordingly. If the goal is reaching decision-makers, SuperAI's earlier editions were more efficient; if the goal is learning about solutions and recruiting engineers, the reverse is true. Extrapolating a three-year-old conference experience onto this year's edition is the most common misjudgment made about fast-scaling conferences like this one.

③ The decentralized-intelligence subtrack is cooling off. This track's presence on the roster has trended downward year over year — narrative heat typically fades before funding conditions actually deteriorate, which makes it worth watching on an ongoing basis as a leading indicator.

• Exa — the industry standing implied by its main-stage placement is noticeably higher than what its company size alone would suggest. It designs search for agents rather than humans, making it one of the few independent suppliers at the agent perception layer.
• Longbridge Securities — a Singapore-licensed brokerage that exposes market data and trading capability to agents via OpenAPI and an MCP server. Making regulated financial services agent-callable is a path that's restricted in most jurisdictions, yet workable under Singapore's regulatory framework — worth tracking to see whether this gets replicated elsewhere.
• Plaud — supply-chain manufacturing plus global subscription revenue, a dual model combining one-time hardware sales with recurring software revenue. One of the most commercially complete AI-native consumer hardware products on the list.

• Whether the September 2027 scheduling shift cements the event's positioning on the autumn Asia-Pacific circuit, and whether it changes the composition of attendees;
• Whether NVIDIA returns — it attended in 2025 and was absent in 2026, so its choice in 2027 will be a direct signal of its stance toward the Asia-Pacific AI market;
• Whether the Genesis prize pool holds at the $2.3M level or pulls back — the most direct reading of the organizer's willingness to invest in early-stage deal flow, and the factor that determines whether it remains an effective entry point into Asia-Pacific early-stage deals.

How to use this: This mapping is best suited to two uses — judging whether a given track has already established a commercialization channel into Asia-Pacific, and locating reachable institutions and capital when planning an international-expansion route (Sections 9 and 10). It is not suited to assessing technical leadership, which requires going back to papers, open-source repositories, and actual products. Any judgment on a specific name should still be grounded in revenue structure, customer concentration, gross margin, and cash flow.