Top 10 Singapore AI Companies Leading Southeast Asia [2026]

TL;DR: The largest AI companies headquartered in Singapore in 2026 are Manus, Patsnap and ADVANCE.AI. Silent Eight, WIZ.AI and Tookitaki sell AI to banks and payment firms. Augmentus, Jan, Hypotenuse AI and Datature complete the ten.

Meta bought Manus, a Singapore AI company, for about US$2 billion in December 2025, and then had to give it back. Beijing ordered the deal withdrawn in April 2026. Two other Singapore AI names changed hands in the same stretch: ViSenze now belongs to London’s Rezolve Ai, and Trax sold its shelf recognition business to Gemspring Capital.

The 10 Singapore AI Companies, Compared

Company	What it sells	Main buyers	Latest disclosed funding
Manus	General AI agent	Individuals and teams	US$75M led by Benchmark (2025)
Patsnap	Patent and R&D search agents	R&D and IP teams	US$300M Series E (2021), IPO filed
ADVANCE.AI	Identity checks and fraud detection	Banks, lenders, exchanges	US$80M group round (2023)
Silent Eight	AI agents for screening alerts	Large banks	US$40M Series B (2022)
WIZ.AI	Voice AI agents	Banks, telcos, lenders	Series B, amount undisclosed (2025)
Tookitaki	AML and fraud detection	Banks, e-wallets, payment firms	Undisclosed investment (2024)
Augmentus	Robot programs from 3D scans	Manufacturers	US$11M Series A+ (2025)
Jan	Offline, open-source AI assistant	Developers, privacy-minded users	None, bootstrapped
Hypotenuse AI	Product content and catalog data	E-commerce brands and retailers	Y Combinator, amount undisclosed
Datature	No-code computer vision platform	Inspection, medical and smart city teams	US$2.7M (2022)

The Top 10 AI Companies in Singapore

1. Manus: a general AI agent, back on its own

Manus is for people who want a whole task finished rather than answered: a market scan, a slide deck, a working website. Given a goal, the agent works on its own virtual computer. Manus said in December 2025 that it had created more than 80 million of those computers for users.

Its help centre lists a free tier with 300 daily credits on Manus 1.6 Lite. Pro starts at US$20 a month for 4,000 credits, adds Manus 1.6 Max and runs up to 20 tasks at once. Xiao Hong, Zhang Tao and chief scientist Ji Yichao founded it in Beijing, then moved the company to Singapore in mid-2025.

The buyback was large. Caixin reports that Tencent, ZhenFund and HSG paid US$2 billion for Meta’s shares, and that annual recurring revenue reached US$400 million by late June 2026.

• Flagship: the Manus agent on web, desktop and mobile, with Wide Research, AI slides and a browser operator.
• Backing: Benchmark led a US$75 million round in 2025. Tencent is set to become the largest shareholder after the buyback, according to Caixin.
• Use it when: a job runs across many steps and tools and has to end as a finished file, not a chat reply.

2. Patsnap: a patent search engine with its own language model

R&D and patent teams use Patsnap to test whether an invention is new before they file, and the company is on its way to the stock market. Bloomberg reported in June 2026 that it had filed confidentially for a dual listing in Hong Kong and Singapore. In January, Bloomberg had put the likely raise at US$300 million to US$400 million.

Jeffrey Tiong and Guan Dian, both NUS alumni, started it in 2007, after Tiong found patent searching painful during a student internship. SoftBank Vision Fund 2 and Tencent Investment led its US$300 million Series E in March 2021, which NUS says valued it above US$1 billion.

Its Eureka platform runs agents for novelty search, freedom to operate search, patent drafting and office action responses.

They sit on PatsnapGPT, the company’s own model, which it says draws on 214 million patents across 172 jurisdictions. After a 14-day free trial, Pro costs US$50 a month for 10,000 credits. Its site lists Singapore first among five offices, with London, Tokyo, Shanghai and Toronto.

• Flagship: Patsnap Eureka, plus an Open Platform that serves the same data through REST APIs and MCP servers for teams building their own agents.
• Backing: SoftBank, Tencent, HongShan (formerly Sequoia China), Shunwei Capital and Vertex Ventures. The company says 18,000+ organisations use it.
• Use it when: an engineering, materials or pharma team needs prior art and patent risk checked before money goes into development.

3. ADVANCE.AI: identity checks tuned for Asian ID documents

Banks, lenders and crypto exchanges use ADVANCE.AI to check that a new customer is a real person holding a real ID. In June 2026 it said its liveness SDK blocked all 600 injection attacks in a BixeLab test built on the CEN/TS 18099 standard. Those attacks feed fake or deepfake video into an app in place of the live camera.

ADVANCE.AI is the business-to-business arm of Advance Intelligence Group, founded in Singapore in 2016. The group’s other arm is Atome Financial, a buy now, pay later lender that reported US$470 million of revenue for 2025. Most group headlines are about lending. The AI products sit in ADVANCE.AI.

Its AdvanGuard platform links document verification, face authentication and Know Your Business checks into one onboarding flow. The homepage says it supports 15,000 global ID templates. It also holds iBeta Level 2 conformance for presentation attacks under ISO/IEC 30107-3, and runs credit information businesses in Indonesia and the Philippines.

• Flagship: AdvanGuard, with a liveness detection SDK for iOS and Android.
• Backing: SoftBank Vision Fund 2 and Warburg Pincus led the group’s Series D of more than US$400 million in 2021, which Bloomberg said valued it at US$2 billion. Its last disclosed equity round was US$80 million in May 2023, led by Warburg Pincus and Northstar.
• Use it when: onboarding customers in several ASEAN markets, each with its own ID cards and fraud patterns.

Five of the seven largest disclosed rounds on this list closed in 2021 or 2022. Manus and Augmentus are the only two with a round size made public in 2025, and WIZ.AI raised its 2025 Series B without naming the amount.

4. Silent Eight: AI agents that clear sanctions alerts for global banks

Silent Eight builds AI that decides sanctions, name screening and AML alerts inside large banks, and one of those banks became an investor. HSBC signed a multi-year deal in 2021. HSBC Ventures then joined the company’s US$40 million Series B in March 2022, led by TYH Ventures.

Its current platform, Iris 7, came out in 2025. Each AI agent gathers context, applies the bank’s own policy and returns clear, escalate or block, with a written evidence record. Pre-built agents cover customer and payment screening, transaction monitoring and adverse news. Its Name Screening Adjudicator won Compliance AI Solution of the Year at the ICA Compliance Awards Europe 2025.

The team started in Singapore’s JFDI.Asia accelerator in 2015 and still lists Singapore as headquarters, with offices in New York, London and Warsaw. Martin Markiewicz is CEO. Standard Chartered was the first global bank to deploy it. Later clients include Emirates NBD, Mashreq, First Abu Dhabi Bank, Abu Dhabi Islamic Bank and the insurer AIA.

• Flagship: Iris 7, sold as a managed service, in the customer’s cloud or on premises.
• Backing: TYH Ventures, HSBC Ventures, OTB Ventures, Wavemaker Partners and Standard Chartered’s SC Ventures.
• Use it when: a large bank’s screening team is buried in false-positive alerts and every auto-closed alert needs an audit trail a regulator can read.

5. WIZ.AI: phone agents that speak Singlish, Thai and Tagalog

Banks, telcos and lenders with busy phone lines in Southeast Asia use WIZ.AI for voice agents that understand local speech, Singlish included. The company says its agents handle more than 100 million calls a month for 300+ enterprise clients in 17 countries. It credits them with a 40 to 50% rise in product activation at Monee, formerly SeaMoney.

Four products share one platform. Wizlynn handles inbound service calls, and Talkbot runs outbound campaigns such as payment reminders and debt collection. QM scores calls for compliance, and the Language Engine tunes speech recognition and voices. The homepage lists 16+ languages, including Bahasa, Thai, Tagalog, Vietnamese and Mexican Spanish, and callers can switch language mid-call.

Founded in Singapore in 2019, WIZ.AI closed a Series B worth “tens of millions” of US dollars in November 2025. SMBC Asia Rising Fund, the venture arm of Sumitomo Mitsui Banking Corporation, led it. The announcement quoted founder and CEO Jianfeng Lu and said revenue more than doubled over 2024 to 2025.

• Flagship: Wizlynn inbound agents and Talkbot outbound calling, on a platform with SOC 2 Type II, PCI DSS and ISO 27001 certification.
• Backing: SMBC Asia Rising Fund, Kasikornbank’s Beacon Venture Capital, SMIC SG Holdings, Singtel Innov8 and Granite Asia.
• Use it when: a contact centre takes calls in several Southeast Asian languages and callers mix English with a local language in the same conversation.

6. Tookitaki: AML software fed by a shared library of scam patterns

Tookitaki sells anti-money-laundering and fraud software to banks, e-wallets and payment firms in Asia Pacific, and part of its detection logic comes from other compliance teams. Its AFC Ecosystem is a free, invite-only network where experts share laundering patterns. The company says the network holds 400+ typologies.

The main product, FinCense, covers transaction monitoring, name screening, onboarding risk, customer risk scoring, case management and an alert prioritisation AI agent. New scenarios from the shared network feed its detection rules. FinCense won Best Transaction Monitoring Solution at the Regulation Asia Awards for Excellence 2025, and Maya, the Philippine digital bank, gives a testimonial on Tookitaki’s homepage.

Abhishek Chatterjee and Jeeta Bandopadhyay started the company in Singapore in 2014. In April 2022 the cross-border payments firm Thunes took a majority stake with over US$20 million, according to Thunes’ announcement, though both firms run independently. Tookitaki opened its own Makati office in March 2026, and press reports name GCash, GoTyme and PayMongo as clients.

• Flagship: FinCense, plus the AFC Ecosystem scenario network and its Scenario Design Studio.
• Backing: Thunes’ majority stake (April 2022) and a strategic investment of undisclosed size from True Global Ventures in October 2024.
• Use it when: you run AML for a bank, e-wallet or remittance firm in Southeast Asia and want detection scenarios built from other institutions’ cases.

Four of the first six companies sell to banks, lenders or payment firms, and three of them list bank-grade certifications or audit trails as selling points. The last four sell to factories, developers, retailers and vision teams. Our list of AI fintech startups in Southeast Asia covers the finance side across the region.

7. Augmentus: robot programs generated from a 3D scan

Augmentus is for factories that sand, spray, blast or weld parts too varied to hand-program a robot for each one. Its software scans the real part sitting in the cell and writes the robot’s toolpath from that scan. The company says it has sold more than 50 systems in over six countries.

The Standard licence covers CAD import, automatic toolpath generation, offline simulation with cycle-time estimates, and program output for ABB, FANUC, KUKA, Yaskawa, Kawasaki, Universal Robots and Techman arms. Pro adds 3D scanning, automatic hand-eye calibration, path correction when a part shifts, and defect detection. A 14-day trial simulates on your own parts but cannot export programs.

Daryl Lim, Chong Voon Foo and Leong Yong Shin founded the company in 2019. Its global head office sits in Singapore Science Park, with a robotics centre in Pflugerville, Texas. Version 3.1, released in July 2026, added rotary scanning that captures a full 360-degree mesh in one turn of a positioner.

• Flagship: the AutoPath robotics stack, sold as Augmentus Standard (software) and Augmentus Pro (software plus 3D vision).
• Backing: Woori Venture Partners led a US$11 million Series A+ in July 2025, with EDBI, Sierra Ventures and Cocoon Capital, per DealStreetAsia. Applied Materials’ venture arm added a strategic investment in October 2025.
• Use it when: high-mix, low-volume parts make manual robot teaching the slowest step on a finishing or welding line.

8. Jan (Menlo Research): a ChatGPT-style app that runs offline

Jan is for people who want a ChatGPT-style assistant that runs on their own computer, and its GitHub repository has passed 44,000 stars. Menlo Research, the Singapore startup behind it, got there without venture funding. Tech in Asia reported that it gives 1% of company equity to staff and contributors every six months.

The desktop app downloads open models such as Qwen, Gemma, Llama and DeepSeek and runs them offline, or connects to cloud models from OpenAI, Anthropic and Google. It also serves an OpenAI-compatible API on localhost port 1337. Version 0.8.4 shipped on 23 July 2026, and the homepage claims more than 6.5 million downloads.

The team trains its own small models too, all on Hugging Face under Apache 2.0. Jan-nano and Lucy arrived in mid-2025, followed by Jan-v1-4B, the Jan-v2-VL vision models and Jan-v3.5-4B in March 2026. Menlo’s own site now leads with Asimov, an open-source humanoid robot kit.

• Flagship: the Jan desktop app for Windows, macOS and Linux, plus Jan Agent and Tokamak, a self-hosted backend that Jan agents connect to.
• Backing: bootstrapped. Menlo lists offices in Singapore, Palo Alto and Ho Chi Minh City, and Jan’s handbook says most of the team works from Ho Chi Minh City.
• Use it when: prompts and files cannot leave the machine, or a developer wants a free local endpoint for testing apps against open models.

9. Hypotenuse AI: product copy and catalog data for online stores

Hypotenuse AI is for e-commerce teams with thousands of product pages to fill, and its homepage says more than 500,000 e-commerce users have signed up. A quote there from Crunchyroll’s head of e-commerce says the tool tripled the team’s output.

Joshua Wong started it in Y Combinator’s Summer 2020 batch as an AI writer for product descriptions, two years before ChatGPT. It now calls itself an AI-first PXM, a product experience management system. The software fills missing attributes from web pages, UPC codes and spec sheets, tags items against Google, Amazon or custom taxonomies, and cleans up catalog photos.

Neither plan on the pricing page shows a price. Basic covers stores with fewer than 100 products and includes one seat, 40+ languages and 1,000 HypoArt images. Ecommerce Enterprise adds PIM and ERP integrations, bespoke models and a checker for retailer guidelines.

The operating company is Hypotenuse Technologies Pte. Ltd. in Singapore, though Y Combinator lists its location as San Francisco.

• Flagship: the product description generator with bulk generation, attribute enrichment, AI image editing and the HypoChat research assistant.
• Backing: Y Combinator, January Capital and executives from Amazon, ShopBack and Carousell, according to a 2024 Scholars’ Choice profile. No round size has been made public.
• Use it when: a brand or retailer has a large catalog with thin product data and needs descriptions and tags that match each marketplace’s format.

10. Datature: label, train and deploy vision models without code

Datature is for teams that want a working computer vision model without writing training code. Nvidia, Rolls-Royce and Singapore’s Land Transport Authority have used it, according to a DigiPen Singapore profile. Keechin Goh and Denzel Lee started the company in 2020.

Work runs in three steps on one platform. Users label images with AI-assisted tools for boxes, keypoints, segmentation masks and 3D data. They then train classification, detection or segmentation models through a drag-and-drop workflow, and deploy the result as an API or to edge devices.

The free plan allows 300 images, 300MB of storage, five exportable models and 300 training minutes a month. Developer raises that to 50,000 images, 5TB and 3,000 minutes, and Professional to 350,000 images and 35TB. Paid tiers show no public price. Enterprise adds deployment on private servers or cloud, plus HIPAA and SOC II compliance.

• Flagship: the Datature platform, split into Label, Train and Deploy products, with solution pages for medical, retail, agriculture, smart city and manufacturing teams.
• Backing: US$2.7 million from Openspace with January Capital taking part, TNGlobal reported in March 2022. No later round has been reported.
• Use it when: an inspection, medical imaging or farming team has photos and wants one tool from annotation to a deployed detection model.

Where Singapore’s AI Money Comes From

Singapore’s pull on regional capital is not new. Its companies took 68% of Southeast Asia’s tech funding in 2024 by venture deal value in the first nine months, on Enterprise Singapore and PitchBook figures. For AI alone, Tracxn counts US$9.3 billion raised by Singapore-based native AI companies across 227 rounds, TNGlobal reported. Vietnam, Malaysia, Indonesia and Thailand together raised less than US$40 million.

One deal explains most of the 2026 jump. Kling AI’s US$2.8 billion Series D made up about 68% of the region’s AI funding this year. Kling is the AI video unit of Kuaishou in Beijing, and its international service runs through Kling AI Pte. Ltd., a Singapore company set up in January 2025.

MiniMax, the other large round in the tally, runs its business from Shanghai and names two Singapore subsidiaries in its Hong Kong prospectus.

Chip sales show the same gap between paperwork and operations. NVIDIA billed US$23.68 billion, 18% of its fiscal 2025 revenue, to customers in Singapore. Its annual report says customers use Singapore to centralise invoicing, and shipments to Singapore were less than 2% of revenue.

Public money is smaller and aimed at people and compute. Budget 2024 set aside more than S$1 billion over five years for AI compute, talent and industry. The goal is to triple the pool of AI practitioners to 15,000, according to the Ministry of Digital Development and Information. ASPIRE 2B, the national supercomputer launched in June 2026, runs more than 1,500 NVIDIA H200 GPUs. The wider picture is in our guide to AI trends in Singapore.

Hiring AI Engineers for a Singapore Company

Several companies on this list build outside Singapore: most of Jan’s team works from Ho Chi Minh City, and Silent Eight runs an office in Warsaw. Hiring locally is hard: Singapore had the highest share of job postings asking for AI skills in the Stanford AI Index 2026, at 4.7%, on Lightcast data.

Second Talent places vetted, AI-native engineers from nine Asian markets at US$3,000 to US$6,000 a month per engineer. Payroll and compliance run through our Singapore employer of record. See our AI engineers and Singapore developers, or tell us the role you need to fill and get a shortlist within 24 hours.

Frequently Asked Questions

What is the largest AI company in Singapore?

By deal value, Manus: Meta paid about US$2 billion for it before China forced the sale to be unwound. By valuation, ADVANCE.AI’s parent group and Patsnap lead, both valued at US$1 billion or more in 2021.

Which Singapore AI companies were sold recently?

Visual search company ViSenze was bought by Nasdaq-listed Rezolve Ai in August 2025, e27 reported. Trax sold its image recognition business to Gemspring Capital in February 2026, which merged it into FORM in Massachusetts.

Does Singapore have its own large language model?

Yes. AI Singapore publishes SEA-LION, an open model family for Southeast Asian languages. Version 4.5 came out on 20 May 2026 as a 27B model built on Qwen. It is tuned for Burmese, Indonesian, Filipino, Malay, Tamil, Thai and Vietnamese as well as English.

How widely is AI used in Singapore?

Microsoft’s AI Diffusion report ranks Singapore second in the world, with 63.4% of working-age people using generative AI in the first quarter of 2026, behind the UAE. For the country count of billion-dollar startups, see our unicorns by country ranking.