NCS: Chinese open AI models gaining ground in Singapore

Chinese open-weight artificial intelligence (AI) models are gaining ground among organisations in Singapore as lower costs and the ability to run the models in their own environments win over customers, according to home-grown technology services provider NCS.

“The best open-weight models in the world are actually coming from China, with very good adoption among our clients,” said Edward Chen, NCS’ first chief AI officer, at a media briefing ahead of the firm’s annual technology forum, NCS AI Impact 2026, held on July 9.

Unlike proprietary models from US AI labs, which are typically accessed over the cloud, open-weight models can be downloaded and deployed on an organisation’s own infrastructure and at a fraction of the cost of Western closed models, Chen said. “Increasingly, our customers – even government customers – are very open about adopting this approach.”

NCS itself has folded Alibaba’s Qwen models into its own products, including its NCS GPT assistant, and used the forum to announce a partnership with Alibaba Cloud to take enterprise AI to organisations across the region.

The Singtel subsidiary, which has 15,000 staff across the Asia-Pacific region, is counting on its fluency in both American and Chinese AI stacks – at a time when the two ecosystems are pulling apart – to gain a competitive edge over rivals.

That East-meets-West positioning was on full display at the one-day forum, which NCS expected to draw more than 1,500 attendees, including over 300 C-suite executives. For the first time, its developer conference put engineers from Google DeepMind, OpenAI, Nvidia and Mistral AI on the same programme as counterparts from Chinese AI players including Alibaba, MiniMax and Moonshot AI, with close to 1,000 developers signed up.

“There are not many companies – or countries, for that matter – that can bring both Eastern and Western engineers together,” said NCS chief executive Sam Liew, who took the helm in April this year.

Straddling both worlds is already paying off for NCS. In China, the company has helped a European multinational integrate local technology stacks with Salesforce. Chinese firms expanding overseas, meanwhile, have tapped NCS to re-platform their products onto Amazon Web Services and Google Cloud. “I’m not locked into any single product, and that’s why customers look for me,” Liew said.

At the forum, NCS launched six additions to its Sunshine.AI suite of what it calls “sovereign, enterprise-grade” AI platforms – systems that customers in regulated sectors such as government, defence and financial services can run entirely within their own environments, on models of their choosing.

They include Sunshine.core, a foundational platform for building and running fleets of AI agents, which Chen said has already been deployed at parent company Singtel; Sunshine.commanderAI, a command centre for robot fleets from multiple vendors; and Sunshine.guardian, an AI governance engine.

There’s also Sunshine.kaisense, a video AI platform built to ingest up to 100,000 camera feeds, and Sunshine.builder, a no-code tool that lets business analysts turn requirements into working software.

The cheekiest launch is Sunshine.chilliclaw, named after Singapore’s beloved chilli crab. It is NCS’ enterprise answer to OpenClaw, the viral open-source AI assistant that Nvidia chief executive Jensen Huang declared “the next ChatGPT” at his GTC keynote earlier in the year, when he said every company needs to have an OpenClaw strategy.

NCS took up the idea but did a total rewrite of the whole software, Chen said, because OpenClaw is fundamentally insecure and “very easy to hack”. Chilliclaw embeds agentic AI into the tools employees already use, such as Microsoft Teams, Outlook and Slack, he added.

The firm also inked a string of sector deals: a joint AI centre of excellence with hospital group IHH Healthcare; agentic AI pilots with NHG Health spanning biometric identification, digital pathology and human resources; an AI tutor at Ngee Ann Polytechnic that adapts to individual learners; and tie-ups with Chinese healthtech players, including exoskeleton makers Fourier Rehab and Hypershell.

Closer to home, NCS will pilot a driverless shuttle for its own staff, integrating Korean firm Autonomous a2z’s ROii vehicle – a level four self-driving shuttle that already ferries Grab employees in Singapore – with its RobotManager software to ply the 1.8km between Yio Chu Kang MRT station and its Ang Mo Kio campus during off-peak hours.

NCS will also hire over 130 AI practitioners over three years with support from Digital Industry Singapore, and is developing applied AI masterclasses for senior executives with the Singapore University of Technology and Design, the National University of Singapore’s School of Continuing and Lifelong Education, AI Singapore and – in another East-West nod – Shanghai Jiao Tong University.

Tying it together is a new NCS AI Playbook distilled from more than 100 projects, built on four execution pillars – cheaper, better, faster and safer, which is an echo of the economic mantra made famous by former labour chief Lim Swee Say, who chairs NCS, with “safer” added to emphasise the importance of cyber security in the AI age.

Gracing the event was Singapore’s health minister Ong Ye Kung, who is also coordinating minister for social policies. He called NCS “more than a vendor or a technology provider” and “a long-term partner in national digital transformation” with deep roots in Singapore's healthcare IT ecosystem.

But he also sounded a note of caution amid the industry’s race to deploy AI. “We cannot charge ahead, driven solely by commercial considerations, even as recursive AI systems gain self-reinforcing intelligence, agency and influence. Otherwise, the machines just seem wiser than their makers,” he said.

“We must be wiser and more humanistic. We must decide deliberately where to embrace AI, where to rein it in, and where human judgement and effort must prevail,” he added.

In healthcare, that means deliberate guardrails, Ong said. AI reading mammogram scans, for instance, may stand in for only one of two human readers, to maintain clinical skills and partly to ensure human judgement remains relevant when it comes to caring for another human being.

He also outlined plans for a common electronic medical record system across all three public healthcare clusters by 2028 and gave an update on Simfoni – the Singapore Medical Foundation AI Model programme set up in 2025 – which will train clinical AI models on local data and practice guidelines, starting with cardiometabolic and eye diseases.