Singtel–NVIDIA Centre of Excellence for Applied AI

<h1>Singtel–NVIDIA Centre of Excellence for Applied AI</h1> <p>The <strong>Singtel Digital InfraCo–NVIDIA Centre of Excellence for Applied AI (COE AAI)</strong> is a private-sector partnership launched in Singapore on <strong>24 February 2026</strong>. It brings together Singtel's Digital InfraCo and NVIDIA to accelerate applied-AI innovation and sovereign-AI capabilities; a Ministry of Digital Development and Information speech marked the launch, but the centre is not a government agency or a statutory funding scheme (<a href="https://www.mddi.gov.sg/newsroom/speech-by-sms-tan-kiat-how-at-launch-of-singtel-digital-infraco---nvidia-centre-of-excellence-for-applied-ai/">MDDI, 24 February 2026</a>; <a href="https://www.singtel.com/about-us/media-centre/news-releases/singtel-digital-infraco-accelerates-sovereign-ai-innovations">Singtel, 24 February 2026</a>).</p> <h2>Applied-AI and sovereign-infrastructure role</h2> <p>The partnership sits within Digital InfraCo's combination of data-centre capacity, connectivity, orchestration and AI services. Singtel describes Digital InfraCo as a unit formed in <strong>2023</strong> to integrate these capabilities, while its current RE:AI platform provides sovereign AI-cloud services and GPU-as-a-Service offerings through AI-ready data centres. In this context, “sovereign AI” concerns control over where data and processing are handled; it does not mean that every customer receives a government designation or unrestricted public compute (<a href="https://www.singtel.com/digital-infraco">Singtel, Digital InfraCo</a>).</p> <p>The centre’s role is therefore best understood as applied capability-building on commercial infrastructure. The existence of compute, connectivity or orchestration offerings does not establish that a particular workload has been accepted, secured or made available to the public (<a href="https://www.singtel.com/digital-infraco">Singtel, Digital InfraCo, rechecked 12 September 2026</a>).</p> <h2>Ecosystem and deployment boundary</h2> <p>The centre is intended to support organisations moving from AI experimentation toward applied deployment through the Singtel–NVIDIA ecosystem. The launch belongs to Singapore's wider AI and digital-infrastructure ecosystem, which includes government policy, research institutions and other company or sector Centres of Excellence. It is not the National AI Strategy, a public grant administrator, an AI regulator, or a guarantee of access to a particular model, GPU allocation or commercial service (<a href="https://www.mddi.gov.sg/newsroom/speech-by-sms-tan-kiat-how-at-launch-of-singtel-digital-infraco---nvidia-centre-of-excellence-for-applied-ai/">MDDI, 24 February 2026</a>; <a href="https://www.singtel.com/digital-infraco">Singtel, Digital InfraCo</a>).</p> <p>“Support” in the launch description should not be converted into a universal onboarding route. An organisation still needs a commercial relationship, technical fit and any required security, data-protection or sector approvals for its own deployment; the announcement does not publish a blanket acceptance or service guarantee (<a href="https://www.mddi.gov.sg/newsroom/speech-by-sms-tan-kiat-how-at-launch-of-singtel-digital-infraco---nvidia-centre-of-excellence-for-applied-ai/">MDDI, 24 February 2026</a>).</p> <h2>Distinction from other Singapore AI centres</h2> <p>The COE AAI should be kept distinct from the <strong>Built Environment AI Centre of Excellence</strong>, an MND–SUTD collaboration aimed at built-environment challenges, and from public research infrastructure such as the National Supercomputing Centre. It is also distinct from Singapore Government adoption programmes: a commercial partnership can contribute to national AI capability without becoming a universal public entitlement or changing the statutory roles of MDDI, IMDA or other regulators (<a href="https://www.mddi.gov.sg/newsroom/speech-by-sms-tan-kiat-how-at-launch-of-singtel-digital-infraco---nvidia-centre-of-excellence-for-applied-ai/">MDDI, 24 February 2026</a>; <a href="/en/knowledge/article/sgkb.business-industry.built-environment-ai-centre-of-excellence">Built Environment AI Centre of Excellence</a>).</p> <p>The comparison is about ownership, mission and access model. A government-supported or sectoral centre may have different partners and target users, while the COE AAI remains a Singtel–NVIDIA private-sector partnership; similar “Centre of Excellence” wording does not make the initiatives interchangeable (<a href="https://www.mddi.gov.sg/newsroom/speech-by-sms-tan-kiat-how-at-launch-of-singtel-digital-infraco---nvidia-centre-of-excellence-for-applied-ai/">MDDI, 24 February 2026</a>).</p> <h2>Ecosystem context and access boundary</h2> <p>At the February 2026 launch, MDDI described Singapore’s AI ecosystem as containing both sectoral centres that bring research institutions, industry and government together, and company centres that build an individual business’s internal capability. The speech said more than 60 AI Centres of Excellence had been set up since the refreshed National AI Strategy; that ecosystem figure is a government speech snapshot, not a count of COE AAI customers or proof that every centre has the same mandate (<a href="https://www.mddi.gov.sg/newsroom/speech-by-sms-tan-kiat-how-at-launch-of-singtel-digital-infraco---nvidia-centre-of-excellence-for-applied-ai/">MDDI, 24 February 2026, rechecked 12 September 2026</a>).</p> <p>Singtel’s current Digital InfraCo description places the partnership’s infrastructure context in a commercial stack: regional data-centre capacity, connectivity and Paragon orchestration, with RE:AI described as a sovereign AI cloud platform and GPU-as-a-Service offering (<a href="https://www.singtel.com/digital-infraco">Singtel, Digital InfraCo, rechecked 12 September 2026</a>). “Sovereign” in this context should be read as an infrastructure and data-control claim made by the provider, not as a Singapore Government certification. The centre’s launch and ecosystem role do not disclose a universal application route, guaranteed capacity, fixed pricing, or approval for regulated deployment; those matters remain service-specific and subject to the customer’s own legal, security and sector requirements.</p>

简介

Singtel–NVIDIA 应用人工智能卓越中心

**Singtel Digital InfraCo–NVIDIA 应用人工智能卓越中心(COE AAI)**是于 2026 年二月 24 日在新加坡启动的私营部门合作项目。该项目汇聚 Singtel 的 Digital InfraCo 与 NVIDIA,以加快应用人工智能创新和主权人工智能能力建设;数码发展及新闻部的一场演讲为启动仪式揭幕,但该中心并非政府机构,也不是法定资助计划(数码发展及新闻部,2026 年二月 24 日;Singtel,2026 年二月 24 日)。

应用人工智能与主权基础设施的作用

该合作项目属于 Digital InfraCo 的综合能力范畴,涵盖数据中心容量、连接、编排和人工智能服务。Singtel 表示,Digital InfraCo 是于 2023 年成立的部门,旨在整合这些能力;其现有 RE:AI 平台通过为人工智能做好准备的数据中心,提供主权人工智能云服务和 GPU 即服务。在此背景下,“主权人工智能”涉及对数据及其处理地点的控制;这并不意味着每位客户都会获得政府指定身份,或不受限制地使用公共算力(Singtel,Digital InfraCo)。

因此,最好将该中心的作用理解为依托商业基础设施建设应用能力。拥有算力、连接或编排服务,并不能证明某项具体工作负载已获接纳、得到保护或向公众开放(Singtel,Digital InfraCo,于 2026 年九月 12 日重新核查)。

生态系统与部署边界

该中心旨在通过 Singtel–NVIDIA 生态系统,支持组织从人工智能试验迈向实际应用部署。这项启动活动属于新加坡更广泛的人工智能与数字基础设施生态系统,该生态系统包括政府政策、研究机构,以及其他企业或行业卓越中心。该中心不是国家人工智能战略、公共资助管理机构、人工智能监管机构,也不保证可使用特定模型、获得 GPU 配额或使用商业服务(数码发展及新闻部,2026 年二月 24 日;Singtel,Digital InfraCo)。

启动说明中的“支持”不应被解读为适用于所有人的通用接入途径。组织仍需建立商业合作关系、满足技术适配要求,并为自身部署取得任何必要的安全、数据保护或行业审批;公告并未公布一概接纳或服务保证(数码发展及新闻部,2026 年二月 24 日)。

与新加坡其他人工智能中心的区别

COE AAI 应与建成环境人工智能卓越中心区分开来;后者是国家发展部与新加坡科技设计大学的合作项目,旨在应对建成环境领域的挑战。COE AAI 也不同于国家超级计算中心等公共研究基础设施。它还不同于新加坡政府的采用计划:商业合作项目可以为国家人工智能能力作出贡献,但这不会使其成为人人均可享有的公共权益,也不会改变数码发展及新闻部、资讯通信媒体发展局或其他监管机构的法定职责(数码发展及新闻部,2026 年二月 24 日;建成环境人工智能卓越中心)。

这种比较着眼于所有权、使命和接入模式。政府支持或行业中心可能拥有不同的合作伙伴和目标用户,而 COE AAI 仍是 Singtel–NVIDIA 的私营部门合作项目;名称中出现相似的“卓越中心”字样,并不意味着这些计划可以互换(数码发展及新闻部,2026 年二月 24 日)。

生态系统背景与接入边界

在 2026 年二月的启动仪式上,数码发展及新闻部表示,新加坡的人工智能生态系统既有汇聚研究机构、业界和政府的行业中心,也有建设个别企业内部能力的公司中心。演讲称,自更新国家人工智能战略以来,已设立 60 多个人工智能卓越中心;这一生态系统数字是政府演讲中的一个时点数据,并非 COE AAI 客户数量,也不能证明每个中心都有相同的任务范围(数码发展及新闻部,2026 年二月 24 日,于 2026 年九月 12 日重新核查)。

Singtel 当前对 Digital InfraCo 的介绍,将该合作项目的基础设施背景置于一个商业技术栈中:区域数据中心容量、连接和 Paragon 编排;RE:AI 则被描述为主权人工智能云平台及 GPU 即服务(Singtel,Digital InfraCo,于 2026 年九月 12 日重新核查)。“主权”在此背景下应理解为服务提供商对基础设施和数据控制所作的表述,而不是新加坡政府认证。该中心的启动和生态系统作用并未披露通用申请途径、容量保证、固定价格,或对受监管部署的批准;这些事项仍取决于具体服务,并须符合客户自身的法律、安全和行业要求。