Production AI Infrastructure Goes Live in Saudi Arabia

The current production deployment is built on AMD Instinct MI355X GPUs and AMD EPYC CPUs, interconnected through Cisco networking systems. The network fabric uses Cisco Silicon One technology and Cisco 800G optics, running on the Cisco N9000 Series platform. This architecture is designed for scale, low latency and operational resilience, connecting the MI355X GPUs in an AI-optimised fabric purpose-built for demanding AI workloads.

With this Saudi AI Infrastructure now operational, HUMAIN is able to offer GPU as a service across a range of use cases including AI model training and inference. The platform is designed to serve governments, enterprises, research institutions and developers, providing access to advanced AI capabilities through open models, open software and locally operated infrastructure.

The companies say the platform gives customers greater control over where their data resides, how their models are customised and how AI systems are deployed and governed. This locally operated approach enables organisations to build AI capabilities aligned with their regulatory requirements, languages and national priorities, reinforcing the sovereign dimension of this Saudi AI Infrastructure deployment.

Dr. Lisa Su, chair and CEO of AMD, said: “Bringing AMD Instinct systems online in Saudi Arabia is an important milestone in our work with HUMAIN. Together, we are building an open, high-performance AI platform that is serving customers today and will scale significantly in the coming years.”

Tareq Amin, CEO of HUMAIN, said: “Putting this infrastructure into production demonstrates HUMAIN’s ability to build, operate and deliver advanced AI infrastructure at scale.”

Chuck Robbins, chair and CEO of Cisco, added: “The region is moving quickly from AI investment to delivery, and through our work with HUMAIN and AMD, Cisco is building the critical infrastructure to make AI vision a reality.”

Planned Expansion Targets 250 MW of Additional Capacity

Building on the live deployment, the three companies plan to deploy up to 250 MW of additional AI infrastructure through their previously announced joint venture. This next phase will use AMD Instinct MI400 Series GPUs, AMD EPYC CPUs, AMD ROCm open software and Cisco networking along with supporting critical infrastructure. Deployment of the 250 MW phase is planned to begin in 2027, with capacity expected to start coming online in the second half of that year.

The broader joint venture remains on track to deploy up to 1 GW of Saudi AI Infrastructure by 2030. The expansion is intended to significantly increase the Kingdom’s AI compute capacity, extending GPU as a service and sovereign AI capabilities to customers across the region and globally.

With production systems already serving customers and a clear roadmap for scaling, the Saudi AI Infrastructure programme now enters its next stage of growth as the companies prepare for the 2027 expansion phase.

The partnership connects a local cloud communications and IT services provider with an international communications software company that has been expanding through regional partnerships. Digital Island, a New Zealand-based provider focused on business collaboration and customer engagement services, will now offer RingCentral’s platform alongside its own local implementation and support capabilities.

New Zealand customers will gain access to RingCentral’s Agentic Voice AI offering through a provider that can manage deployment and support on the ground. The two companies are already working with early adopter customers ahead of wider availability later in 2026.

Agentic Voice AI Expands Through Local Partner

For Digital Island, the partnership adds an international software platform to its portfolio at a time when many businesses are reviewing their communications systems and customer service operations. The company will provide end-to-end services spanning initial design, implementation and ongoing support.

Leon Sheehan, Chief Executive Officer of Digital Island, said: “As RingCentral’s first partner in New Zealand, Digital Island is proud to bring Kiwi businesses access to AI-powered unified communications and contact centre solutions that modern organisations require. We are committed to helping businesses transform both customer and employee experiences with our New Zealand-based team of experts.”

RingCentral sells business telephony, messaging, video, contact centre and related software. Homayoun Razavi, Executive Vice President and General Manager of Global Strategic Providers at RingCentral, said: “We built our global strategy around partnering with local market leaders who understand what their customers need on the ground, and Digital Island is exactly that partner. Together, we’re giving New Zealand businesses a direct path to Agentic Voice AI, running on a cloud platform built for reliability, data privacy, and enterprise-grade security.”

RingCentral Services Target New Zealand Businesses

The partnership model reflects a common route to market for overseas software providers entering smaller national markets. Local sales, migration work and support are often critical to winning contracts in cloud communications, particularly for customers replacing older phone systems or moving contact centre operations to cloud-based platforms.

Krishna Baidya, Head of Connected Work and Customer Contact Research, Asia-Pacific, at Frost & Sullivan, pointed to the importance of local execution. “Moving off legacy communications takes more than software licences. It takes a local partner who can own strategy, deployment, and ongoing support, end to end. Pairing RingCentral’s cloud platform with Digital Island’s local service model gives New Zealand businesses a practical, lower-risk path to modernisation,” said Baidya.

The agreement arrives as competition in cloud communications and AI-assisted customer engagement continues to intensify across Asia-Pacific. Vendors are seeking local partners that can translate broad software offerings into sector-specific projects, while customers weigh the benefits of automation against integration costs and the risks of disrupting day-to-day service. Digital Island will now handle not only sales for RingCentral but also the deployment and support work that often determines whether a unified communications overhaul succeeds.

The agreement, signed between Maxis and Melaka ICT Holdings, is designed to transition telecom towers currently relying on microwave backhaul to advanced optical fibre connectivity. The fibreisation programme is expected to increase the proportion of fibre-connected telecom towers from approximately 40 percent to 80 percent across the affected sites.

Fibre Backhaul Upgrade Covers 151 Towers

Under the partnership, Maxis will upgrade 151 towers owned by Melaka ICT Holdings in a phased implementation. The core objective is to replace wireless microwave backhaul with optical fibre, establishing a more robust transport layer between telecom towers and the wider network. This shift from microwave to fibre backhaul is intended to deliver greater capacity and more reliable connectivity across the state’s telecom infrastructure.

The approximately 200 km of fibre being deployed will form a digital backbone that connects towers previously served by microwave links. The RM30 million investment reflects the scale of physical infrastructure work involved in laying fibre across Melaka’s tower network.

Beyond serving mobile network traffic, Maxis and Melaka ICT Holdings also plan to use the fibre network to offer wholesale mobile backhaul services to other operators, a move intended to foster broader competition and shared use of the new telecom infrastructure.

Maxis CEO Goh Seow Eng said the fibreisation effort lays the foundation for a future-proof ecosystem. “This network offers the scalability needed to meet the exponential rise in data demand, providing the high-speed, low-latency backbone that mobile connectivity and AI traffic demands,” he stated.

Melaka ICT Holdings CEO Dr. Nazdiana binti Ab Wahab noted that the collaboration will accelerate migration from microwave-based connectivity to high-capacity fibre networks, enhancing network performance and supporting growing digital demands across the state.

Fibreisation Supports 4G and 5G Infrastructure

The fibre backhaul programme is being implemented to support both current 4G services and future 5G requirements. Moving from microwave backhaul to optical fibre gives telecom towers the capacity needed to handle increasing data traffic generated by next-generation mobile services. Fibre-connected towers are better positioned to support the low-latency and high-throughput demands that 4G and 5G networks require.

The project also builds on a memorandum of understanding signed by Melaka ICT Holdings and Maxis in October 2024 to enhance broadband connectivity across Melaka. The initiative aligns with the state’s Smart Melaka Blueprint 2035, which identifies the development of a digital highway as a key foundation for IoT, AI, smart cities, and the digital economy.

Melaka’s Chief Minister Ab Rauf bin Yusoh said the fibreisation programme supports the state’s broader vision for strengthening digital infrastructure and enabling future technology-driven services. The phased deployment of optical fibre across 151 telecom towers represents a significant upgrade to the backhaul layer of Melaka’s telecom infrastructure, positioning the state’s network for the connectivity demands ahead.

The four agreements include a Letter of Intent with PEACE Cable International Network, two Memoranda of Understanding with GTI and SM+ respectively, and an Open API and Open Digital Architecture Manifesto commitment with TM Forum.

PEACE Cable Deal Expands International Connectivity

The Letter of Intent signed with PEACE Cable International Network is the most infrastructure-specific of the four agreements. Under the LOI, Telin plans to increase its connectivity resources with PEACE Cable to around 1.4 Tbps across strategic routes connecting Asia, Africa and Europe. The agreement also covers dark spectrum resources connecting Egypt and France, along with end-to-end capacity supporting cross-Egyptian connectivity.

This expansion of subsea cable resources is designed to strengthen Telin’s international connectivity capabilities while providing greater capacity to support growing demand for digital services across multiple markets. The PEACE Cable system serves as a critical link between continents, and the planned capacity increase positions Telin to serve enterprise and wholesale customers requiring reliable intercontinental routes.

Agreements Extend Fibre and Data Centre Links

Telin also signed a Memorandum of Understanding with Graha Telekomunikasi Indonesia, known as GTI, which is part of the Remala Group. The MoU establishes a collaborative framework to explore fibre pair capacity and international connectivity opportunities through the BSCS Fiber Pair Project. The arrangement focuses on fibre infrastructure that could enhance connectivity between key regional nodes.

On the data centre side, Telin signed an MoU with SM+, an Indonesian data centre operator backed by the Sinar Mas conglomerate. Under this agreement, Telin will support SM+ in expanding its data centre colocation solutions and reaching new customers in international markets through Telin’s global network. The arrangement includes integration with Telin’s TNeX platform, its global data centre connectivity service, which would link SM+ facilities to Telin’s broader international connectivity portfolio.

The fourth agreement saw Telin sign the TM Forum’s Open API and Open Digital Architecture Manifesto. This commitment involves adopting Open APIs and Open Digital Architecture as foundational elements of Telin’s software architecture. The move is intended to align Telin’s technology stack with global industry standards and increase interoperability across digital ecosystems. This is a standards and architecture commitment rather than a physical network deployment.

“For Telin, building global capabilities is no longer about expanding the network, but also how we can connect various technologies, platforms, and players in the ecosystem to create more choices and value for customers,” said Telin CEO Abdul Rahman Ansyori. “We want to continue to be at the meeting point of various capabilities to open a new growth space for Telin, TelkomGroup, and partners.”

Taken together, the four agreements extend Telin’s reach across subsea cable capacity, fibre infrastructure, data centre connectivity and open architecture standards, reinforcing its role as a provider of international connectivity services linking Asia with Africa and Europe.

The AI for All Project is designed to go beyond general-purpose chatbots by creating AI agents that connect to actual public, medical, financial, education and housing services. Rather than users navigating multiple apps and websites, the AI finds the relevant service and carries them through each step. Services under the AI for All Project will roll out in stages, with the Kakao consortium planning a beta service in September and SK Telecom unveiling its beta in October before a full launch within the year.

SK Telecom, KT and Kakao Selected for AI for All

Each of the three consortiums brings a distinct approach to delivering nationwide AI services.

The SK Telecom consortium, which includes Goodoc, Shinhan Card and TMAP Mobility, has proposed an action-oriented daily-life AI assistant. The service would be accessible through phone calls and text messages, handling tasks ranging from government paperwork and health management to job hunting. By linking with Government24 and PASS, users could obtain 83 types of certificates. The assistant would also gather medical checkup results, explain them in plain language and guide users to nearby hospitals. For job seekers, it would track application deadlines and assist with cover letters and interview preparation. SK Telecom plans to build its service around its proprietary AI foundation model while also drawing on a range of domestic AI models.

KT’s consortium includes Upstage, Musinsa and Zigbang. KT has proposed a personalised AI service that ties together housing, education, finance and shopping into a single integrated assistant linked to commerce platforms. With user consent, the AI remembers interests, schedules and tasks in progress, and proactively suggests what needs to be done next. KT’s consortium will draw on multiple Korean AI models and will also deploy domestic AI infrastructure technologies, including Korean-made neural processing units.

The Kakao consortium, which includes LG Uplus and Seoul National University Hospital, plans to make KakaoTalk the main point of contact for AI services. Users would access a general-purpose AI chatbot within KakaoTalk and be connected to specialised AI agents covering medical, financial, tax, education, employment and care services. The service will run on in-house AI models including Kakao’s Kanana and K-EXAONE.

Nationwide AI Services Planned After Beta Testing

The government plans to sign agreements with the three selected operators and will provide them with 512 Nvidia B200 GPUs this year to support the AI for All Project. The nationwide service is planned to launch after the beta phase is completed.

A key priority of the AI for All Project is accessibility. SK Telecom is making its AI available through phone calls and text messages, while Kakao and LG Uplus plan to offer access through KakaoTalk and phone calls. KT will build a dedicated app and website while also applying AI functions to existing services. These channels are intended to ensure that older adults and people less comfortable with digital devices can use AI services through familiar tools.

The project also aims to create real-world use cases for the domestic AI industry. All three consortiums plan to use multiple Korean AI models, turning large-scale public services into a proving ground for domestically developed artificial intelligence technologies. The role of major telecom operators in delivering these nationwide AI services underscores the infrastructure capabilities required to support the AI for All Project at scale.

The regulation applies to commercial, industrial and residential buildings, as well as real estate complexes. Developers and property owners will be required to provide infrastructure such as ducts, cable routes and other facilities needed to support telecommunications networks.

The new framework is designed to prevent buildings from being structured around a single telecom operator. It requires internal network infrastructure to support access by different licensed operators on fair and non-discriminatory terms.

The rules also establish responsibilities for building owners and telecom licensees. Owners are responsible for creating and maintaining internal passive connections unless they appoint a licensed operator to handle the work.

In-building coverage systems must be designed and built by a contracted licensee for larger developments meeting specified thresholds. These include complexes with more than 48 property units, six or more above-ground floors, two or more underground floors, or a single floor exceeding 3,000 square metres.

New Compliance Rules Cover Building Approvals

The regulation makes Passive Telecom Infrastructure part of the building approval process. Before a municipality issues a building completion certificate, an authorised or licensed entity must inspect the internal telecommunications connections and confirm that they meet TRA requirements.

For larger developments where an operator is contracted to build and operate the infrastructure, other licensed operators may be invited to share construction and operating costs. The framework also provides for shared maintenance responsibilities where such arrangements are accepted.

The rules cover fibre and copper cabling, ducts, equipment and protection rooms, power supplies, backup batteries and other passive network elements. Fibre-optic cabling will be required in schools, hospitals and other facilities designated by the TRA.

Most of the new requirements will take effect six months after publication. Buildings with construction permits issued before the regulation takes effect are exempt, while certain residential requirements have a longer implementation period.

The TRA has also established penalties for non-compliance. Fines include OMR 500 for non-compliant internal connections, OMR 300 for arrangements that give one operator exclusive service in a building, and OMR 800 per 1,000 square metres of affected area for certain larger buildings.

The regulation is intended to ensure Passive Telecom Infrastructure is planned alongside building development rather than added later. By setting common technical and access requirements, the TRA is creating a framework that allows telecom operators to use shared internal infrastructure while giving developers clearer compliance obligations.

This Starlink Satellite Internet Licence represents an important milestone in the country’s regulatory framework for satellite communications. It introduces a space-based connectivity layer to the UAE’s national digital infrastructure, complementing existing terrestrial fibre-optic and 5G networks rather than replacing them. The move is intended to diversify internet access options, enhance network resilience, readiness and continuity, and support critical sectors including maritime and aviation transport, energy, logistics and emergency response.

UAE Grants Starlink 10-Year Satellite Licence

From a regulatory standpoint, the Starlink Satellite Internet Licence reflects TDRA’s approach to building a flexible and forward-looking framework that embraces emerging technologies while expanding user choice and promoting competition. TDRA said this is expected to contribute to service quality, customer experience and continuity of telecommunications services. The authority also noted that the licence reinforces the UAE’s position as an attractive destination for global satellite system operators, in support of the UAE Digital Agenda and the “We the UAE 2031” vision.

Majed Sultan Al Mesmar, Director General of TDRA, said: “This licence represents a significant addition to the UAE’s telecommunications sector and reflects the country’s commitment to adopting advanced technologies and fostering a flexible regulatory environment that supports innovation and investment. The introduction of advanced satellite internet services will expand connectivity options, enhance network resilience and business continuity, and support the UAE’s ambition to strengthen its position as a regional and global hub for telecommunications and the digital economy.”

He added: “TDRA is committed to ensuring that the introduction of emerging and new technologies delivers tangible benefits to customers, service quality, and sector competitiveness, while maintaining the highest standards of security, reliability, and consumer protection.”

Licence Expands Satellite Connectivity Across Key Sectors

The scope of the Starlink Satellite Internet Licence extends well beyond individual consumers. It covers businesses and government entities, as well as satellite connectivity services for the maritime and aviation sectors, all in accordance with the UAE’s approved regulatory and technical frameworks. This broadens the licence’s strategic and economic significance across key productive sectors of the UAE economy.

TDRA said the broadband satellite internet services covered by the licence are expected to contribute to digital transformation, develop connectivity solutions for vital sectors and areas requiring additional satellite connectivity options, and enhance the UAE’s readiness to respond to emergencies and crises.

The services authorised under the Starlink Satellite Internet Licence are subject to comprehensive regulatory and technical requirements. These include obligations related to security and the protection of information infrastructure, service quality, reliability and continuity, consumer rights and data privacy, as well as compliance with spectrum-use regulations and technical coordination with relevant authorities.

TDRA stated that these requirements form an essential part of licensing decisions of this nature. The authority’s regulatory approach seeks to balance enabling access to advanced telecom infrastructure and satellite connectivity technologies with ensuring that deployment takes place within a robust framework that safeguards network security and consumer rights. The Starlink Satellite Internet Licence positions satellite communications as a complementary element within the UAE’s broader digital infrastructure strategy.

A network digital twin is more than a visual copy of a network. It is a virtual representation that can reflect the physical and operational state of the real network using data such as equipment configuration, device status, events, topology, performance and service information. The ITU’s Y.3093 recommendation, approved in August 2025, establishes requirements for the data domain supporting network digital twins and specifically highlights real-time interaction between physical and virtual networks.

That real-time relationship is what makes digital twins relevant to autonomous telecom infrastructure. A static model can show what a network looks like, but an operational twin can help represent what the network is doing now and how it may respond to a proposed change. AI can then use that environment to test configurations, evaluate possible outcomes and identify a preferred action without immediately applying every experiment to the live network.

The Network Needs a Virtual Environment for AI

The need becomes more important as networks move toward higher levels of autonomy. An autonomous system has to do more than detect a fault. It needs to understand the network state, evaluate possible responses and determine whether a change could create a new problem somewhere else.

A digital twin can support that process by bringing different network data sources into one environment. ITU’s framework includes historical and operational data, device status, network topology, performance information, events and service data, while also requiring compatibility with existing databases, data lakes and physical network equipment.

This creates a foundation for AI-driven simulation and optimisation. Instead of making a network change first and observing the result afterwards, operators can increasingly test potential decisions against a virtual representation of the infrastructure.

The concept is already moving into next-generation network research. Work on network digital twins is now part of formal 3GPP standardisation, with the Release 19 study TR 28.915 focused specifically on the management aspects of Network digital twins.

Research demonstrations are also beginning to show measurable performance benefits. A 2026 6G RAN study using digital-twin-based AI control reported up to 20% improvement in uplink throughput in its demonstrated scenario. The digital twin was used to evaluate and validate the AI-driven network-control approach before applying it to the real-world environment.

The significance is not the 20% figure alone, which comes from a specific demonstration rather than an industry benchmark. It is that a virtual representation of the network can become part of the process through which AI decisions are evaluated.

At its core, AI network slicing allows multiple logical networks to operate over the same physical infrastructure, with each slice designed around particular performance requirements. GSMA identifies slicing as a core 5G capability for applications that need differentiated quality of service, including high bandwidth, low latency and predictable performance for areas such as industrial robotics and autonomous systems.

The arrival of 5G-Advanced strengthens the foundation for this model. 3GPP describes Release 18 as the first 5G-Advanced release, introducing further enhancements to the 5G system. Its work also includes enhancements around AI and machine learning for NG-RAN, including support for AI/ML-assisted network slicing and coverage and capacity optimisation.

This matters because future telecom infrastructure will have to support applications with very different requirements at the same time. A factory automation system may need highly predictable latency and reliability, while a large public event can suddenly generate a major increase in traffic. A single fixed network configuration cannot optimise equally for every situation.

AI can provide an additional intelligence layer by analysing network conditions and helping determine how resources should be allocated between slices.

From Virtual Networks to Intelligent Network Resources

Traditional network slicing already separates services according to defined requirements. The next step is making those slices more responsive.

An AI-enabled system can continuously examine information such as traffic levels, latency, network performance and service conditions. Instead of relying entirely on manual configuration, it can help determine when a slice needs additional capacity, when resources can be released and when network policies should change.

This is moving from predefined slicing toward adaptive slicing.

The development is already being demonstrated in live-network environments. In February 2026, an industry collaboration demonstrated an agentic AI-powered 5G-Advanced slicing solution in a live 5G network. The system monitored network KPIs including bitrate and latency, combined them with contextual information such as traffic, incidents, locations, events and weather, and automatically adjusted RAN policies to meet service requirements.

The demonstration covered the RAN, transport and core, showing why AI-driven slicing is becoming an infrastructure challenge rather than a feature confined to the radio network. It also used four operating modes: chatbot, on-demand, scheduled and autonomous.

That architecture points toward a broader change in telecom infrastructure. The network is not simply providing separate virtual paths for different services. It is beginning to acquire the intelligence needed to decide how those virtual resources should respond as conditions change.

Key takeaway: Network slicing is evolving from a static virtualisation capability toward a more adaptive infrastructure layer that can respond to changing service and network conditions.

The change is therefore not about replacing network slicing with AI. It is about adding intelligence to the infrastructure that already performs the slicing, allowing 5G-Advanced networks to move closer to intent-driven and dynamically managed operation.