Freyr AI Expands Southeast Asia Colocation Capacity 164%

A very important question sits behind all these numbers: What does Thailand actually gain from this data center investment wave? THB 750 billion of approved investment and 3,400 MW of approved IT load are significant. But for me, investment value and MW alone should not be the final measurement of success. The bigger question is how much long-term value remains in Thailand. Does this investment create: • Thai technical and management capability? • Local supply-chain opportunities? • Better power and digital infrastructure? • Access to AI and compute capacity for Thai businesses? • More renewable-energy development? • Tax and economic value beyond the construction phase? • Opportunities for Thai companies to participate in the AI infrastructure ecosystem? • Infrastructure that strengthens Thailand’s own digital and AI capabilities? A hyperscale data center can bring billions of baht of CAPEX, but it does not automatically translate into the same level of local economic impact. Much of the equipment is imported. Data centers are highly automated. And the value created after construction depends heavily on who owns the infrastructure, who operates it, who supplies it, who uses the compute, and where the economic value ultimately flows. This is why I think Thailand is entering an important second phase. The first phase was: How do we attract global data center investment? The next phase should be: How do we convert that investment into capability for Thailand? That means looking beyond approved MW toward local talent, energy infrastructure, technology transfer, domestic suppliers, AI compute access, and the development of a stronger Thai digital ecosystem. Foreign investment is important. But the strongest outcome is not simply having global infrastructure physically located in Thailand. It is when that infrastructure also helps build Thailand’s own capability around it. That, in my view, is the real opportunity behind this data center investment cycle.

Thailand’s data center market is moving beyond traditional colocation into large-scale, multinational digital infrastructure. As of September 2026, BOI data indicates: • 42 approved projects during 2024–2026 • Approximately 3,400 MW of approved IT load • Around THB 750 billion in approved investment • Investors from 11 nationality groups Thailand leads by project count with 15 projects, followed by China with 10, the United States with 5, the UAE with 4, and Japan with 2, alongside investors from Singapore, Hong Kong, India, Australia, France and Malaysia. However, the registered name of a Thai project company does not always reveal the true source of capital. Many investments are structured through local SPVs, Singapore holding companies and regional subsidiaries. This is why both BOI and DBD data matter: • BOI shows approved investment and ultimate investor nationality. • DBD reveals the Thai entities, SPVs, capital increases and corporate structures behind the projects. Three points stand out: 1. Thailand is attracting diversified global capital. The market is not dependent on one nationality or hyperscaler. 2. Ownership structure matters. A Singapore-registered company may ultimately be controlled by investors from China, the U.S. or elsewhere. 3. Execution is now the key challenge. The critical questions are: Where is firm power available? How quickly can grid capacity be delivered? What clean-energy options exist? What local economic value will be created? Who will ultimately use the capacity? Thailand has entered the hyperscale investment cycle. The next challenge is converting approved projects into operational, sustainable and commercially utilized infrastructure. AI & DC Solutions Building the infrastructure behind Thailand’s AI and digital economy.

IREN: AI INFRASTRUCTURE EXPANSION DRIVES WALL STREET OPTIMISM • Wall Street analysts covering IREN currently show generally positive sentiment, with consensus assessments ranging from moderate buy to strong buy according to the information provided. • Average analyst price targets are shown in the approximately $77 to $84 range, although individual forecasts vary and can change as business conditions evolve. • One of the biggest developments in the IREN story is the company's expansion beyond its Bitcoin-mining roots into AI infrastructure and high-performance cloud computing. • IREN is leveraging its existing data-center footprint, power capacity and computing infrastructure to pursue growing demand for AI workloads. • Relationships and deployments involving major technology companies such as NVIDIA and Microsoft have increased attention on IREN's ability to participate in the expanding AI data-center ecosystem. • The company is also reporting multi-billion-dollar contracted annualized revenue associated with upcoming capacity, highlighting the potential scale of its AI infrastructure expansion. • Increasing demand for accelerated computing could create additional opportunities as enterprises and hyperscalers seek access to GPUs, data-center capacity and the power required to operate increasingly demanding AI systems. • Several financial firms have recently maintained positive ratings or increased their price targets, reflecting expectations for continued growth in IREN's AI and high-performance computing operations. • The opportunity comes with substantial execution risk. Building AI infrastructure requires enormous capital expenditures for GPUs, data centers, electrical systems, networking and cooling. • IREN can also experience significant stock-price volatility as investors reassess growth expectations, financing requirements, capital spending and the economics of its expanding infrastructure platform. • The investment story therefore centers on whether IREN can successfully convert its power and data-center assets into durable AI infrastructure revenue while managing the considerable capital required for expansion. • Analyst ratings and price targets represent opinions and estimates, not guarantees of future performance. For informational purposes only. Not investment advice.

China’s AI Infrastructure Boom: The Real Race Is Compute + Power + Speed SemiAnalysis’ new bottom-up mapping of China’s data-centre market challenges a common simplification: that China has many underutilised data centres and therefore excess capacity. Both can be true. China can have legacy vacancy and an AI-capacity shortage at the same time. SemiAnalysis tracks 1,000+ facilities across 60+ operators and estimates >24 GW of delivered capacity in China — larger than EMEA or the rest of APAC individually. Yet much of the older stock was built for low-density retail workloads and is poorly matched to today’s AI racks. New demand is for something very different: wholesale, high-density, power-rich infrastructure. AI is reshaping the market Combined ByteDance/Alibaba/Tencent/Baidu capex rose from about US$35B in 2024 to >US$50B in 2025 and is reportedly on course toward US$100B in 2026. GDS and VNET alone signed ~1.3 GW of wholesale capacity in 1H26. This is not simply “more data centres”. It is a change in the physical architecture of digital infrastructure. Power is becoming geography China’s “Eastern Data, Western Compute” strategy is shifting compute toward regions with abundant land and cheaper electricity. Inner Mongolia is emerging as a major AI hub because power can cost roughly half Tier-1-city levels. AI training loosens the historic link between compute and population centres. Once fibre latency becomes sufficiently low, electrons, land and grid access matter more. Speed is a competitive capability SemiAnalysis reports that China can routinely deliver ~100 MW facilities in around 12 months, with some projects faster. Prefabrication, steel-hall construction, parallel MEP production and modular deployment compress the programme substantially. Industrialisation of data-centre construction is becoming part of AI competitiveness. The real engineering question: useful AI, not just MW The next phase will not be won by whoever announces the most gigawatts. The more meaningful metric is how much useful AI compute can be delivered per constrained unit of electricity, water, land, capital and carbon. That pushes the engineering frontier toward: • high-density liquid cooling; • warm-water operation and lower cooling overhead; • modular power and thermal infrastructure; • grid-aware workload orchestration; • higher utilisation of installed compute; and • co-design across chips, racks, cooling, power and the grid. The AI data centre is increasingly one integrated machine. China’s build-out shows what happens when compute policy, grid investment, manufacturing capacity and hyperscale demand move together. For the rest of Asia, including Singapore and Johor, the key question is which elements can be adapted to deliver AI infrastructure that is faster, denser, more efficient and more sustainable. #AI #DataCenters #AIInfrastructure #LiquidCooling #EnergyEfficiency #Sustainability #DigitalInfrastructure #China https://lnkd.in/gh5Z4thc

BROADCOM: AI INFRASTRUCTURE GROWTH CONTINUES TO BUILD Broadcom remains one of the major semiconductor and infrastructure technology companies benefiting from the enormous investment flowing into artificial intelligence. Its combination of custom AI silicon, high-speed networking and infrastructure software gives the company exposure to several critical layers of the AI buildout. • AI semiconductor revenue has been expanding rapidly as hyperscale customers increase spending on specialized accelerators and networking infrastructure. • Broadcom has established a significant position in custom AI silicon, designing specialized accelerators for some of the world's largest technology companies. • Alphabet is among the major technology companies associated with Broadcom's custom AI semiconductor business, demonstrating the importance of specialized chips alongside traditional GPUs in large-scale AI computing. • Networking is another important part of the opportunity. Massive AI clusters require extremely fast connections between accelerators, servers and data centers, creating additional demand for Broadcom's semiconductor technology. • Broadcom's opportunity extends beyond chips. Its infrastructure software business, strengthened by VMware, provides another major source of revenue and cash generation. • Strong free cash flow and high operating profitability provide Broadcom with substantial financial resources to invest in innovation, reduce debt and return capital to shareholders. • The bullish case centers on continued hyperscaler spending, accelerating demand for custom AI accelerators and the enormous networking requirements created by increasingly powerful AI systems. • Risks remain. Broadcom trades at a substantial valuation, and a meaningful portion of its AI opportunity depends on spending by a relatively concentrated group of very large technology customers. • Investors will be closely watching Broadcom's upcoming earnings for evidence that AI semiconductor demand and custom silicon deployments continue to accelerate. • The larger investment thesis is that AI infrastructure will require far more than GPUs alone. Custom accelerators, networking, connectivity and infrastructure software are all becoming critical components of the global AI buildout — areas where Broadcom has positioned itself as an important supplier. Broadcom's long-term opportunity could ultimately depend on how successfully it converts today's extraordinary AI infrastructure spending cycle into sustained revenue, earnings and free-cash-flow growth. For informational purposes only. Not investment advice.

This is exactly what “Smarter AI for All” looks like in practice. AI at scale is no longer only about GPUs and compute. It starts with the fundamentals power, cooling, water, data-center design and sustainability and then brings together infrastructure, platforms, services and real business use cases. Great insights from Linda Yao , who continues to drive Lenovo’s Hybrid Cloud & AI vision with a very practical message, AI needs to be deployable, scalable, sustainable and economically viable from edge to data center to cloud. Lenovo’s Hybrid AI strategy is increasingly focused on taking AI from experimentation into production and measurable business outcomes. Proud to be part of Lenovo mission to bring smarter AI to everyone, everywhere building the AI factories and Hybrid AI foundations that will power the next generation of innovation. #Lenovo #SmarterAIForAll #HybridAI #AIFactory #AIInfrastructure #SustainableAI #AgenticAI #DataCenter #ArtificialIntelligence

AI infrastructure planning now starts with land, power, and water, not GPU count. On-prem deployments are running into a physical ceiling before they hit a compute one. At the Six Five Summit: AI Unleashed 2026, Patrick Moorhead opened the Sustainability Track with Linda Yao, VP and General Manager of Hybrid Cloud and AI Solutions at Lenovo's Solutions and Services Group. "The site, the power source, and the cooling strategy now matter just as much as the compute." — Linda Yao, VP and General Manager, Hybrid Cloud and AI Solutions, Lenovo Key Insights: 🔹 Power and water availability now set the ceiling on AI capacity, ahead of GPU count. 🔹 On-premises AI deployment can run up to eight times more cost-effective than cloud token-maxing when planned holistically from the start. 🔹 Rack density has more than doubled since 2021, pushing liquid cooling into standard requirement territory. 🔹 Neptune liquid cooling customers have cut energy costs 40% and sustained 10% higher performance. 🔹 Circularity practices let enterprises extend infrastructure life without a full AI refresh. Enterprises that plan for power from the beginning will be positioned to scale AI demand for years ahead. Those that treat it as an afterthought will find it limits how much capacity they can bring online. Watch the full conversation: https://lnkd.in/gxdh-nRD

Private AI is becoming the control plane for Enterprise AI. The Enterprise AI debate has spent too much time on models and GPUs - and not enough on control. - Where does your data live? - Where does inference happen? - Who controls the infrastructure? - And what ultimately limits your ability to scale? Increasingly, the answer comes down to power, cooling, water, and the physical infrastructure underneath AI. That makes On-prem Private AI strategically important. At the Six Five Summit: AI Unleashed 2026, Patrick Moorhead and Linda Yao of Lenovo discussed the infrastructure realities enterprises need to confront as AI moves from experimentation to production. This is bigger than a data-center discussion. Private AI gives enterprises greater control over their data, economics, performance, security - and ultimately their AI capacity. But Private AI only works at enterprise scale when infrastructure is designed around it from the beginning. Compute is just one piece. Power. Cooling. Data. Security. Lifecycle. Economics. That is the foundation of Enterprise AI. The future won't be cloud vs. on-prem. It will be Enterprise AI built around the right workloads - with Private AI at the core and hybrid cloud extending its reach. Watch the full conversation: https://lnkd.in/gxdh-nRD #PrivateAI #EnterpriseAI #AIInfrastructure #OnPremAI #HybridAI #DataCenters #AI

China’s AI data center market has reached an inflection point, with over 24GW of operational capacity across 1,000+ facilities, surpassing EMEA and APAC ex-China. The Chinese AI Infrastructure Boom: Introducing the SemiAnalysis China Datacenter Model highlights how AI demand from ByteDance, Alibaba, Tencent, and Baidu is transforming a historically retail-focused sector into a hyperscale, wholesale-driven market. Government-led Eastern Data, Western Compute (EDWC) policies, abundant power, rapid permitting, modular construction, and lower costs enable China to deliver 100MW-scale facilities within 12 months, positioning the country as a formidable competitor in the global AI infrastructure race. 1. China has emerged as the world's second-largest data center market, operating more than 24GW of capacity, supported by strong hyperscaler investment and state-backed infrastructure expansion. 2. AI demand is reshaping the sector, with wholesale AI facilities rapidly filling while many legacy retail facilities remain underutilized due to inadequate power density and outdated designs. 3. EDWC has successfully redirected infrastructure growth inland, where regions such as Inner Mongolia benefit from cheaper power, abundant land, and favorable policies, becoming China's equivalent of a hyperscale capacity hub. 4. China's competitive advantage lies in speed and execution, combining modular construction, streamlined approvals, and domestic supply chains to deploy AI-ready infrastructure significantly faster and cheaper than many Western markets. Forward-Looking Question (Current Situation) 1. As export controls continue to constrain advanced GPU availability, will China's AI industry sustain its current infrastructure expansion pace, or will compute shortages become the primary bottleneck despite abundant power and data center capacity? Challenging & Inspiring Future Development Question 2. If China can consistently deploy hyperscale AI campuses in less than a year while scaling domestic AI models and semiconductor capabilities, how might this reshape the global balance of AI infrastructure leadership by 2030, and what strategic responses should Singapore, Malaysia, and other APAC data center markets make today to remain competitive?

APAC Data Centre Weekly Snapshot | 25 August–8 September 2026 Another busy fortnight across the APAC data centre sector, with significant company-led announcements spanning AI infrastructure, power generation, investment and next-generation computing. Here are six stories that caught our attention: Firmus Technologies has signed a multi-year agreement to provide OpenAI with AI computing capacity from two Malaysian data centres. The agreement takes Firmus’s contracted capacity beyond 900MW and will involve the deployment of Nvidia’s next-generation Vera Rubin processors across APAC. DayOne Data Centers and TNB Genco Power Generation are exploring up to 1.5GW of dedicated on-site power generation and battery storage for DayOne’s planned Selangor campus. The project would significantly expand DayOne’s Malaysian presence beyond Johor and demonstrates the increasingly important relationship between data centre development and power strategy. Keppel DC REIT and Keppel have agreed to acquire a 90% effective interest in two hyperscale data centres in Greater Tokyo for approximately US$1.19 billion. The acquisition is expected to increase Japan’s contribution to the REIT’s portfolio rental income from approximately 9% to 23%. Tata Consultancy Services subsidiary HyperVault AI Data Center and its partners plan to invest up to US$7.4 billion in a 1GW AI data centre campus in Hyderabad. The 264-acre development will support high-density GPU infrastructure for AI training and inference. Key ASIC Inc. and CT Vision have signed an agreement to develop a 300MW green AI data centre in Malaysia, targeting operation through 100% renewable energy. The partnership combines Malaysian semiconductor expertise with Hong Kong-based investment and green-technology capabilities. Diraq and Equinix are preparing to deploy the world’s first silicon-spin quantum computer inside a shared commercial data centre. Scheduled for installation at an Equinix facility in Sydney by October 2026, the system will operate alongside conventional AI, cloud and high-performance computing infrastructure while consuming less than 20kW of power. From a talent perspective, the continued acceleration of AI-ready infrastructure is intensifying competition for professionals with experience across critical facilities, hyperscale operations, construction delivery, commissioning, power strategy and high-density cooling. Singapore remains the region’s central hub for leadership, investment and connectivity, while some of the fastest-growing demand for operational and project talent is emerging across Malaysia, Indonesia, India, Japan and Australia. For operators expanding across multiple APAC markets, securing experienced leaders early will be every bit as important as securing land, capital and power. #NuboSearch