Australian AI infrastructure startup Firmus Technologies has finalised a US$10 billion debt financing facility led by funds managed by Blackstone Tactical Opportunities and Blackstone Credit & Insurance, with support from technology investor Coatue Management. The transaction, announced on 9 February 2026, ranks among the largest private debt financings in Australian history. The capital will fund Project Southgate — a national rollout of Firmus's AI Factory platform built on NVIDIA's DSX reference architecture — developed in partnership with CDC Data Centres and Nvidia. The initiative is designed to reach a capacity of up to 1.6 gigawatts over the next three years, serving AI training and inference workloads for enterprise and government customers.

Firmus is not a newcomer to large-scale capital raises. The company raised A$830 million in equity across two placements in 2025, backed by Nvidia and Australian investor Ellerston Capital, which established the foundation for Project Southgate. The new US$10 billion debt package represents a categorical shift in scale, moving Firmus from expansion mode into what backers describe as national-scale execution. Blackstone framed the deal as a long-term infrastructure conviction play rather than a short-term technology bet, drawing a parallel to its broader strategy of financing large AI compute and data-centre platforms globally — a strategy that also underpinned its 2024 acquisition of Sydney-based AirTrunk alongside Canada's CPPIB.

The deal carries meaningful near-term significance beyond the capital figure itself. Firmus is preparing for an IPO on the Australian Securities Exchange, with co-founder Oliver Curtis indicating to the Australian Financial Review that a listing is expected in 2026, targeting approximately A$2 billion in proceeds. The company's ability to execute at the scale it has promised — US$70 billion in projected total capital expenditure through 2028 — will be scrutinised by institutional investors ahead of that listing. Critics note that the debt-fuelled model echoes CoreWeave's aggressive US expansion, a comparison that cuts both ways: CoreWeave's revenue growth has been strong, but its post-IPO stock volatility has kept risk questions alive.

For Australia, the Firmus deal is a landmark moment in the country's ambition to be more than a consumer of AI — it signals a bid to become a producer of the infrastructure that powers it. Project Southgate's use of Tasmania's existing hydroelectric grid as a clean energy backbone addresses one of the key constraints facing AI data centres globally: the cost and carbon intensity of power. Blackstone's senior managing director explicitly stated a belief that 'Australia can play a central role' in the AI infrastructure buildout, a framing that carries weight for federal and state policymakers evaluating data sovereignty and digital industrial policy. The forthcoming ASX IPO will be the first major test of whether domestic public markets can support AI infrastructure at this scale, and its outcome will influence how the next cohort of Australian deep-tech companies thinks about capital pathways.

Firmus's $10B Blackstone Loan Positions Australia as a Global AI Compute Hub

While the Firmus deal is Australia-centred, its context is inseparable from the broader APAC consolidation underway. STT GDC — now majority-owned by KKR and Singtel — operates data centres in India as a core part of its 12-market Asia-Pacific and European footprint. As AI-driven demand for colocation and high-density compute accelerates across Indian enterprise and government sectors, the ownership change at STT GDC could influence investment priorities, capacity expansion timelines, and pricing dynamics in the Indian market. India's own data centre buildout — estimated at US$100 billion in capital expenditure — is drawing parallel interest from global infrastructure funds, and the Firmus and STT GDC deals may serve as a benchmark for the scale of financing that Indian AI infrastructure platforms could eventually attract.

India Features in the STT GDC Footprint as APAC's AI Infrastructure Consolidation Deepens

Singapore's digital infrastructure story took on new dimensions the week prior, when KKR and Singtel announced their US$10.9 billion full acquisition of STT GDC — the largest M&A transaction in Singapore in four years. The near-simultaneous arrival of the Firmus financing demonstrates that global capital is treating the two markets as complementary rather than competing nodes in the same regional AI infrastructure network. STT GDC operates across 12 markets including Singapore, India, Southeast Asia, the UK and Europe, with 2.3 GW of design capacity; Firmus targets a 1.6 GW buildout concentrated in Australia. For Singapore-based investors and infrastructure funds, the combined signal is clear: the window for positioning in APAC AI compute infrastructure is open and accelerating rapidly.

Dual $10B Deals in Two Weeks Put Singapore and Australia at the Centre of APAC's AI Infrastructure Race

australia

firmus-technologies

blackstone-tactical-opportunities

blackstone-credit-insurance

coatue-management

ai-factory

nvidia-dsx

cdc-data-centres

nvidia

ellerston-capital

airtrunk

sydney

canada

cppib

australian-securities-exchange

coreweave

Blackstone and Coatue's $10B loan to Firmus is among Australia's largest-ever private debt deals — learn what's at stake for regional AI infrastructure and the coming ASX IPO.

$10B debt deal tests Australia's AI infrastructure ambition

Nvidia-backed Firmus secures one of Australia's largest-ever private debt financings as it races toward an ASX IPO and 1.6 gigawatts of AI capacity.

The scale of the Firmus financing — one of the largest private debt deals in Australian history — signals that institutional capital is no longer treating APAC as a secondary market for AI infrastructure investment. Blackstone, which has deployed more than US$70 billion into data centre assets globally since 2021, chose Australia for a flagship credit transaction, citing energy cost advantages, regulatory flexibility, and proximity to fast-growing regional AI demand. The deal arrives weeks after KKR and Singtel's US$10.9 billion acquisition of STT GDC, underscoring that both Singapore and Australia are now anchoring separate but complementary legs of the region's AI infrastructure buildout.

For APAC investors and policymakers, the paired mega-deals represent a structural shift in how the region is being capitalised. Project Southgate's planned 1.6 GW capacity — built around Nvidia chip architecture and Tasmania's hydroelectric power — positions Australia as a potential sovereign AI compute hub, with ambitions that extend beyond domestic demand into serving hyperscale and AI-native customers globally. The critical test will come when Firmus attempts its ASX IPO later in 2026: whether public markets price the company as a durable infrastructure asset or a heavily leveraged growth bet will set a benchmark for how the region's next wave of AI infrastructure companies access capital.

Blackstone's conviction bet on an Australian AI startup reshapes the regional investment thesis for data centre capital.

Australia's $10B Firmus Deal Signals APAC Is Now Primary AI Infrastructure Territory

Can Australia's Firmus Justify a $10 Billion AI Infrastructure Bet?

Samsung Electronics reported Q1 2026 operating profit of ₩57.2 trillion (approximately $41.8 billion), an eightfold increase year-on-year that shattered analyst estimates and surpassed the company's total profit for all of 2025. Revenue reached a record ₩133.9 trillion ($90 billion), up 50% year-on-year, driven almost entirely by explosive demand for AI memory chips. The company's Device Solutions division — which houses its memory and foundry operations — accounted for more than 93% of total group profit, posting ₩53.7 trillion in operating income on chip sales that rose 225% year-on-year. Samsung's memory business set a new quarterly sales record, with high-bandwidth memory (HBM) chips for AI accelerators the primary driver.

The results crystallise a structural shift in the global semiconductor industry. Samsung has now mass-produced HBM4, the sixth generation of high-bandwidth memory, and is shipping to customers including hyperscalers and AI accelerator makers such as Nvidia. The company warned that its demand fulfilment rate has reached a record low, and that customers concerned about supply shortages are already pulling forward orders for 2027 — a signal that the AI infrastructure buildout is consuming memory capacity faster than the industry can replenish it. Samsung's foundry operations are also running at full utilisation on advanced nodes, with 2nm production ramping for the second half of 2026 and 1.4nm development on schedule.

For the APAC technology market, the implications are significant and two-sided. On one hand, Samsung's supply constraints are directly inflating the cost of memory components used in every AI server, cloud data centre, and enterprise device across the region. AWS, Google, Meta, and Microsoft — the hyperscalers driving the largest APAC cloud expansions — are among Samsung's biggest customers, meaning their aggressive infrastructure buildout is itself intensifying the supply crunch. On the other hand, Samsung's record profit validates the region as the decisive supplier of the AI era's most critical resource. The company stated that server memory demand is expected to remain strong through the second half of 2026 as agentic AI adoption accelerates, making the Korean giant's output cadence a core variable in every APAC enterprise AI roadmap.

Australia's data centre pipeline is among the most active in APAC, with hyperscalers including AWS, Microsoft, and Google committing billions to new capacity. Samsung's warning that memory demand fulfilment rates are at record lows — and that the supply-demand gap will widen into 2027 — introduces a material constraint on how quickly those facilities can be populated with AI-capable servers. Australian enterprises already navigating Gartner's forecast of A$172.3 billion in domestic IT spending in 2026 will need to factor extended memory lead times and elevated component pricing into infrastructure budgets and AI project timelines.

Memory Crunch Threatens Australia's AI Data Centre Buildout Timeline

India's hyperscale cloud buildout — supported by government initiatives and surging enterprise demand — is directly exposed to Samsung's supply warnings. The country's rapidly expanding technology services sector, which increasingly differentiates on AI capabilities, sources compute infrastructure from the same global supply chain now dominated by Samsung's constrained HBM output. With Samsung customers pulling forward 2027 orders and the company running memory fulfilment at record lows, Indian cloud providers and large-scale AI deployers face a structural cost headwind that is unlikely to ease before mid-2027. The timing is particularly challenging as Indian enterprises scale from AI pilots to production deployments.

India's AI Infrastructure Surge Collides With a Global Chip Shortage

Singapore's financial institutions, government agencies, and technology firms are among the most aggressive early adopters of agentic AI in Southeast Asia. Samsung's earnings confirm that the memory chips underpinning AI inference workloads are in critically short supply globally, with hyperscale cloud providers at the front of the queue. For Singapore-based enterprises building on AWS or Azure, this means the cost of cloud-based AI compute is likely to remain elevated, and procurement teams sourcing on-premises AI hardware face extended delivery windows. The IMDA's continued push for sovereign AI infrastructure adds further urgency: any plan to deploy locally-hosted AI systems depends on the same constrained HBM supply chain.

Singapore's AI Ambitions Run Into Samsung's Global Supply Queue

samsung

google

microsoft

hbm4

apac

Samsung's Q1 2026 operating profit hit ₩57.2 trillion — its best quarter ever, already exceeding all of 2025. Here's what the supply crunch means for APAC cloud buyers.

756% profit surge: Samsung's AI chip quarter breaks every record

A 756% profit surge in a single quarter reveals how South Korea's chip giant has become the indispensable bottleneck in the global AI infrastructure race.

Samsung's earnings call delivered an unusually candid warning: its demand fulfilment rate is at a record low, and the gap between supply and demand is expected to widen further into 2027. For APAC enterprises — from Singapore's financial institutions deploying agentic AI to Indian technology services firms building inference infrastructure — this supply crunch translates directly into longer lead times, higher server costs, and compressed margins on AI-enabled products. The hyperscalers driving APAC's cloud expansion, including AWS with its $200 billion 2026 capex plan, are themselves the largest consumers of Samsung's constrained HBM output, creating a feedback loop that disadvantages smaller regional buyers competing for the same chips.

The competitive landscape for HBM supply is also shifting. Samsung is racing to close the gap with SK Hynix, which seized an early lead in HBM3E. Samsung's announcement in February that it had begun mass-producing HBM4 chips — the most advanced generation yet — is strategically critical: whichever Korean supplier dominates HBM4 volumes will effectively control the pace at which APAC AI infrastructure can scale through 2027 and beyond. Regional procurement teams, cloud architects, and CIOs in Australia, Singapore, and India should treat Samsung's supply warnings not as a quarterly earnings footnote but as a multi-year constraint on their AI deployment timelines.

Samsung's record results are simultaneously good news for Korea and a warning signal for every enterprise building AI infrastructure across Asia Pacific.

APAC Cloud Buyers Face Widening Memory Supply Gap Through 2027

Samsung Posts All-Time Profit Record as AI Memory Demand Overwhelms Supply

Australian data center builder Firmus Technologies raised $505 million in a funding round led by Coatue Management, with participation from Nvidia, valuing the company at $5.5 billion. This marks the third equity round in six months, bringing total equity raised to approximately $1.35 billion. The company is planning an ASX IPO in June or July, seeking to raise about $2 billion in additional capital.

The funding will accelerate Project Southgate, Firmus's plan to build renewable energy-powered data center capacity across Australia, starting with a Tasmania facility. The site is designed as modular AI factories running Nvidia GPU clusters at high density, eventually housing 36,000 Nvidia GB300 Grace Blackwell chips. In February, the company secured a $10 billion debt financing facility led by Blackstone, one of the largest private credit transactions in Australian history.

Firmus exemplifies Australia's emergence as a destination for AI infrastructure investment, leveraging renewable energy resources, political stability, and proximity to Asia-Pacific markets. Tasmania's hydroelectric power provides consistent baseload energy at costs below fossil fuel alternatives, with the state generating 90% of electricity from renewables. The planned IPO would be one of the largest in Australia this decade for a tech company.

Australia has long struggled to produce globally significant technology companies, with its economy dominated by mining, banking, and real estate. A successful Firmus IPO could signal a shift in the country's economic identity, demonstrating it can compete in AI infrastructure by leveraging advantages like cheap energy and stable governance. The Australian government has signaled support for domestic AI infrastructure, allocating funding for digital infrastructure development in its 2025-2026 budget.

Firmus IPO Could Signal Australia's Tech Transformation

Firmus, now Singapore-based after pivoting from bitcoin mining to AI infrastructure, has data center projects in both Australia and Singapore. The company's regional headquarters in Singapore positions it to serve growing demand across Southeast Asia's fintech and AI sectors.

Singapore Emerges as Regional AI Infrastructure Hub

blackstone

tasmania

asia-pacific

project-southgate

nvidia-gb300-grace-blackwell

Australian startup's massive funding round led by Coatue targets Asia-Pacific AI infrastructure expansion

$505M Raise Fuels Firmus's $5.5B AI Data Center Push

Nvidia-backed data center builder reaches $5.5B valuation ahead of ASX IPO plans

The funding positions Firmus to accelerate high-density AI data center construction across Asia-Pacific, a region experiencing exploding demand for next-generation compute power with Nvidia chips powering these facilities. This aligns with the sovereign AI narrative, where governments prioritize keeping information within national borders, making Australia's stable regulatory environment increasingly attractive.

Region attracts massive capital for sovereign computing capabilities

Asia-Pacific AI Infrastructure Boom Accelerates

Firmus Raises $505M for Australia AI Factories

Malaysian Prime Minister Anwar Ibrahim revealed that the government has been quietly blocking data center projects unrelated to AI development since late 2024, citing rising energy and water consumption that could push electricity tariffs higher for ordinary consumers. The restriction specifically targets non-AI facilities to preserve grid resources as the country faces infrastructure pressure from explosive data center growth. Projects demonstrating benefits for high-technology and AI development receive easier approvals, while all new applications unrelated to AI advancement have been stopped.

Malaysia emerged as the Asia-Pacific region's fastest-growing data center market, gaining 850 MW in potential electricity demand in 2024's first half alone as capacity climbed from 10 MW in 2021 to approximately 1.3 GW by 2024. The surge followed Singapore's 2019-2022 data center moratorium, redirecting major investments from companies like Google, Amazon, ByteDance, Nvidia, and Alibaba to Malaysia's cheaper land and electricity markets. According to DC Byte analysis, over two-thirds of Southeast Asia's data center capacity under construction across five main markets is now committed to Malaysia.

The bottleneck stems from grid connection delays rather than electricity generation capacity, creating infrastructure approval challenges for new developments. Water shortages and power supply constraints in key states like Johor and Selangor have already forced authorities to slow approvals, with the government announcing a comprehensive pause on non-AI data center development. Critics warn Malaysia risks subsidizing foreign corporations' computational needs without securing commensurate returns in local innovation capacity or strategic control, as major technology companies acquire land from palm oil plantations for facilities.

Australia approved data center applications for 22 facilities with 3.67 GW combined capacity in NSW alone during 2025, equivalent to powering more than 3 million households annually and increasing pressure on available generation capacity. On March 27, 2026, Australia's Investment Delivery Authority endorsed 15 data center projects worth A$51.9 billion for prioritized government support, signaling continued commitment despite Malaysia's cautionary approach to non-AI facilities.

Australia Watches Malaysia's Grid Constraints Amid Own Expansion

India maintains Tier 2 status under US AI chip export controls with quota restrictions, while Malaysia benefits from unrestricted GPU access, creating different infrastructure investment pressures. Major global hyperscalers and domestic enterprises made significant investments and operational expansions in India during 2025, with Hyderabad emerging as a focal point for continued growth in 2026 without implementing Malaysia-style restrictions.

India Avoids Malaysia's Restrictions With Strategic Power Planning

Singapore's 230kV transmission network cannot handle 400kV connections required by 100MW+ data centers, with upgrading requiring $2 billion investment and 5-year construction timeline for just 50km of high-voltage lines. Singapore currently holds about 60% of Southeast Asia's data center capacity but faces limitations due to small size and reliance on imported energy, validating Malaysia's preemptive approach to grid management before constraints become critical.

Singapore's Moratorium Experience Validates Malaysia's Caution

malaysia

singapore

johor

selangor

southeast-asia

amazon

bytedance

alibaba

dc-byte

Southeast Asia's fastest-growing hub restricts approvals as grid strain threatens consumer tariffs.

The Asia-Pacific region faces a 165% increase in data center power demand by 2030, rising from 320 TWh in 2024 to 780 TWh, creating unprecedented collision between AI compute growth and struggling power infrastructure. Singapore previously lifted its data center moratorium with strict sustainability mandates while Japan announced plans to relocate data centers near offshore wind and nuclear sites to address similar constraints.

Chinese firms are providing critical infrastructure solutions, including Tianneng Group's 1 GWh "Solar-Storage-Computing" project announced in early 2026, offering clean energy integration while stabilizing local power supply as Malaysia's data center boom becomes intertwined with Chinese capital and infrastructure capabilities. As Malaysia struggles with grid constraints, limited water capacity, and geopolitical pressure from US tariffs, the country faces a narrow window to reinforce its grid, expand clean energy integration and secure long-term water resilience to maintain its structural advantages as ASEAN's primary data center hub.

Malaysia's selective AI-only approval process reflects broader regional grid constraints.

APAC Data Center Policy Shift Signals Infrastructure Reality Check

Malaysia Blocks Non-AI Data Centers Over Power Constraints

Malaysia's Prime Minister Anwar Ibrahim confirmed that his government has been rejecting applications for data centers unrelated to AI since mid-2024, with restrictions expected to remain in place for almost two years. The official reason for restricting non-AI data centers is resource pressure from electricity and water consumption, with concerns that continued growth could push tariffs higher for ordinary consumers. However, the moratorium is designed to protect space for high-value projects including NVIDIA's $4.3 billion AI data center partnership with YTL, Microsoft's second site acquisition, and ByteDance's $2.1 billion expansion commitment.

Since 2024, Malaysia has emerged as ASEAN's main data center hub, challenging Singapore's dominance, with 34 existing data centers and 33 upcoming facilities as of early 2026. Growth has concentrated in Johor state, bordering Singapore and offering cheaper land and lower electricity tariffs, with the state projected to hold 60% of Malaysia's total data center capacity by 2030. The government announced the pause due to water shortages and power supply constraints, but officials see this as temporary while Malaysia works to secure sustainable power and water supply pathways.

The policy reveals Malaysia's strategic ambition to position itself as a high-technology destination by selectively approving the most energy-intensive AI facilities, though questions remain about who actually benefits from foreign investment. Chinese firms like Tianneng Group's 1 GWh "Solar-Storage-Computing" project announced in early 2026 are providing both capital for facilities and technology to meet utility demands, making the partnership with China fundamental to Malaysia's ability to sustain its data center boom. Prime Minister Anwar emphasized that data center investment must deliver tangible returns to ordinary people, not just serve investor profit.

Australia received Tier 1 designation under the US AI Diffusion Rule, granting unrestricted access to Nvidia H100, H200, and Blackwell-class GPUs, while Singapore, India, and Malaysia were placed in Tier 2 with quotas, redirecting AI cluster deployment toward Australia. In one week in April 2026, over A$6 billion in committed capital flowed to Australia's data center market through NEXTDC's A$1 billion hybrid securities raise, Stockland's A$3.94 billion Western Sydney campus, and ESR's US$850 million equity raise.

Australia Benefits from Malaysia's Restrictions

Major global hyperscalers and domestic enterprises made significant investments and operational expansions in India in 2025, a trend set to continue in 2026, with Hyderabad emerging as a focal point. India encourages development with land-value rebates up to 50%, full exemptions on stamp duty, and 100% exemptions on electricity and transmission charges, with 77 Tier 3 data centres certified by 2022.

India Faces Similar Power Grid Pressures

Singapore had only 7.2 MW of available data center capacity in late 2024 with a vacancy rate of about 1%, remaining the most power-constrained data center market globally with less than 4 MW of available capacity. Singapore's two new tranches of data center development capacity totalling 1.2 GW should ensure the city-state remains a premium high-quality hub for specialized AI workloads while pushing larger requirements to neighbouring Southeast Asia.

Singapore's Capacity Constraints Drive Regional Spillover

asean

tianneng-group

solar-storage-computing

Prime Minister confirms 2024 moratorium to protect power grid and water infrastructure.

The Asia-Pacific region confronts an unprecedented collision between explosive AI compute growth and power infrastructure struggling to keep pace, with electricity consumption projected to climb from 320 TWh in 2024 to 780 TWh by 2030—a 165% increase. Singapore lifted its data center moratorium with strict sustainability mandates, while Malaysia emerged as the world's hottest data center destination despite power infrastructure concerns.

Malaysia's cost advantage is around 22% lower than Singapore and comparable to Indonesia and Thailand, combined with a skilled workforce and relatively stable power supply, giving Malaysia structural advantages to become ASEAN's primary data center hub. However, this advantage is not guaranteed as electricity demand accelerates and environmental concerns intensify, forcing Malaysia into a narrow window to reinforce its grid, expand clean energy integration and secure long-term water resilience.

Regional power constraints force governments to prioritize high-value AI infrastructure over traditional facilities.

APAC Power Grid Crisis Drives Malaysia's Selective Data Center Policy

Malaysia Bans Non-AI Data Centers for Two Years

The Australian federal government released its 'Expectations of data centres and AI infrastructure developers' on March 23, outlining how facilities should contribute to national interest, energy transition, and water security. NSW's Investment Delivery Authority endorsed 15 data centre projects worth A$51.9 billion for priority government support on March 27, following approval of 22 additional facilities with 3.67 gigawatts combined capacity in 2025. The policy shift follows Australia ranking second globally after the US as the most attractive destination for data centre investment in 2024.

Data centres in Sydney already consume 4% of NSW's grid-supplied electricity, potentially rising to 11% by 2030, with national consumption reaching 11% of Australia's total electricity by 2035. The Australian Energy Market Commission has proposed new technical standards requiring large electricity users to remain connected during grid disturbances and respond to instability, reflecting concerns about hyperscale AI facilities' impact on system stability. The policy signal is increasingly clear that data centre proponents may be expected to fund infrastructure upgrades directly.

Australia received Tier 1 designation under the US AI Diffusion Rule in January 2025, granting unrestricted access to Nvidia H100, H200, and Blackwell-class GPUs, redirecting AI cluster deployment toward Australia as Singapore, India, and Malaysia were placed in Tier 2 with quotas, driving a 30% surge in NEXTDC's forward order book. The government's new expectations framework represents Australia positioning itself as a regional AI hub while ensuring public infrastructure costs are managed.

NSW's Investment Delivery Authority endorsed 15 data centre projects worth A$51.9 billion for priority government support through the approvals process on March 27, 2026. The state had previously approved 22 additional facilities with 3.67 gigawatts combined capacity in 2025, equivalent to powering over 3 million households annually and increasing pressure on available generation capacity.

Data centre development is heavily concentrated in NSW and Victoria, particularly around Sydney and Melbourne, hosting 80% of Australia's data centre capacity. These areas face land constraints, network congestion and competing industrial demand, creating localised grid exposure when multiple large facilities seek connection to the same substations within compressed timeframes.

NSW Approves A$51.9B in Data Center Projects Amid Grid Concerns

Southeast Asia is poised for further expansion in 2026, with Malaysia gaining traction not only as an offshoot of Singapore but increasingly for local workloads as the government digitises the economy. Amid limited opportunities in Singapore, investors and operators are exploring Malaysia (Johor) as a nearby alternative, with Malaysia and Indonesia being the data centre investment hotspots of Southeast Asia.

Malaysia's strong data centre development pipeline will see its population per megawatt ratio consolidate faster than any other Asia Pacific market between now and the end of the decade, according to Cushman & Wakefield analysis. Malaysia is prioritising energy incentives as part of its strategy to attract international data centre operators.

Malaysia Emerges as Singapore Alternative for Data Center Investment

Singapore was placed in Tier 2 under the US AI Diffusion Rule, subject to quotas and Total Processing Performance caps, while Australia received unrestricted Tier 1 access to advanced GPUs. Sydney's western corridor has been absorbing hyperscale demand that Singapore's land constraints had displaced, with institutional capital formation accelerating.

Despite strong demand and supply constraints driving high utilisation rates and upward pressure on rental rates, investors and operators are exploring nearby markets, especially Malaysia (Johor) and Thailand for expansion opportunities. Singapore remains highly selective in its approach to new developments, contrasting with other APAC markets offering extended incentives.

Singapore Loses AI Investment to Australia Under US Trade Rules

india

investment-delivery-authority

australian-energy-market-commission

nextdc

nvidia-h100

nvidia-h200

nvidia-blackwell

New federal requirements demand AI facilities fund power upgrades amid surge in NSW applications.

Singapore, India, and Malaysia were placed in Tier 2 under US AI export controls, subject to quotas and processing performance caps, while Australia received Tier 1 designation with unrestricted GPU access. Australia's data centre market builds on aggressive 2025 hyperscaler infrastructure buildout, with OpenAI's recent partnership with NEXTDC announcing a AUD 7 billion hyperscale AI campus in western Sydney.

Singapore continues as a key APAC market despite power constraints promoting growth in secondary hubs like Johor, Jakarta and Delhi, while India is expected to reach 3GW capacity by 2030 with Malaysia, Indonesia, Philippines and Thailand developing regulatory frameworks to attract operators. APAC currently hosts 12.7GW operational capacity with 3.2GW under construction and 13.3GW in planning, with total capacity expected to exceed 17GW by 2026.

Tier 1 GPU access drives unprecedented investment surge across region.

APAC Data Centers Pivot to Australia After US Trade Rules

Australian Government Sets Grid Investment Expectations for Data Centers

Malaysia's Prime Minister Anwar Ibrahim announced the country will restrict development of data centers not related to artificial intelligence, with the restrictions expected to last almost two years to reduce pressure on the power grid and water supply. Anwar explicitly cited rising energy and water consumption, and flagged the risk that continued growth could push tariffs higher for ordinary consumers. The prime minister said that if data centers do have benefits in terms of AI development and technology, approvals will be easier.

EY's Asia-Pacific energy and resources leader Mark Bennett has estimated that data centres alone could require between 5 and 6 gigawatts of electricity by 2035 – equivalent to roughly one-fifth of Peninsular Malaysia's current total power capacity. The government's position is that infrastructure upgrade costs for data centres must be fully borne by developers, not socialised onto the wider grid. Malaysia is now gaining data centre capacity faster than any other state in the Asia-Pacific region, with about 850 MW in potential electricity demand announced in the first half of 2024 alone.

The investment names are familiar: NVIDIA partnered with YTL on a US$4.3 billion AI data centre, Microsoft acquired a second site, and ByteDance committed US$2.1 billion in expansion. Growing investor interest in Malaysia is largely due to Singapore's moratorium on new data centers from 2019 to 2022, during which many companies moved projects to the neighboring country. The harder work is in translating foreign compute investment into domestic ability, managing grid expansion without tariff blowback, and defining what AI sovereignty actually means in practice for Malaysia.

On 27 March 2026, the IDA endorsed 15 data centre projects worth A$51.9 billion for prioritised government support through the approvals process. Without appropriate policy settings, this growth could place pressure on electricity infrastructure planning, system reliability and ultimately consumer costs. On 23 March 2026, the Australian Government released its Expectations of data centres and AI infrastructure developers, outlining explicit government expectations about how data centres should contribute to Australia's national interest.

Australia Accelerates Grid Planning After Malaysia's Moratorium Warning

Major global hyperscalers and domestic enterprises made significant investments and operational expansions in India in 2025; a trend set to continue this year, with Hyderabad a focal point. India encourages development with land‑value rebates up to 50 %, full exemptions on stamp duty, and 100 % exemptions on electricity and transmission charges. Australia, India and Malaysia are rapidly emerging as significant markets as demand for storage, analytics and processing accelerates.

India Expands Data Centers as Malaysia Imposes Regional Constraints

In late 2024, Singapore had only 7.2 MW of available data center capacity with a vacancy rate of about 1%. In Q1 2025, the vacancy rate remained at about 2%, far below the Asia-Pacific average of 14%. Singapore lifted its moratorium on data centers in 2022 but introduced much stricter sustainability standards, constraining expansion. Singapore's two new tranches of data centre development capacity totalling 1.2 GW should ensure the city-state remains a premium high-quality hub for specialised AI workloads.

Singapore's Strict Standards Drive Regional Data Center Spillover

peninsular-malaysia

PM Anwar cites power grid strain as government selectively approves only high-tech AI facilities.

The Asia-Pacific region confronts an unprecedented collision between explosive AI compute growth and power infrastructure struggling to keep pace, with electricity consumption projected to climb from 320 TWh in 2024 to 780 TWh by 2030—a 165% increase. Singapore lifted its data center moratorium with strict sustainability mandates, while Malaysia emerged as the world's hottest data center destination despite power infrastructure concerns.

In a high digital adoption scenario, data centre electricity consumption in Asia Pacific could grow from under 200 TWh in 2025 to more than 1,000 TWh by the mid-2030s, with data centres accounting for about 2.3% of regional electricity demand by 2030. Governments across the region are scrambling to augment the energy generation, transmission and distribution infrastructure necessary to enable this.

Malaysia's selective moratorium highlights regional struggle with AI infrastructure demands.

APAC Data Center Power Crisis Intensifies as Malaysia Restricts Growth

Malaysia Imposes Two-Year Ban on Non-AI Data Centers

Firmus Technologies raised US$505 million led by Coatue Management with Nvidia co-investing, valuing the company at US$5.5 billion — its third equity raise in six months and total capital of US$1.35 billion in that period. The data center builder focuses on AI infrastructure across Asia-Pacific markets, positioning itself at the center of the region's artificial intelligence buildout. Today's rounds highlight a clear shift toward infrastructure-first investing—from AI chips and GPU cloud platforms to spacecraft mobility and carbon-negative fuels, as demand for compute, automation, and real-world AI applications accelerates.

The funding comes as Asia Pacific venture capital reaches $73.6 billion across 4,308 transactions in the past year, with investors putting $27.4 billion to work in Q1 alone — up about 20% from the prior quarter and nearly double year-ago levels. AI, autonomous vehicles, robotics, and industrial enablement are the primary drivers of venture capital investment heading into 2026, giving founders building in these verticals a structural tailwind. Investment to startups in India, Singapore and South Korea rose sequentially in Q1, while the quarterly numbers show increasing momentum in the regional startup ecosystem.

The region will spend over US$437 billion on acquiring new technology between 2025 and 2030, with Forrester estimating total spending on technology will grow by 9.3%, driven by investments in software, services, communications equipment, and tech outsourcing. India is forecast to lead major Asia Pacific markets with technology spending growth of 13.4%, followed by China at 10.7%, Australia at 8.6% and Singapore at 6%. Firmus's massive funding round signals investor confidence in the infrastructure backbone needed to support this expansion.

Australia's technology spending is expected to reach nearly AUD $110 billion in 2026, with software prices rising at almost five times the pace of general inflation as vendors add AI features to contract renewals. Firmus's Australian operations are positioned to capture significant market share as local enterprises accelerate their AI infrastructure investments. The company's $5.5 billion valuation makes it one of Australia's most valuable tech unicorns, reflecting the critical role of data center infrastructure in the nation's digital transformation.

Australian Data Center Unicorn Reaches $5.5B Valuation

India leads Asia Pacific with 13.4% technology spending growth, propelled by rapid cloud adoption and data localization rules driving major onshore infrastructure investment. India received $3.8 billion in Q1 investment, the highest in four quarters, with a significant portion going to AI infrastructure including a $600 million financing for AI systems developer Neysa. While Firmus hasn't disclosed specific India operations, the country's massive AI infrastructure requirements present significant expansion opportunities for the newly-funded company.

India AI Infrastructure Demand Drives Regional Investment

Singapore's market is expected to be supported by AI transformation spending and hyperscaler expansion, though a shortage of skilled workers remains the main brake on technology adoption. Singapore-based startups are expected to raise over $18.4B in new funding in 2026, driven by rising investor confidence, strong government support, and more global VCs entering the market. Firmus's Singapore presence positions it to serve as a regional hub for Southeast Asian AI infrastructure deployment.

Singapore Hub Strategy Attracts Record Infrastructure Investment

south-korea

china

forrester

Australia-Singapore data center firm attracts Nvidia backing in largest APAC tech funding of 2026.

Australia, China, India, Singapore, and Saudi Arabia all feature in this week's top funding rounds, reflecting the breadth of Asia-Pacific and GCC startup ecosystems driving AI infrastructure investment. The funding positions companies to meet the growing demand for AI compute in the Asia-Pacific region, where specialized AI infrastructure startups are rapidly gaining traction.

China's AI infrastructure spending will exceed US$70 billion in 2026, fueled by major investments from Alibaba and ByteDance as well as government industrial digitalization mandates, though weak domestic demand will slow traditional enterprise IT spending. India, as Asia Pacific's fastest-growing market, is being propelled by rapid cloud adoption and data localization rules driving major onshore infrastructure investment.

Regional startups raise over $1.2B this week as infrastructure funding dominates.

APAC AI Infrastructure Investment Surge Continues

Firmus Raises $505M in Record AI Infrastructure Round

Amazon CEO Andy Jassy disclosed Thursday that AWS AI services now carry over $15 billion in annualized revenue run rate, marking the first time the company has revealed financial returns from its artificial intelligence investments. The figure represents roughly 10% of Amazon Web Services' $142 billion revenue run rate and comes as the company projects $200 billion in capital expenditure for 2026, primarily focused on AI infrastructure. The disclosure was part of Jassy's annual shareholder letter defending the company's unprecedented spending.

The AI revenue milestone provides crucial validation for Amazon's massive infrastructure investments amid investor concerns about returns. Three years after AWS launched commercially, it had a $58 million revenue run rate, while three years into this AI wave, AWS's AI revenue run rate is over $15 billion—nearly 260 times larger. Amazon's custom chip business, including Graviton, Trainium, and Nitro components, has also doubled to over $20 billion annual revenue run rate, reducing reliance on external suppliers like Nvidia.

The announcement signals a pivotal moment for the cloud computing industry as AI transitions from experimental to revenue-generating at scale. The AI run-rate is strong validation that AWS is successfully turning the AI boom into real, high-growth revenue, positioning Amazon as a leader in AI infrastructure. With customer commitments already securing a substantial portion of 2026 capex spending, Amazon appears well-positioned to monetize its AI investments through 2027-2028.

Amazon's AI revenue breakthrough provides Australian businesses with access to proven, enterprise-grade AI infrastructure that local providers cannot match in scale or capability. With AWS already serving major Australian enterprises and government agencies, the $15 billion revenue validation could accelerate AI adoption across sectors from mining to financial services, where data sovereignty and compliance requirements favor established cloud providers with local presence.

Australian Enterprises Benefit from AWS AI Scale

Amazon's AI revenue milestone presents opportunities and challenges for India's technology sector. While Indian enterprises gain access to world-class AI infrastructure, local cloud providers and AI startups face intensified competition from AWS's proven revenue model. India's large developer ecosystem and cost advantages in AI services could create partnership opportunities, but the scale of Amazon's AI business demonstrates the capital requirements for competing in enterprise AI infrastructure.

India's Tech Sector Weighs AWS AI Opportunities

Singapore's role as AWS's APAC headquarters becomes more strategic as the company's AI services generate massive revenue. The city-state's digital government initiatives and smart nation programs could benefit from AWS's proven AI capabilities, while local startups and enterprises gain access to infrastructure that supports scaling AI applications regionally. Singapore's regulatory environment and data protection frameworks align well with enterprise AI deployment requirements.

Singapore Positions as APAC AI Hub Amid AWS Growth

graviton

trainium

nitro

Tech giant reveals AI services generating record revenue amid massive infrastructure investment.

Amazon's $15 billion AI revenue disclosure intensifies competitive pressure on APAC cloud providers who have been building regional AI capabilities. The scale of AWS's AI monetization—equivalent to many entire national tech sectors—demonstrates the infrastructure requirements and customer demand driving the global AI economy. Regional providers must now prove they can deliver comparable AI services while competing against Amazon's massive capital deployment and established enterprise relationships.

For APAC enterprises, Amazon's AI revenue milestone signals the maturation of cloud-based AI services, potentially accelerating enterprise adoption across the region. The validation of AI as a revenue-generating business model, rather than experimental spending, could drive increased C-suite confidence in AI investments throughout Asian markets, from Singapore's financial services to India's technology sector.

Arc Brief

Nvidia

News Coverage

Can Australia's Firmus Justify a $10 Billion AI Infrastructure Bet?

Samsung Posts All-Time Profit Record as AI Memory Demand Overwhelms Supply

Samsung Posts All-Time Profit Record as AI Memory Demand Overwhelms Supply

Firmus Raises $505M for Australia AI Factories

Malaysia Blocks Non-AI Data Centers Over Power Constraints

Malaysia Bans Non-AI Data Centers for Two Years

Australian Government Sets Grid Investment Expectations for Data Centers

Malaysia Imposes Two-Year Ban on Non-AI Data Centers

Firmus Raises $505M in Record AI Infrastructure Round

Amazon AWS AI Revenue Hits $15 Billion Milestone