In the 100 days between April 28 and August 6, Blaize formed a hybrid AI inference alliance with Nokia and Datacomm Diangraha, Firmus and NVIDIA unveiled a 170,000-GPU data center in Batam, CoreWeave committed 360MW for its first Asia-Pacific capacity, and Indosat launched the 1GW Zankore AI factory platform with Ooredoo, Nokia, and NVIDIA. Futurum views the convergence as the assembly of a hub-and-spoke model for AI infrastructure in Indonesia with AI factories generating intelligence at gigawatt scale, and a distributed inference fabric delivering it across thousands of sites, with Blaize wiring the distribution layer first.
What Is Covered in This Article:
- The Blaize, Nokia, and Datacomm Diangraha alliance to deploy hybrid AI inference across Indonesia and Southeast Asia, validated at Nokia’s Network Innovation Lab in Singapore, with Datacomm reporting a 50%+ surge in customer AI inference demand over six months
- The centralized buildout: CoreWeave’s 360MW across three data centers by 2028, the Firmus-NVIDIA 170,000-GPU, 360MW facility in Batam targeting Q1 2027, and the Zankore by Indosat platform targeting 1GW of NVIDIA DSX AI factory capacity with roughly 200MW in H1 2027
- Indonesia’s national AI strategy: the 2025-2045 roadmap from the Ministry of Communication and Digital Affairs, the Danantara-led Sovereign AI Fund planned for 2027-2029, and the Indosat Ooredoo Hutchison and Twimbit projection that sovereign AI adds up to $140 billion to GDP by 2030
- Nokia’s position on both sides of the split, as networking partner inside the NVIDIA-aligned Zankore factory platform and as Blaize’s channel to the distributed enterprise edge
The News: In April, Blaize (Nasdaq: BZAI) announced a three-way strategic collaboration with Nokia and PT Datacomm Diangraha, one of Indonesia’s leading IT service providers, to deploy hybrid AI inference infrastructure across Indonesia and the wider Asia-Pacific region. The alliance combines Nokia’s networking backbone for GPU-intensive telco workloads, Blaize’s programmable, energy-efficient inference platforms for the enterprise edge, and Datacomm’s 30 years of Indonesian infrastructure expertise, with a joint Nokia-Blaize reference architecture validated at Nokia’s Network Innovation Lab in Singapore. Datacomm reports a 50%+ surge in customer AI inference demand over the past six months, and the partners position Indonesia as the reference market for expansion into Vietnam, the Philippines, and the wider Southeast Asian corridor. “GPU economics simply do not scale to thousands of distributed sites — but Blaize does,” said Dinakar Munagala, CEO of Blaize.
The alliance opened a cascade of commitments to the same market. In late June, NVIDIA-backed Firmus announced a 170,000-GPU, 360MW AI data center in Batam, built by DayOne and targeting operations in Q1 2027. On August 4, CoreWeave announced 360MW of contracted IT power across three Indonesian data centers, its first capacity commitment in the Asia-Pacific region, expected online in 2028. Two days later, Indosat Ooredoo Hutchison, Ooredoo Group, Nokia, and NVIDIA launched Zankore by Indosat, a full-stack platform targeting 1GW of NVIDIA DSX AI factory capacity beginning with roughly 200MW in Jakarta in H1 2027.
From Blaize to NVIDIA: Why Is AI Infrastructure Converging on Indonesia?
Analyst Take: Four major commitments to AI infrastructure in Indonesia landed within 100 days, from vendors as different as a sub-10-watt edge inference specialist and the world’s largest accelerated computing franchise. Futurum views what is being assembled in Indonesia as a single operating model that treats AI infrastructure the way electrification treated power, as generation plus distribution. AI factories generate intelligence at gigawatt scale. A distributed inference fabric delivers it to the thousands of sites where data originates and decisions execute, across an archipelago of more than 17,000 islands and roughly 285 million people whose priority workloads, agricultural monitoring, port logistics, public safety, and smart city operations, run where the data is generated. The 100-day cascade supplied every component: Zankore and Firmus the committed generation capacity, CoreWeave the merchant capacity layer, the Blaize-Nokia-Datacomm alliance the distribution fabric, and Danantara’s Sovereign AI Fund the financing vehicle to buy both. Vendors that package the assembled model with a replicable blueprint, pre-validated architectures, local operators, and sovereign capital will set the terms of AI infrastructure buildouts across the Global South.
The Generation Layer Attracted 1.7GW of Targets in 100 Days
The generation layer filled fastest, and nearly all of it aligns with NVIDIA. Zankore by Indosat is the most ambitious entry: a platform sponsored by Ooredoo Group as lead investor, targeting 1GW of NVIDIA DSX AI factory capacity on GB300 NVL72 racks, with Nokia supplying AI-native networking and roughly 200MW planned for Jakarta in H1 2027 before regional expansion. NVIDIA claims its DSX MaxLPS design delivers up to 40% additional compute within the same power budget, a vendor figure Futurum has tracked since GTC 2026 as part of the DSX reframing of infrastructure competition from cores per dollar to tokens per megawatt. Firmus brings the largest single facility: 170,000 GPUs and 360MW in Batam, constructed by DayOne with backing from Coatue and SoftBank, with GPU deliveries running from Q1 2027 into early 2028 and company projections of up to $30 billion in revenue over the first six years based on committed customer agreements. CoreWeave committed 360MW across three owned-and-operated facilities, its first Asia-Pacific capacity, though not online until 2028. “CoreWeave goes where we can bring AI pioneers the capacity, performance, and reliability they need to build and scale their ideas,” said Sachin Jain, Chief Operating Officer of CoreWeave. Every megawatt in this layer is a target, and Indonesia’s grid, permitting, and power pricing will decide which targets convert.
Indonesia’s National Strategy Is Structured to Buy Both Tiers
The strategy behind that demand is deliberate and funded. Indonesia’s Ministry of Communication and Digital Affairs, under Minister Meutya Hafid, launched a national AI roadmap in August 2025 with a 2025-2045 horizon, phased into 2025-2027, 2028-2035, and 2035-2045 stages. Its infrastructure pillar specifies GPU and TPU expansion, a national cloud hosted in sovereign data centers, and green data center development through public-private partnerships. Danantara, the sovereign wealth fund, is tasked with spearheading AI financing and standing up a Sovereign AI Fund with blended financing models between 2027 and 2029.
The economic stakes attached to the program are large. Sovereign AI alone is forecast to add up to $140 billion to Indonesia’s GDP by 2030, equivalent to annual economic growth of up to 6.8%, according to the Empowering Indonesia Report 2025 by Indosat Ooredoo Hutchison and Twimbit, a projection from an interested operator that should be read as a ceiling rather than a base case. Indonesia’s AI sector is expanding at a 31% CAGR, the fastest rate in Southeast Asia, and 89% of enterprises rank digital sovereignty as a key factor in technology decisions. The talent pillar targets 100,000 AI-trained workers annually and 20 million AI-literate citizens by 2029. Geography is the strategy’s quiet constraint. A national cloud in Jakarta serves a bank’s core systems yet may not serve the roadmap’s priority sectors of food security, logistics, and public services.
Distributed Inference Is the Tier Where GPU Economics Break
This is the gap the Blaize alliance targets, and the demand signal preceding the centralized announcements came from the distribution channel rather than the cloud. Datacomm Diangraha, a Nokia partner founded in 1990 with more than 450 employees serving enterprise, telecommunications, government, and military customers, reported a 50%+ surge in customer AI inference demand over the past six months, a company-supplied figure that nonetheless comes from purchase conversations rather than market projections. In a briefing with Futurum, Blaize leadership described the workload composition behind that surge: geospatial computer vision first, then video analytics aggregated across thousands of nodes operating as micro-sites, with some deployments required to run air-gapped. The three-way stack assigns Nokia the networking backbone from optical interconnect to last-mile edge connectivity, validated with Blaize’s compute as a joint reference architecture at Nokia’s Network Innovation Lab in Singapore, and assigns Datacomm the in-country deployment. Munagala’s claim that GPU economics do not scale to thousands of distributed sites alludes to Blaize’s Graph Streaming Processor architecture, which spans 42U rack-scale systems down to modules drawing under 10 watts. The claim that its cost and power profiles beat GPU infrastructure at distributed scale awaits third-party benchmarking.
Futurum’s research supports the hybrid case for Sovereign AI deployment. The February 2026 Futurum report Sovereign AI: What Nations Want (And What They’ll Actually Get), by Nick Patience and Fernando Montenegro, concluded that full-stack national sovereignty is unachievable and that nations are settling for strategic autonomy, with hybrid and edge deployments growing as public cloud market share peaks. Blaize leadership made a parallel observation in our briefing: sovereign no longer means only a country, since banks, investment offices, and regional governments now demand independence from hyperscale cloud providers for specific models and use cases. Blaize wrapping its hardware in a video-analytics-as-a-service model and an AI services API matters as much as the silicon.
Futurum expects sovereign AI programs globally to formalize this split by 2028. National tenders will separate centralized AI factory capacity from distributed inference networks as distinct procurement categories, much as telecom regulators long separated core networks from access networks. Blaize is running the same playbook beyond Indonesia, with a $120 million minimum 18-month contract with Starshine Computing Power Technology spanning India, Indonesia, Japan, South Korea, and China signed in July 2025, a June 2026 MOU with Korea’s SPHERE AX signed at the National Assembly in Seoul, and stated ambitions in the Middle East and Africa. Indonesia serves as the reference deployment the partners intend to replicate in Vietnam and the Philippines.
Nokia Sits on Both Sides of the Split
One vendor appears in both tiers, and its position will shape how the split resolves. Nokia supplies AI-native networking to the Zankore factory platform alongside NVIDIA, which took a $1 billion stake in Nokia in October 2025 to co-develop AI-RAN, and Nokia simultaneously anchors the Blaize alliance at the distributed edge through the January 2026 partnership and the Singapore lab validation. “Nokia’s AI-native networking provides trusted connectivity underpinning this AI infrastructure generation,” said Justin Hotard, CEO of Nokia, in the Zankore announcement. Nokia connectivity is the layer both tiers must buy, and the company monetizes GPU factories and edge micro-sites with the same portfolio. For Blaize, Nokia’s presence validates the hybrid architecture Blaize needs enterprises to adopt, and Nokia’s deepening NVIDIA alignment means the flagship channel partner also distributes the incumbent’s path to the same telco edge. The signal to track is whose compute populates Nokia’s reference architectures as they replicate across markets. Blaize as the default enterprise-edge silicon would confirm the co-expansion thesis, while NVIDIA Jetson slotting into those blueprints would suggest the channel conflict risk is materializing.
Blaize’s $130 Million Guidance Prices the Thesis Before the Execution
The bear case is arithmetic. Blaize reported Q1 2026 revenue of $2.7 million, up 172% year over year, while reaffirming full-year guidance of $130 million, which requires more than $127 million across the remaining three quarters. Gross margin of 58% and a $35 million equity raise in May improve the footing, and the NeoTensr contract, with $70 million of total potential value, was expected to begin fulfilling at $11 million or more in Q2. Guidance still depends on a second-half ramp with little precedent for a company at this scale, and Datacomm’s contribution is not expected until Q4 2026. The alliance structure compounds the risk with counterparties concentrating Blaize’s pipeline, Starshine, and NeoTensr, among them, that lack the balance sheet depth of the hyperscale customers funding the centralized tier.
Crowding adds a new pressure. Three centralized entrants are now competing for the same national budget and the same grid capacity, and Zankore’s Ooredoo capital plus Firmus’s committed customer agreements could absorb sovereign procurement before distributed inference tenders are written. The competitive field is also converging on the edge tier itself: NVIDIA’s Jetson Thor line targets the same distributed vision and robotics sites, Qualcomm is pushing its Dragonwing IQ series into industrial edge AI, and Hailo and Axelera are underpricing incumbents in vision inference silicon. Microsoft’s $1.7 billion Indonesia cloud and AI commitment and AWS’s multibillion-dollar Jakarta region give hyperscalers a route to serve edge workloads from the center via services, which would shrink the addressable share of the distributed tier if latency and sovereignty requirements soften. Public tenders written around the two-tier model would validate Blaize’s positioning while tenders written around centralized GPU capacity alone would suggest the distributed tier remains an aspiration.
What to Watch:
- Blaize’s Q2 print and second-half ramp should show the NeoTensr fulfillment beginning at $11 million or more and revisit the $130 million full-year guidance.
- Slippage on power, permitting, or financing will show which of the 1.7GW in announced targets is real.
- Whether Indonesia’s first tenders separate centralized AI factory capacity from distributed inference networks, the structural signal Futurum expects by 2028.
- Whether Blaize remains the default enterprise-edge compute in joint blueprints that Nokia extends to Vietnam and the Philippines within the next 12 months.
- Qualcomm Dragonwing, Hailo, and Axelera design wins in Southeast Asia will show whether the distributed tier consolidates around GPU-adjacent architectures before Blaize reaches scale.
See the complete press release on the Blaize website.
Sources
- Blaize and Datacomm Sign Technology Alliance MOU to Explore AI Inference Solutions Across Indonesia, Blaize, August 2026
- Sovereign AI: What Nations Want (And What They’ll Actually Get) – Report Summary, Futurumgroup
- Nokia, Blaize, and Datacomm Diangraha Unite to Deliver Hybrid AI Inference Across Indonesia and Southeast Asia, Prnewswire
- CoreWeave Expands Cloud AI Platform to Indonesia, Marking First Move Into Asia-Pacific Region, Coreweave
- Indosat becomes first mobile operator in SEA to roll out AI-RAN with Nokia and NVIDIA, Nokia
- Nvidia-backed Firmus plans 170,000-GPU Batam AI data centre, Techwireasia
Declaration of generative AI and AI-assisted technologies in the writing process: This content has been generated with the support of artificial intelligence technologies. Due to the fast pace of content creation and the continuous evolution of data and information, The Futurum Group and its analysts strive to ensure the accuracy and factual integrity of the information presented. However, the opinions and interpretations expressed in this content reflect those of the individual author/analyst. The Futurum Group makes no guarantees regarding the completeness, accuracy, or reliability of any information contained herein. Readers are encouraged to verify facts independently and consult relevant sources for further clarification.
Disclosure: Futurum is a research and advisory firm that engages or has engaged in research, analysis, and advisory services with many technology companies, including those mentioned in this article. The author does not hold any equity positions with any company mentioned in this article.
Analysis and opinions expressed herein are specific to the analyst individually and data and other information that might have been provided for validation, not those of Futurum as a whole.
Read the full Futurum Group Disclosure.
Other Insights From Futurum:
CoreWeave Q1 FY 2026: Capacity Constraints Amid Accelerating AI Demand
Can Appian and Synechron’s Open Underwriting Stack Transform Insurance Underwriting?
Is the Supply Chain AI Accountability Gap a Recipe for Failure?
Author Information
Brendan is Research Director, Semiconductors, Supply Chain, and Emerging Tech. He advises clients on strategic initiatives and leads the Futurum Semiconductors Practice. He is an experienced tech industry analyst who has guided tech leaders in identifying market opportunities spanning edge processors, generative AI applications, and hyperscale data centers.
Before joining Futurum, Brendan consulted with global AI leaders and served as a Senior Analyst in Emerging Technology Research at PitchBook. At PitchBook, he developed market intelligence tools for AI, highlighted by one of the industry’s most comprehensive AI semiconductor market landscapes encompassing both public and private companies. He has advised Fortune 100 tech giants, growth-stage innovators, global investors, and leading market research firms. Before PitchBook, he led research teams in tech investment banking and market research.
Brendan is based in Seattle, Washington. He has a Bachelor of Arts Degree from Amherst College.
