The Global AI Infrastructure Opportunity
The AI economy is live. We build the physical GPU infrastructure to power it, turning institutional capital into high-density compute and recurring infrastructure revenue.
Why AI Infrastructure Matters
As organisations deploy AI across software development, healthcare, manufacturing, financial services and scientific research, demand for computing infrastructure continues to expand. Every new AI application increases the need for processing capacity, storage, networking and energy, reinforcing the importance of the physical infrastructure supporting the AI economy.
Outlays are no longer projections. They are hard commitments.
Recent analyses from S&P Global Ratings, Aviva Investors and the UN's Independent International Scientific Panel on AI indicate the combined 2026 capital expenditure guidance for the five major hyperscalers is closer to $750–800 billion, largely directed toward AI infrastructure and data centre expansion.
Microsoft
Continuing multi-billion-dollar investment in AI infrastructure, with cloud expansion and GPU deployments forming a significant share of capital expenditure.
Amazon
Projected to lead hyperscaler capital expenditure in 2026, driven by continued investment in AI infrastructure and AWS capacity expansion.
Meta
Accelerating investment in large-scale AI infrastructure to support next-generation model training and inference across its platforms.
Oracle
Expanding data centre infrastructure and AI cloud capacity through significant increases in capital expenditure and long-term enterprise compute commitments.
NVIDIA
Continued record data centre revenue underscores sustained global demand for AI computing infrastructure and accelerated processing platforms.
Market Deficit
Demand is compounding. Supply is constrained.
Demand grows exponentially, but physical bottlenecks choke supply. NexusCore Compute solves this by financing and owning the bare-metal layer, securing enterprise GPU infrastructure.
To understand why this infrastructure deficit is so severe, we map the core physical bottlenecks:
Fabrication Lead Times
Urgent market demand has completely outpaced fabrication; global pre-orders for core H100/H200 accelerators routinely triple total available component supply.
Upstream Duopolies
Global hardware deployment remains throttled by a severe component bottleneck, with SK Hynix and Samsung controlling 95% of high-bandwidth memory production.
Packaging Constraints
TSMC's CoWoS packaging capacity—vital for AI accelerators—is expanding at a modest ~1.6x annually, barely keeping pace with immediate enterprise-level orders.
Facility Timelines
Constructing high-density, liquid-cooled physical spaces requires multi-year horizons, while enterprise computational demand aggressively scales quarter-over-quarter.
Grid Capacity
Grid access and the availability of localized clean energy matrices have become the primary constraints dictating where new AI infrastructure can physically be deployed.
Structural Imbalance
This deficit is structural, not cyclical. Compute demand continues to outpace supply, meaning the massive gap between what the market needs and what it possesses continues to widen.
Silicon is no longer a tool. It is an infrastructure asset class.
The era of viewing the GPU as an internal component for visual workstations or gaming setups is over. Today, enterprise GPU infrastructure is the primary processing unit powering the most critical, capital-intensive macro infrastructure on earth: the AI data center.
Independent industry forecasts project the global data centre GPU market to reach approximately $400 billion by 2032, while investment across the broader AI infrastructure ecosystem continues to expand rapidly.
This massive shift has fundamentally reorganized how institutional wealth views technology allocation:
The Investment Vector: Institutional investors are increasingly shifting their focus from speculative AI applications to the physical infrastructure enabling AI at scale. Independent industry research indicates that nearly 80% of investors interested in GPU infrastructure identify generative AI as the primary driver of demand, reflecting a broader move toward long-life, asset-backed digital infrastructure.
Alternative Asset Migration: Industry surveys indicate growing institutional interest in physical AI infrastructure as investors seek exposure to long-term digital infrastructure assets.
Software fades. Topologies endure.
In every historical industrial leap, long-term capital finds its ultimate protection not in consumer applications, but in the physical layer supporting them. Individual AI models will encounter rapid obsolescence, software platforms will consolidate, and commercial applications will fluctuate. However, the raw physical compute infrastructure underpinning the entire ecosystem will persist, retain strategic relevance, and compound in baseline value.
Over 70% of all global corporate AI capital is deployed directly into the physical layer. This is a reflection of a hard reality: machine intelligence is a relentless, hardware-heavy discipline. Every single model trained, every inference query answered, and every enterprise layer deployed demands physical GPU servers running inside physical data center racks.
For institutional investors seeking direct exposure to the AI macro-trend without absorbing the binary risk of picking a single winning software application, infrastructure represents the most defensible, asset-backed, and structurally supported position available in the global market.
Unit Economics
Illustrative deployment economics
Representative 50,000 GPU AI Infrastructure Campus — illustrative unit economics across capital structure, margins, and hold period.
Infrastructure Investment
$2.8 Billion
Equity Contribution
65% ($1.82B)
Asset-Backed Financing
35% ($980M)
Target Utilisation
85%
Infrastructure Gross Margin
63%
Expected Payback Period
4 Years
Investment Hold Period
8 Years
Fund Overview
Target performance & current raise
Headline return targets for the vehicle, alongside capital committed against the current institutional raise.
Target Financial Metrics
Target Net IRR
19–22%
Target Annual Cash Yield
10–12%
Investment Horizon
7–8 Years
Target Infrastructure Utilisation
85%+
Maximum Loan-to-Value
45%
Fund Size
$2.9B Committed
of $7.5B Target
Fund Structure
Preferred Equity
Min. Commitment
$25 Million
Target Final Close
Q2 2027
Roadmap
Capital deployment roadmap
Phased build-out of GPU capacity and capital deployment through 2028.
Phase I · 2026
Capacity
150 MW
Estimated GPUs
50,000
Capital Deployed
$3.2 Billion
Phase II · 2027
Capacity
450 MW
Estimated GPUs
150,000
Capital Deployed
$8.7 Billion
Phase III · 2028
Capacity
1.0 GW
Estimated GPUs
350,000
Capital Deployed
$18.5 Billion
2028 Outlook
Where the capital goes
Planned allocation of raised capital, and the operating targets NexusCore Compute aims to hit by 2028.
Capital Allocation
GPU Procurement
42% ($3.15B)
Data Centre Construction & Expansion
31% ($2.33B)
Renewable Energy Infrastructure
12% ($900M)
Strategic Acquisitions
9% ($675M)
Working Capital & Operations
6% ($450M)
2028 Portfolio Targets
Over 350,000 enterprise GPUs under management
1 Gigawatt (GW) of AI compute capacity
More than 250 enterprise customers
Operations across 12 European data centre campuses
100% renewable-powered infrastructure
Annual recurring revenue exceeding $4.5 Billion
EBITDA margin of approximately 58%
Assets Under Management (AUM) exceeding $20 Billion
