top of page
Business & Enterprise
AI Technology
Cloud Infrastructure
Nscale
Built a vertically integrated AI infrastructure company by converting low-cost European power and data-center assets into contracted GPU capacity, cloud software, and sovereign AI deployments, despite the execution risk of financing capacity ahead of deployment.
1
MODEL
BUSINESS MODEL
IaaS
- Originated within Arkon Energy, initially built around low-cost power assets and crypto-mining infrastructure, before Josh Payne repositioned it as an AI infrastructure company.
- Launched from stealth in May 2024 and raised a $155 million Series A in December 2024 to build AI-ready data centers.
- Developed vertically integrated capacity: powered land, data centers, GPU clusters, high-performance networking, and the Nscale Cloud Platform.
- Secured long-term take-or-pay contracts with Microsoft and Anthropic, using committed demand to support GPU procurement and infrastructure build-out.
- Expanded beyond raw infrastructure by agreeing in July 2026 to acquire Anyscale, adding Ray-based distributed-compute software and a developer ecosystem to its infrastructure platform.
1. Secure an input advantage—power, land, cooling, fiber, or data-center capacity—before competing for compute demand.
2. Convert that input into a deployable infrastructure product with GPU provisioning, networking, orchestration, and operational support.
3. Win anchor customers whose multi-year commitments can finance capacity, but validate delivery timelines before scaling obligations.
4. Add a software layer only when it reduces customer workload complexity or increases retention; infrastructure without software remains exposed to commodity GPU pricing.
5. Use contracted revenue to finance expansion only when deployment milestones, power access, and supply chains are credible.
6. Avoid treating signed contract value as realized revenue; the common mis-build is building a backlog that outruns the ability to deliver power, GPUs, and operational capacity.
DISTRIBUTION MODEL
Nscale scales through long-term take-or-pay contracts, owned or controlled power and data-center capacity, GPU deployment, and cloud software. It reported approximately $50.2 billion in committed total contract value as of April 30, 2026, $103.4 billion in active and contracted value as of August 31, 2026, and $140.6 million revenue in H1 2026 versus $33.0 million for all of 2025. However, only about 25,000 GPUs were reportedly active against roughly 461,000 active or contracted GPUs, making deployment speed, power availability, financing, and customer concentration the decisive scalability constraints.
HOW THEY BUILT IT
HOW TO ARCHITECT IT
Enterprise Sales, Direct Sales, Partnerships, Product Integrations
HOW THEY OPERATIONALIZED
HOW TO REPLICATE WHAT WORKED
dm
- Sold long-duration dedicated AI capacity directly to hyperscalers and frontier AI labs, especially Microsoft and Anthropic.
- Used sovereign AI programs and government relationships in Norway, the UK, and Portugal to establish strategic infrastructure credibility.
- Partnered with NVIDIA, Dell, Nokia, Vast, Verne, Aker, Start Campus, and Caterpillar to secure hardware, networking, power, and construction capability.
- Reached developers and AI-native companies through the Nscale Cloud Platform, on-demand compute, serverless inference, and Anyscale’s Ray-based distributed-compute software following the acquisition agreement.
What worked: anchor a small number of massive, creditworthy contracts to finance vertically integrated capacity, then use sovereign and partner ecosystems to establish delivery credibility.
Trap if copied blindly: anchor-contract distribution becomes dangerous when two customers dominate contract value and revenue depends on future construction and GPU deployment; without diversified demand and delivery capacity, the model becomes a financed construction project rather than a scalable cloud business.
| PATTERNS OF THIS MODEL
PATTERNS IN VERTICALLY INTEGRATED AI INFRASTRUCTURE:
1. POWER IS THE UPSTREAM CONSTRAINT. Access to low-cost electricity, grid interconnection, land, and cooling increasingly determines which AI infrastructure providers can deliver capacity.
2. TAKE-OR-PAY CONTRACTS FINANCE THE BUILD-OUT. Long-duration commitments convert future demand into collateral for GPUs, construction, and debt, but they also create delivery obligations before revenue recognition.
3. SOFTWARE REDUCES COMMODITY EXPOSURE. Infrastructure providers that add orchestration, developer platforms, or workload software can retain more pricing power than raw GPU suppliers.
THE WEAKNESS: vertically integrated AI infrastructure can become a high-leverage construction business; contract value may grow much faster than active capacity, revenue, and operating cash flow.
What companies with this model reveal
| OPPORTUNITY INTELLIGENCE
GOLDMINE 1 — POWER-TO-MODEL PLATFORMS. Most AI clouds sell GPU-hours, while buyers need a complete path from power and land to trained or served models. Bundling energy, data centers, GPUs, orchestration, and model tooling can capture value across the stack.
GOLDMINE 2 — SOVEREIGN AI FACTORIES. Governments need local compute, data residency, energy strategy, and national model capability. Nscale’s Norway, UK, and Portugal projects show that sovereign demand can anchor large infrastructure investments.
GOLDMINE 3 — INFRASTRUCTURE-TO-WORKFLOW BUNDLES. The Anyscale acquisition points to a larger opportunity: sell not only GPU capacity but distributed-compute software, workload orchestration, and developer adoption.
THE PIT — BACKLOG-LED VALUATION. A large contract backlog can overstate near-term business strength when only a fraction of GPUs are active and revenue recognition depends on future construction.
THE SECOND PIT — ANCHOR-CUSTOMER DEPENDENCE. Microsoft and Anthropic reportedly represented about 85% of contract value, exposing Nscale to renegotiation, delay, or strategic-change risk.
MOVE WITH CAUTION — Do not finance GPU and data-center expansion based on future take-or-pay commitments without conservative assumptions on construction timelines, power delivery, hardware availability, and customer credit risk.
Untapped Business Model / Gaps / Goldmines / Pits
Patterns & Insights
2
MARKET
mkt mt es
Nscale competes against AWS, Microsoft Azure, Google Cloud, Oracle, CoreWeave, Nebius, Lambda, and Crusoe. Rather than matching hyperscalers across general cloud services, it attacked the AI capacity bottleneck through vertically integrated power, data centers, GPU clusters, and sovereign deployments. Its attack surface was Europe’s energy advantage and the hyperscalers’ need for supplemental AI capacity. Microsoft and Anthropic became anchor customers because they required dedicated capacity beyond their internal infrastructure, while NVIDIA, Dell, Nokia, and Aker strengthened Nscale’s supply and delivery ecosystem.
MARKET TYPE
Emerging Market
ENTRY STRATEGY
Brownfield Entry
WHY THEY WON
AI infrastructure is an Emerging Market characterized by explosive demand, GPU and power scarcity, evolving contract structures, and uncertain long-term utilization. Nscale exploited an opening created by the shift from general cloud computing to AI factories requiring dedicated power, dense GPU clusters, and sovereign capacity. Its advantage was upstream control of energy and land assets combined with downstream cloud software. The transferable lesson is that emerging infrastructure markets reward companies that secure scarce inputs early and convert them into contractable capacity before demand normalizes or supply becomes abundant.
EXECUTION
Nscale entered through a corporate spin-out and asset repurposing rather than a conventional startup launch. It began as an offshoot of Arkon Energy, which had acquired low-cost power assets in Europe and North America and operated crypto-mining infrastructure. Josh Payne initially planned a data-center business, then repositioned the company around AI infrastructure as GPU demand surged. Nscale launched from stealth in May 2024 and raised $155 million in Series A funding in December 2024. The critical bet was that power assets, data-center capability, and GPU operations could be converted into AI-ready infrastructure. This was viable because the company already controlled energy-related assets and understood high-density compute operations. The approach required rapid enterprise sales, financing, construction, and cloud-software capability; it would fail if legacy assets could not be converted into reliable, contracted AI capacity.
FOOTHOLD STRATEGY
Lighthouse Customer Strategy
Nscale’s early foothold came from European power and data-center assets, then shifted to strategic AI customers requiring dedicated capacity. Microsoft became the decisive lighthouse customer through projects in Norway, the UK, Portugal, and the U.S., including a five-year agreement valued at about $6.2 billion for capacity in Narvik and an October 2025 agreement covering approximately 200,000 NVIDIA GB300 GPUs across Europe and the U.S. Anthropic later became another anchor through a GPU-infrastructure agreement at the Monarch Compute Campus worth up to $44.6 billion. These relationships validated Nscale’s ability to structure and deliver hyperscale AI capacity.
GROWTH CAMPAIGN
CAMPAIGNS THAT WORKED
SO WHAT: Nscale’s most powerful campaigns combine sovereign infrastructure, anchor-customer commitments, and ecosystem partnerships to make a capital-intensive build-out credible.
KEY LEARNING: in AI infrastructure, buyers and capital providers evaluate delivery capability—power, land, construction, and operational readiness—not merely contract size.
HOW TO USE: publish a verifiable capacity ladder—contracted capacity, under-construction capacity, active capacity, and revenue-recognized capacity—so buyers can distinguish future commitments from operating reality.
KEY LEARNING
THE STANDARD: In an Emerging Market, entrants win by securing scarce inputs and defining the operational standard before supply becomes abundant.
RULE 1 — UPSTREAM CONTROL CREATES DOWNSTREAM PRICING POWER. Power, land, and grid access can matter more than GPU-hour price.
RULE 2 — CONTRACTS FINANCE CAPACITY BUT CREATE OBLIGATIONS. Take-or-pay agreements provide visibility while requiring timely delivery and capital discipline.
RULE 3 — SUCCESSFUL ENTRANTS INTEGRATE THE STACK. They win by combining energy, data centers, GPUs, networking, software, and operations.
RULE 4 — EMERGING MARKETS REWARD SPEED BUT PUNISH OVEREXTENSION. Rapid backlog growth without deployment capacity creates financing and execution risk.
MARKET TYPE: Emerging Market (category).
gc
Market Context
| MARKET INTELLIGENCE
PATTERNS IN ENTERPRISE-SALES LED, PARTNERSHIP-AMPLIFIED INFRASTRUCTURE DISTRIBUTION:
1. SOVEREIGN REFERENCES SUBSTITUTE FOR CATEGORY EDUCATION. Government-backed projects and national AI programs make a new infrastructure provider credible to enterprises and capital markets.
2. PARTNERS AMPLIFY DELIVERY CAPABILITY AND DEPENDENCY. NVIDIA, Dell, Nokia, and construction partners improve execution capacity but tie the vendor to external supply chains.
3. CAPACITY METRICS BECOME COMMERCIAL PROOF. Active GPUs, deployed megawatts, and revenue per megawatt matter more to sophisticated buyers than headline contract value.
4. THE WEAKNESS: enterprise-led distribution can concentrate revenue among a few customers with strong negotiating power and changing AI strategies.
| MARKET ENTRY PLAYBOOK
THE STANDARD: Brownfield entry is appropriate when an operator can repurpose existing assets, capabilities, or infrastructure into a higher-value market without starting from zero.
RULE 1 — Enter through existing assets only when they solve a scarce input problem, such as power, land, logistics, or operational expertise.
RULE 2 — Validate that the new market values the same underlying capability; Nscale’s power and data-center assets transferred into AI infrastructure because AI factories require both.
RULE 3 — Rebuild the commercial layer for the new buyer; crypto-mining economics did not automatically translate into hyperscaler contracts or cloud software.
RULE 4 — Reject brownfield entry if legacy assets create cost, regulatory, or technical constraints that outweigh the head start.
How to enter
| FOOTHOLD STRATEGY PLAYBOOK
THE STANDARD: A lighthouse-customer strategy uses one demanding customer to prove the product, unlock financing, and create reference gravity.
RULE 1 — Choose a customer whose requirements force enterprise-grade delivery, not merely a pilot.
RULE 2 — Convert the relationship into contractual and operational proof before pursuing broad sales.
RULE 3 — Avoid dependence on one lighthouse; use its credibility to diversify into adjacent segments and geographies.
RULE 4 — In capital-intensive infrastructure, a lighthouse should also improve financing access, procurement scale, or operational learning. Nscale’s Microsoft relationship did all three, but its concentration with Microsoft and Anthropic remains a structural risk.
How to get the first strong position
MARKET PATTERNS & PLAYBOOK
3
MONEY
money rev pri
Nscale monetizes through long-term take-or-pay contracts for reserved AI compute, managed infrastructure services, and usage-based cloud access. Its Microsoft and Anthropic agreements create multi-year committed revenue, while Nscale Cloud offers on-demand compute and serverless inference billed by consumption. As of April 30, 2026, it reported approximately $50.2 billion in committed TCV, but H1 2026 revenue was only $140.6 million, showing that predictability depends on converting contracts into deployed, revenue-generating capacity.
REVENUE MODEL
Contract Revenue, Usage-Based
PRICING MODEL
Custom Enterprise Pricing, Usage-Based Pricing, Volume-Based Pricing
WHY THEY WON
Customers pay for dedicated GPU nodes, full clusters, networking, storage, managed infrastructure, and cloud software under long-term take-or-pay contracts or consumption-based pricing. Revenue expands as data centers and GPU tranches enter service, as customers add capacity, and as serverless inference usage grows. Committed contracts provide revenue visibility, but revenue recognition follows deployment and service delivery. Margins depend on power cost, construction and financing expense, GPU depreciation, utilization, and the spread between contracted capacity pricing and operating cost.
Nscale prices reserved private-cloud clusters through long-term take-or-pay contracts negotiated by capacity, duration, GPU type, and location. Its Nscale Cloud offers on-demand compute and serverless inference charged by consumption or tokens. The strategic consequence is a two-tier model: hyperscaler and frontier-lab contracts finance large infrastructure, while cloud consumption serves AI-native companies and sovereign buyers.
TARGET AUDIENCE
Hyperscalers such as Microsoft (enterprise committed capacity; want supplemental AI infrastructure across regions).
Frontier AI labs such as Anthropic (enterprise committed capacity; want dedicated GPU infrastructure for training and inference).
Governments and sovereign-AI programs (enterprise/custom; want local data residency, energy-aligned capacity, and national AI infrastructure).
AI-native companies and developers (on-demand or serverless; want GPU compute, distributed-compute tooling, and inference access).
CUSTOMER BUYING BEHAVIOUR
Large customers begin with a multi-year AI capacity forecast, evaluate power, location, GPU availability, networking, reliability, and delivery timelines, then negotiate long-term take-or-pay contracts. Procurement involves technical validation, legal negotiation, financing review, and phased deployment milestones. Developers and smaller AI companies more often test on-demand compute or serverless inference before expanding into reserved capacity.
| PRICING INTELLIGENCE
What makes this model effective & make customers pay
Customers are not buying GPUs; they are buying guaranteed future access to AI capacity located where power, sovereignty, and performance requirements can be met.
RULE 1 — SELL CAPACITY CERTAINTY, NOT JUST COMPUTE. Take-or-pay contracts price the buyer’s need to reserve future AI capacity against scarcity and launch risk.
RULE 2 — MATCH CONTRACT DURATION TO ASSET LIFE. Multi-year commitments finance data centers and GPUs, but must reflect realistic technology and demand cycles.
RULE 3 — PRICE POWER AND CONSTRUCTION INTO CAPACITY. Energy cost, grid access, construction, debt service, and utilization risk must be embedded in contract pricing.
RULE 4 — SEPARATE CONTRACT VALUE FROM RECOGNIZED REVENUE. A large backlog is valuable only if the company can deliver capacity on schedule and maintain utilization.
The buyer is really buying assured AI production capacity, proven by contracted dollar commitments, deployed megawatts, active GPUs, and revenue recognized after delivery.
PRICE & REVENUE
| Revenue Risk - The biggest threat to revenue stability
Deployment timing risk: Nscale had roughly 25,000 active GPUs against approximately 461,000 active or contracted GPUs, so much of its contracted value depends on future data-center and GPU deployment.
Customer concentration: Microsoft and Anthropic together accounted for about 85% of contract value, creating exposure to renegotiation, delay, or strategic change by either customer.
Financing and construction risk: Nscale’s model uses long-term contracts to support borrowing for GPUs and infrastructure; construction delays, rising power costs, or financing constraints can pressure margins and liquidity.
GPU obsolescence: Rapid NVIDIA architecture transitions can reduce the economic life of installed hardware before depreciation and debt are serviced.
Sovereign and geopolitical risk: Projects in Norway, Portugal, the UK, and the U.S. depend on permitting, grid access, government support, and cross-border infrastructure execution.
Nscale reported $19.1 million revenue in 2024, $33.0 million in 2025, and $140.6 million in H1 2026, alongside a roughly $1.02 billion H1 2026 net loss; it raised $2 billion in Series C funding in March 2026 at a reported $14.6 billion valuation and filed for a reported $30 billion NYSE IPO in September 2026. These are company-reported or media-reported figures, not assured future revenue.
Where the model can break
4
MOTION
motion ge cs
Nscale sells high-value, capital-intensive, reliability-sensitive AI infrastructure through structured enterprise processes. Its Microsoft and Anthropic agreements involve multi-year take-or-pay commitments, phased deployment, power and data-center development, and executive-level negotiation. The company’s $50.2 billion committed TCV and $103.4 billion active and contracted value show that enterprise contracting, not self-serve adoption, is the primary revenue motion.
GROWTH EXPANSION MODEL
Vertical Integration, Acquisition-Led Growth
HOW THEY EXPAND
Nscale expanded vertically by controlling more of the AI infrastructure stack: power assets, data centers, GPU clusters, networking, cloud software, and distributed-compute tooling. Its July 2026 agreement to acquire Anyscale added Ray-based distributed-compute software, developer ecosystem reach, and workload orchestration above raw infrastructure. It also expanded geographically through Norway, Portugal, the UK, Texas, and other U.S. sites. The strategic logic was to move from a power-and-data-center supplier to a full-stack AI infrastructure platform, but each move increased capital intensity, integration complexity, and execution risk.
COMPETITIVE STRATEGY
Differentiation
HOW THEY COMPETE
Nscale faced AWS, Microsoft Azure, Google Cloud, CoreWeave, Nebius, Lambda, and Crusoe. Its battleground was hyperscale AI capacity, sovereign infrastructure, and power-constrained AI factories. Its comparative advantage was upstream control of power and land assets, combined with GPU deployment and cloud software. It used vertical integration to offer dedicated capacity from energy to model workloads, rather than competing only as a GPU reseller. The constraint was that hyperscalers possessed greater capital, distribution, and operational scale, while other neoclouds competed aggressively for GPU supply and anchor contracts.
GROWTH ENGINE
GTM
Content Flywheel
Nscale’s growth engine begins with a large customer’s need for dedicated AI capacity. Enterprise sales and infrastructure teams translate that need into a long-term take-or-pay contract covering GPU type, location, power, networking, delivery tranches, and pricing. Nscale uses contracted demand to finance land, data centers, GPUs, and networking, then recognizes revenue as capacity enters service. Proven deployment supports additional contracts, lender financing, supplier procurement, and geographic expansion. The engine compounds when contracted revenue funds capacity that attracts further contracted demand. It breaks if construction delays, GPU supply constraints, customer concentration, or financing costs prevent contracted capacity from becoming active, revenue-generating infrastructure.
Nscale runs a sales-led, enterprise-first GTM system. It targets hyperscalers, frontier AI labs, governments, and large enterprises with dedicated AI capacity requirements, using direct sales, technical solution engineering, sovereign partnerships, and executive relationships to structure multi-year take-or-pay contracts. The Nscale Cloud Platform and Anyscale software layer support developers and AI-native companies, but they are secondary to the contracted-capacity engine. Partnerships with NVIDIA, Dell, Nokia, Vast, Verne, Aker, Start Campus, and Caterpillar strengthen supply, construction, and delivery capability. The founder-level implication is that this GTM requires financing capacity, construction execution, and operational reliability; it is unsuitable for a company that cannot deliver large-scale infrastructure on contractual timelines.
SUSTAINING MOATS
High Customer Lock-In
moat
The moat is Nscale’s control of scarce AI infrastructure inputs—power, land, data centers, GPU capacity, and financing—combined with long-term customer contracts. Committed take-or-pay agreements support debt financing, which funds more capacity; operational delivery then supports further contracts. The moat weakens if financing costs rise, construction delays occur, GPU assets depreciate faster than expected, or customer concentration increases.
| MOAT INTELLIGENCE
THE STANDARD: A capital moat lasts when access to financing, scarce inputs, and operational delivery creates a capacity position competitors cannot quickly replicate.
RULE 1 — POWER AND LAND ARE SCARCE INPUTS. Nscale’s origins in low-cost power assets and data-center development give it an upstream position many GPU resellers lack.
RULE 2 — COMMITTED REVENUE SUPPORTS CAPITAL ACCESS. Long-term Microsoft and Anthropic agreements provide lenders and suppliers with visibility into future demand.
RULE 3 — SOFTWARE DEEPENS THE PLATFORM. The Anyscale acquisition adds Ray-based distributed-compute capability, reducing Nscale’s dependence on raw GPU capacity alone.
THE SIGNAL: Nscale reported $50.2 billion committed TCV as of April 2026, $103.4 billion active and contracted value as of August 2026, and $140.6 million H1 2026 revenue. The moat does not protect against delayed construction, customer concentration, GPU obsolescence, or a financing-market shock; capital advantage can become capital risk when leverage is high.
Why this company remains defensible
ARR & TAKEAWAY
ARR Journey - what to do at each stage
PRE-$1M ARR — SECURE THE UPSTREAM ASSET. Nscale originated within Arkon Energy around low-cost power assets and crypto-mining infrastructure before repositioning toward AI.
$1–5M ARR — REPURPOSE POWER INTO AI CAPACITY. The company began converting data-center and energy assets into AI-ready infrastructure, launching from stealth in May 2024.
$5–10M ARR — BUILD THE AI INFRASTRUCTURE PRODUCT. Nscale developed GPU clusters, networking, cloud software, and sovereign AI capability.
$10–50M ARR — WIN SOVEREIGN AND HYPERSCALE DEMAND. It built Stargate Norway, Stargate UK, Start Campus Portugal, and U.S. projects while deepening Microsoft, NVIDIA, Dell, Nokia, and Aker relationships.
$50–100M ARR — CONVERT ANCHOR DEMAND INTO CONTRACTED CAPACITY. Microsoft and Anthropic agreements created multi-year take-or-pay commitments and enabled large-scale GPU procurement.
$100M+ ARR — FINANCE THE CAPACITY FLYWHEEL. Nscale reported $140.6 million H1 2026 revenue, $50.2 billion committed TCV, and $103.4 billion active and contracted value, while agreeing to acquire Anyscale and pursuing an IPO. WATCH: active capacity, construction timelines, and customer concentration now matter more than headline contract value.
Rule: In vertically integrated AI infrastructure, scale only when contracted demand, power access, construction capacity, and financing capacity advance together.
COPY PLAYBOOK : What Worked → What Failed → What to Replicate → What to Avoid
THE STANDARD: Secure a scarce input advantage, convert it into contracted AI capacity, and add software only after the infrastructure can be delivered reliably.
SEQUENCE:
1. Secure a defensible input advantage—power, land, data centers, logistics, or specialized operations—before committing to major capacity.
2. Win a demanding anchor customer and convert its requirement into a contracted, phased capacity deployment.
3. Use committed revenue to finance expansion only when construction timelines, GPU supply, and utilization assumptions are credible.
WORKED: Nscale used European power assets and data-center capability to build AI infrastructure, then used Microsoft and Anthropic commitments to finance rapid capacity expansion.
CAUTION: Do not copy the contract-financed capacity flywheel without diversified customers and conservative delivery assumptions; concentration plus leverage can turn growth into fragility. Also avoid treating contract backlog as revenue—capacity must be built, powered, and utilized before it becomes durable revenue.
bottom of page