---
title: "Nscale — Form S-1 (SEC EDGAR, September 18 2026)"
url: https://stacklist.com/card/12dcccfe-a16c-4a02-9c49-954b75f26c53
source_url: "https://www.sec.gov/Archives/edgar/data/0002110365/000119312526395475/ck0002110365-20260918.htm"
stack: https://stacklist.com/c/finance/stack/0646d51e-dad0-4c1e-a033-869814d86f1c
summary: "Nscale Limited filed a Form S-1 registration statement with the SEC on September 18, 2026, seeking to register securities for a public offering. The England and Wales-incorporated company is classified as a smaller reporting company and emerging growth company."
tags: "sec-filing, form-s1, ipo, nscale, registration-statement, uk-company, emerging-growth"
key_entities: "Nscale Limited (organization), Securities and Exchange Commission (organization), Ian D. Schuman (person), Michael Benjamin (person), Jennifer Engelhardt (person), Adam J. Gelardi (person), Phoebee Gahan (person), Rod Miller (person), David Dixter (person), Jaime E. Ramirez (person), Latham & Watkins LLP (organization), Milbank LLP (organization), London (location), Houston, Texas (location), England and Wales (location), Form S-1 Registration (event)"
classification: "reference"
content_hash: "sha256:3434638c656b916eb57ad69252d4495c99b042f7bcb49cb1b276df13bf2d411a"
acp_version: "0.2"
token_counts_approximate: 18908
visibility: public
agent_accessible: true
status: "final"
---

# Nscale — Form S-1 (SEC EDGAR, September 18 2026)

Table of Contents As filed with the Securities and Exchange Commission on September 18, 2026. Registration No. 333‑     UNITED STATES SECURITIES AND EXCHANGE COMMISSION Washington, D.C. 20549     FORM S‑1 REGISTRATION STATEMENT UNDER THE SECURITIES ACT OF 1933     Nscale Limited* (Exact Name of Registrant as Specified in its Charter)   England and Wales (State or Other Jurisdiction of Incorporation or Organization) 7372 (Primary Standard Industrial Classification Code Number) Not Applicable (I.R.S. Employer Identification No.) Level 5, 16 New Burlington Place London W1S 2HX United Kingdom +44 (0) 208 740 7575 (Address, including zip code, and telephone number, including area code, of Registrant’s principal executive offices)     Nscale Operations US, LLC 109 N Post Oak Lane, Suite 140 Houston, Texas 77024 (832) 551-3300 (Name, address, including zip code, and telephone number, including area code, of agent for service)     Copies to:   Ian D. Schuman Michael Benjamin Jennifer Engelhardt Adam J. Gelardi Latham & Watkins LLP 1271 Avenue of the Americas New York, New York 10020 (212) 906‑1200   Phoebee Gahan Chief Legal Officer Nscale Limited Level 5, 16 New Burlington Place London W1S 2HX United Kingdom +44 (0) 208 740 7575   Rod Miller David Dixter Jaime E. Ramirez Milbank LLP 55 Hudson Yards New York, New York 10001 (212) 530‑5000     Approximate date of commencement of proposed sale to the public: As soon as practicable after the effective date of this registration statement. If any of the securities being registered on this Form are to be offered on a delayed or continuous basis pursuant to Rule 415 under the Securities Act of 1933, as amended, or the Securities Act, check the following box. ☐ If this Form is filed to register additional securities for an offering pursuant to Rule 462(b) under the Securities Act, check the following box and list the Securities Act registration statement number of the earlier effective registration statement for the same offering. ☐ If this Form is a post‑effective amendment filed pursuant to Rule 462(c) under the Securities Act, check the following box and list the Securities Act registration statement number of the earlier effective registration statement for the same offering. ☐ If this Form is a post‑effective amendment filed pursuant to Rule 462(d) under the Securities Act, check the following box and list the Securities Act registration statement number of the earlier effective registration statement for the same offering. ☐ Indicate by check mark whether the registrant is a large accelerated filer, an accelerated filer, a non-accelerated filer, a smaller reporting company, or an emerging growth company. See the definitions of “large accelerated filer,” “accelerated filer,” “smaller reporting company” and “emerging growth company” in Rule 12b-2 of the Securities Exchange Act of 1934, as amended, or the Exchange Act.   Large accelerated filer ☐ Accelerated filer ☐ Non-accelerated filer ☒ Smaller reporting company ☐     Emerging growth company ☒   If an emerging growth company that prepares its financial statements in accordance with U.S. GAAP, indicate by check mark if the registrant has elected not to use the extended transition period for complying with any new or revised financial accounting standards provided pursuant to Section 7(a)(2)(B) of the Securities Act. ☒ The registrant hereby amends this registration statement on such date or dates as may be necessary to delay its effective date until the registrant shall file a further amendment which specifically states that this registration statement shall thereafter become effective in accordance with Section 8(a) of the Securities Act, or until the registration statement shall become effective on such date as the U.S. Securities and Exchange Commission, acting pursuant to said Section 8(a), may determine . * On May 4, 2026 and May 5, 2026, we consummated a corporate reorganization and as a result Nscale Limited, a private company limited by shares under the laws of England and Wales, is the ultimate holding company of Arkon Energy Pty Ltd. and, directly and indirectly, of Nscale Global Holdings Limited. Prior to the consummation of this offering, Nscale Limited will re-register as a public limited company and change its legal name to Nscale plc.       Table of Contents The information in this prospectus is not complete and may be changed. We may not sell these securities until the registration statement filed with the Securities and Exchange Commission is effective. This prospectus is not an offer to sell these securities, and we are not soliciting an offer to buy these securities in any state where the offer or sale is not permitted.     PRELIMINARY PROSPECTUS SUBJECT TO COMPLETION, DATED , 2026 Shares Nscale Limited Ordinary Shares     This is the initial public offering of the ordinary shares of Nscale Limited. We are offering ordinary shares, par value $0.01 per share. Prior to this offering, there has been no public market for our ordinary shares. We currently expect the initial public offering price to be between $ and $ per ordinary share.     We intend to apply to list our ordinary shares on the New York Stock Exchange (“NYSE”) under the symbol “NSCL”. Investing in our ordinary shares involves risks. See “Risk Factors” beginning on page 21 .     We are an “emerging growth company,” as defined under the federal securities laws, and, as such, may elect to comply with certain reduced public company reporting requirements. See “ Prospectus Summary — Implications of Being an Emerging Growth Company .”         Per Share     Total   Initial public offering price   $       $     Underwriting discounts and commissions (1)   $       $     Proceeds, before expenses, to us (2)   $       $       (1) We refer you to “ Underwriting ” for additional information regarding underwriting compensation. (2) Assumes no exercise of the underwriters’ option to purchase additional ordinary shares. We have granted the underwriters an option for a period of 30 days after the date of this prospectus to purchase up to an additional ordinary shares from us, at the initial public offering price, less underwriting discounts and commissions.   Neither the SEC nor any other state securities commission has approved or disapproved of these securities or passed upon the adequacy or accuracy of this prospectus. Any representation to the contrary is a criminal offense. The underwriters expect to deliver the ordinary shares to purchasers against payment on or about , 2026.   *listed in alphabetical order   Lead Bookrunners Goldman Sachs & Co. LLC* J.P. Morgan* Morgan Stanley Bookrunners RBC Capital Markets BofA Securities Deutsche Bank Securities Credit Agricole CIB TD Securities Mizuho KeyBanc Capital Markets Cantor SMBC Nikko Wolfe | Nomura Alliance Co-Managers Citizens Capital Markets Loop Capital Markets Roth Capital Partners ABN AMRO Compass Point DNB Carnegie Rosenblatt SEB Tigress Financial Partners   Prospectus dated , 2026   Table of Contents       Table of Contents   TABLE OF C ONTENTS   About This Prospectus ii Market and Industry Data iii Trademarks, Service Marks and Trade Names iv Presentation of Financial and Other Information v Prospectus Summary 1 The Offering 16 Summary Consolidated Financial and Other Data 18 Risk Factors 21 Cautionary Statement Regarding Forward‑Looking Statements 72 Use of Proceeds 73 Dividend Policy 74 Corporate Reorganization 75 Capitalization 76 Dilution 79 Management’s Discussion and Analysis of Financial Condition and Results of Operations 81 Business 103 Management 121 Executive and Director Compensation 127 Principal Shareholders 141 Certain Relationships and Related Party Transactions 143 Description of Certain Indebtedness 149 Description of Share Capital and Articles of Association 153 Ordinary Shares Eligible for Future Sale 167 Material U.K. Tax Considerations 172 Material U.S. Federal Income Tax Considerations 175 Underwriting 180 Legal Matters 188 Experts 189 Enforcement of Civil Liabilities 190 Where You Can Find More Information 192 Index to Consolidated Financial Statements F- 1   Neither we nor the underwriters have authorized anyone to provide you with any information or to make any representations other than those contained in this prospectus, any amendment or supplement to this prospectus, or any free writing prospectus we have prepared, and neither we nor the underwriters take responsibility for, and can provide no assurance as to the reliability of, any other information others may give you. We and the underwriters are offering to sell ordinary shares and seeking offers to purchase ordinary shares only in the United States and certain other jurisdictions where offers and sales are permitted. The information contained in this prospectus is accurate only as of the date on the cover page of this prospectus, regardless of the time of delivery of this prospectus or the sale of ordinary shares. Our business, financial condition, results of operations and prospects may have changed since the date on the cover page of this prospectus. For investors outside the United States: neither we nor the underwriters have done anything that would permit this offering or possession or distribution of this prospectus in any jurisdiction, other than the United States, where action for that purpose is required. Persons outside the United States who come into possession of this prospectus must inform themselves about, and observe any restrictions relating to, the offering of the ordinary shares and the distribution of this prospectus outside the United States. i Table of Contents   Ab out This Prospectus Prior to this offering, we conducted our business through Nscale Global Holdings Limited. On May 4, 2026 and May 5, 2026, we consummated a corporate reorganization and as a result Nscale Limited, a private company limited by shares under the laws of England and Wales, is the ultimate holding company of Arkon Energy and, directly and indirectly, of Nscale Global Holdings Limited. Prior to the consummation of this offering, Nscale Limited will re-register as a public limited company and change its legal name to Nscale plc. The foregoing transactions are herein referred to collectively as the Reorganization. See “ Corporate Reorganization ” for a description of the Reorganization. This registration statement, including the prospectus contained herein, includes the audited consolidated financial statements, unaudited interim condensed consolidated financial statements, summary consolidated financial and other data of Nscale Global Holdings Limited, which holds all of our operating subsidiaries and, after the Reorganization and this offering, will be a direct wholly owned subsidiary of Nscale plc. The securities issued to investors in this offering will be ordinary shares of Nscale plc. Except where the context otherwise requires or where otherwise indicated, the terms “Nscale,” the “Company,” the “Group,” “we,” “us,” “our,” “our company” and “our business” refer to (i) Nscale Limited, a privately held company limited by shares, incorporated in England and Wales and its consolidated subsidiaries prior to the completion of the Reorganization and (ii) Nscale plc and its consolidated subsidiaries after the completion of the Reorganization. See “ Corporate Reorganization ” and “ Description of Share Capital and Articles of Association .” ii Table of Contents   Mark et and Industry Data Within this prospectus, we reference information and statistics regarding (i) our industry, (ii) the markets for our products, and (iii) the size and growth rate of the markets in which we participate. We are responsible for these statements included in this prospectus. We have obtained this information and statistics from our own internal estimates, surveys and research, as well as from various independent third‑party sources and publicly available data. Industry publications, research, surveys, studies and forecasts generally state that the information they contain has been obtained from sources believed to be reliable, but that the accuracy and completeness of such information is not guaranteed. These forecasts and forward‑looking information are subject to uncertainty and risk due to a variety of factors, including those described under “ Cautionary Statement Regarding Forward-Looking Statements ” and “ Risk Factors .” These and other factors could cause results to differ materially from those expressed in the forecasts or estimates from independent third parties and us. iii Table of Contents   Trad emarks, Service Marks and Trade Names We have proprietary rights to certain trademarks, service marks and trade names used in this prospectus that are important to our business, including but not limited to Nscale and the Nscale logo, certain of which are registered, or for which applications for registration are pending, in various jurisdictions. This prospectus contains additional trademarks, service marks and trade names of others, which are, to our knowledge, the property of their respective owners. We do not intend our use or display of other companies’ trademarks, service marks or trade names to imply a relationship with, or endorsement or sponsorship of us by, any other companies. Solely for convenience, the trademarks, service marks and trade names referred to in this prospectus appear without the ® and ™ symbols, but such references are not intended to indicate, in any way, that we will not assert, to the fullest extent under applicable law, our rights or the rights of the applicable licensors to these trademarks, service marks and trade names. iv Table of Contents   Pre sentation of Financial and Other Information On May 4, 2026 and May 5, 2026, we consummated a corporate reorganization and as a result Nscale Limited, a private company limited by shares under the laws of England and Wales, is the ultimate holding company of Arkon Energy and, directly and indirectly, of Nscale Global Holdings Limited. Prior to the consummation of this offering, Nscale Limited will re-register as a public limited company and change its legal name to Nscale plc. See “ Corporate Reorganization ” for a description of the Reorganization. Except as otherwise disclosed in this prospectus, the historical consolidated financial statements, the summary historical consolidated financial data, and the other financial information included elsewhere in this prospectus have been prepared in U.S. dollars in accordance with accounting principles generally accepted in the United States (“GAAP”). This historical financial information gives effect to the 60-for-1 forward stock split that occurred in May 2026 pursuant to the Reorganization but does not give effect to the Subsequent Reorganization Steps (as defined below) or this offering. We present our consolidated financial statements in U.S. dollars, which is our functional and presentational currency. All references in this prospectus to “dollar,” “USD” or “$” mean U.S. dollars, all references to “£,” “GBP” or “Pounds Sterling” mean British pounds sterling and all references to “Euro” or “€” mean the currency of the member states of the European Monetary Union that have adopted or that adopt the single currency in accordance with the treaty establishing the European Community, as amended by the Treaty on European Union. Certain monetary amounts, percentages, and other figures included in this prospectus have been subject to rounding adjustments. Accordingly, figures shown as totals in certain tables may not be the arithmetic aggregation of the figures that precede them, and figures expressed as percentages in the text may not total 100% or, as applicable, when aggregated may not be the arithmetic aggregation of the percentages that precede them. The acquisitions undertaken during the six months ended June 30, 2026, fiscal years ended December 31, 2025 and December 31, 2024, whether taken into consideration individually or as a group of related businesses, are not “significant” for purposes of Rule 3‑05 of Regulation S‑X. Therefore, we are not required to, and have elected not to, provide separate historical financial information in this prospectus relating to these acquisitions. Non‑GAAP Financial Measures Certain parts of this prospectus contain non‑GAAP financial measures, including Adjusted EBITDA, Adjusted EBITDA margin, Adjusted operating loss and Adjusted operating loss margin. These non‑GAAP financial measures are presented for supplemental informational purposes only and should not be considered a substitute for net loss, net loss margin, operating loss and operating loss margin or any other financial information presented in accordance with GAAP and may be different from similarly titled non‑GAAP measures used by other companies and may not be identical to corresponding measures in our various agreements. For additional information regarding our non-GAAP financial measures, and for a reconciliation of each such non-GAAP measure to its most directly comparable GAAP measure, see “ Management’s Discussion and Analysis of Financial Condition and Results of Operations—Non-GAAP Financial Measures. ” Glossary of Certain Terms The following are abbreviations, acronyms and definitions of certain terms used in this document: • “active capacity” means the IT megawatt capacity at each of our wholly-owned, colocation and leased facilities that are online and revenue generating. • “active data center sites” means all wholly-owned, colocation and leased sites that are ready-for-service. • “AI” means artificial intelligence. • “Aker” means Aker ASA and affiliated companies. • “Arkon Energy” means Arkon Energy Pty Ltd. • “API” means a set of rules and protocols that enable different software systems to communicate and share data, often to abstract complex functionality into simpler commands. • “APAC region” means Asia-Pacific region. • “colocation sites” means third-party sites where we rent capacity (including space, power, cooling, and network connectivity) and own and operate the GPUs and related compute infrastructure. v Table of Contents   • “contracted capacity” means the IT megawatt capacity at each of our wholly-owned, colocation and leased facilities that are under development pursuant to signed contracts with customers. • “contracted data center sites” means all wholly-owned, colocation and leased facilities that are under development pursuant to signed contracts with customers. • “CPU” means central processing unit. • “EMEA” means Europe, the Middle East, and Africa. • “FLOP” means floating point operations and is typically calculated per second as a common industry measure of a GPU’s computation power during training. • “GDPR” means the EU and U.K. general data protection regulations. • “GPU” means graphics processing unit. • “gross power” means total electrical power demand including all equipment and support infrastructure in a data center such as cooling and lighting. • “GW” means gigawatt. • “HMRC” means His Majesty’s Revenue and Customs. • “HPC” means high-performance computing. • “IT” means information technology. • “ITGCs” means information technology general controls. • “IT load power” means power consumed by AI or IT equipment. • “leased sites” means sites where we have entered into a long-term lease of land and power capacity to develop data center capacity (including space, power, cooling, and network connectivity). • “LLMs” means large language models. • “MOIC” means multiple on invested capital. • “MW” means megawatt. • “NKS” means Nscale Kubernetes Service. • “operating data center sites” means all wholly-owned, colocation and leased facilities that are either ready-for-service or under development pursuant to signed contracts with customers. • “potential power capacity” means the capacity expected to be available at our wholly-owned, colocation or leased facilities, following the development of behind the meter power generation and/or the approval of grid capacity applications. • “PUE” means power usage effectiveness. • “SAFE” means Simple Agreements for Future Equity. • “Sandton” means Sandton Capital Solutions Master Fund V, L.P. collectively with its related parties. • “SLA” means service level agreement. • “SOP” means standard operating procedures. • “TCV” or “total contract value” means the aggregate revenue contracted across the full committed term of signed customer agreements, measured at date the contract is signed and excluding any optional extensions, renewals, unexercised capacity, or any significant financing components relating to upfront payments. • “Tier 1,” “Tier 2,” and “Tier 3” refer to commonly accepted industry standard classifications for data center markets/locations, and indicate primary, secondary, and emerging markets/locations, respectively. • “Time-To-First Token” measures the elapsed time between a user submitting a request and an AI model generating its first output token. • “Token” is a unit of text (typically a word or subword) that AI models process during inference—i.e., when a model generates a response to a user request. • “Token volume” is a widely used AI industry measure of inference activity and reflects the scale of AI workloads processed on a platform. vi Table of Contents   Letter from Our Founder To Our Prospective Shareholders, I founded Nscale based on a simple and powerful belief: artificial intelligence is the fourth industrial revolution. The world is observing the largest technological shift in history. The response has been an explosion of building. Every sector is being reshaped. Drug discovery timelines are shrinking. Engineering cycles are becoming automated end-to-end. Scientific research is compressing from years to weeks. Entire categories of software are being rebuilt. Frontier labs, enterprises, and sovereign states are committing capital at an unprecedented pace and scale, all racing toward the same objective: to build intelligence and put this technology into the hands of the many. But every one of them runs into the same constraint. These extraordinary AI capabilities do not simply materialize. They are manufactured, requiring physical infrastructure of extraordinary scale and complexity. Ambition may be unlimited, but the capacity to deliver it is not. Infrastructure is the defining constraint of the age, and it will decide which nations lead, which companies win, and which products become ubiquitous. The constraint is not simply a shortage of chips. It is land, in the right place, with the right connectivity. It is power, secured, and delivered at gigawatt scale. It is the data center, built for densities and cooling loads that are increasing every year. It is the silicon, integrated and clustered in some of the largest single machines in the world, optimized for the highest performance. It is the cloud platform that orchestrates and runs the workload efficiently. And it is the AI services layer that turns raw capability into intelligence that a business, a hospital, or a government can use and deploy. Every layer is a bottleneck. If any layer is constrained, it impacts everything above it. If any layer is not optimized, it slows down everything above it. The entire vertical stack must be integrated from the ground up as a single unified system, delivered globally, at scale, with the highest performance. That is the foundation. Whoever builds the foundation upon which the AI market sits will build a generation-defining company. Nscale is building that foundation. The Founding Thesis When I set out on this journey in August 2023, I had just finished consolidating hundreds of megawatts of land and power in the lowest-cost markets in the world, with the intention of building a data center platform. The “ChatGPT moment” was nine months earlier, and by the time GPT-4 shipped in March 2023, data center demand was compounding quickly. As we engaged prospective customers for our data center platform, I ran a deep analysis on the sector. It became clear that the market was both fragmented and in its infancy, and not prepared to deliver on the wave of incoming demand. So I asked a simple question: “What does it take to build the infrastructure platform that delivers this technology at a global scale?” The answer was full vertical integration. Own the land, the power, the data center, the chips and the software. Delivered end-to-end in a single unified system, purpose-built from the ground up to deliver speed, performance, and efficiency on the most demanding AI workloads. Upon this realization, we went into stealth, hired world-class leaders in each function who have operated at scale, and we launched the platform as Nscale in May 2024. From inception, we built Nscale with an infrastructure-first thesis, building against contracted customer demand, underwriting projects to attractive long-term returns, maintaining prudent leverage and seeking to match the duration of our capital commitments with the strong revenues supporting them. That discipline remains central to how we scale. Since launch, we have experienced exponential growth and demand for our services. We have also compounded every layer of our vertical integration, bringing more capability in-house, adding margin, removing cost, and further extending our control of each layer of the AI value chain. All of this is translating to less risk and more value for our customers. In only two and a half years, we have scaled from $100 million in total contracted value to over $103 billion. Our power pipeline has grown from 750 MW to 10 GW+ owned and controlled. Our team has grown from 40 to over 1,000. We’ve expanded our customer base and now operate in 14 regions. Each milestone is built on the last, and we are just getting started. Nscale Infrastructure Nscale Infrastructure is the physical foundation of our platform – land, power, data centers, and large-scale AI clusters owned and controlled end-to-end that allow our hyperscale customers to train, deploy, and run inference at scale. vii Table of Contents   Today, Nscale Infrastructure represents the majority of our business. We serve some of the largest companies on Earth through strategic partnerships and long-term agreements. This is the foundation that funds the continued buildout of our global footprint, and we believe the opportunity ahead here is vast. With a 10 GW+ portfolio of owned and controlled power, we expect significant expansion in Nscale Infrastructure as we grow with our existing customers, and contract and scale new customers building and delivering frontier intelligence. Nscale Cloud Nscale Cloud extends our platform beyond large-scale AI infrastructure. Our vision is to build the engine of superintelligence, the backbone of the intelligence economy. Nscale Cloud expands the capabilities of Nscale Infrastructure with a full-stack cloud platform that allows AI natives and enterprises to generate value with AI. With Nscale Cloud, we abstract away the complexity of managing infrastructure, expanding into platform and AI services including fine-tuning and inference services. Nscale Cloud captures more of our customers' workloads, and drives high-value platform and service revenues on top of our global AI infrastructure platform. Through Nscale Cloud we will unlock high-margin full-stack AI cloud revenues while building a more diversified customer portfolio of frontier labs, developers, AI natives, and enterprises. We believe many of these customers will become some of the largest companies on Earth. We also see a significant and growing opportunity in sovereign AI, as nations increasingly seek to build and control their own AI infrastructure and capability rather than rely on others. Nscale is well positioned to serve this demand as a trusted, sovereign-capable platform. This is particularly true in Europe, where the need for sovereign AI infrastructure is especially acute, and where our British domicile positions us as a natural, trusted partner. Central to our offering is our announced acquisition of Anyscale, the company behind Ray, the leading open-source framework for distributed compute. This is the first step in a broader strategy to build out our platform layer. It reflects conviction that open-weight models, alongside open-source infrastructure, are foundational to how AI gets built and deployed. It also moves us further up the stack, closer to owning the workload itself, not just the infrastructure beneath it. We expect to continue innovating and expanding our platform capabilities as the market and our customers ’ needs evolve. Impact When I started this journey, the premise was simple: we couldn ’ t just build at scale; we had to build responsibly. Our commitment to the communities in which we live and work and our stewardship of the environment remain paramount. We maintain these founding commitments, whether it ’ s harnessing renewable energy, drawing on behind-the-meter energy generation to protect regional grids and consumers, capturing waste heat for local industries and agriculture, or investing deeply in providing services to the communities where we build. We make sure our growth strengthens communities across the globe. The Future AI will redefine how we live and work for generations to come, creating new opportunities for people, businesses and economies around the world. We are not simply building for the demand we see today; we are building the platform for what comes next. We move with speed. We approach opportunity with conviction. We approach risk with discipline. We have leaders in each function who have operated at global scale. We have a deep bench of directors who have operated platforms of trillion-dollar scale, with products used by billions of people. And while we have come a long way in a short time, measured against the scale of the opportunity before us, we are just getting started. Thank you for joining us on this journey.   Josh Payne Founder and CEO   viii Table of Contents   Pros pectus Summary This summary highlights information contained elsewhere in this prospectus. This summary does not contain all the information that may be important to you before deciding to invest in our ordinary shares, and we urge you to read this entire prospectus carefully, including the “Risk Factors , ” “Cautionary Statement Regarding Forward-Looking Statements,” “Business” and “Management’s Discussion and Analysis of Financial Condition and Results of Operations” sections and our consolidated audited financial statements, and unaudited interim condensed consolidated financial statements including the notes thereto, included in this prospectus, before deciding to invest in our ordinary shares. Overview Nscale is a full-stack AI hyperscaler, building the engine of superintelligence. The artificial intelligence market is catalyzing a fourth industrial revolution, as AI-driven productivity gains are integrated into virtually every product, industry, and job. This shift is driving the largest infrastructure build-out in history, as companies and nation-states race to build leading AI capabilities across every major vertical. Over the next five years, we believe the largest companies will be born from this next great exponential technology cycle. Success will be defined not only by algorithms or models, but by access to scalable, reliable, and cost-efficient AI infrastructure capable of supporting them. AI infrastructure is the foundation of this market. The company that builds and operates that foundation on which the AI market sits will become the hyperscaler of tomorrow. Nscale is building that foundation. We believe Nscale is positioned to be one of the lowest cost producers of AI compute, delivering the infrastructure on which the AI economy is being built. With our AI cloud platform designed as part of our vertically integrated model, we deliver AI infrastructure and services to some of the most important technology companies globally. We plan, source, build and operate multiple stages of the AI infrastructure value chain, including powered land and power access, behind-the-meter power generation, data center design and build, facility ownership and operations, GPU fleet deployment and management and our unified software control plane. This integrated approach maximizes our ability to service our customers reliably and economically. See “ Management's Discussion and Analysis of Financial Condition and Results of Operations—Maximizing the Value of Our Infrastructure .” Our platform is built on a global portfolio of low-cost powered land and behind-the-meter power sites that support the deployment of AI infrastructure to serve large-scale AI training and inference workloads. We develop large-scale AI campuses in low-cost power regions to enable cost-effective delivery of AI compute across our portfolio while smaller, distributed GPU clusters allow us to serve sovereign workloads or those that require in-country infrastructure deployment. This model allows Nscale to deliver low-cost, reliable, and performant compute to a global customer base running large scale training and inference workloads. Our unified cloud software platform delivers a consistent, secure orchestration layer across our global portfolio, managing scheduling, fleet operations, and system health across heterogeneous environments to provide a reliable, standardized operating model for running AI training and inferencing workloads at scale. At present, our degree of ownership and operational control varies by project and location: our current portfolio includes wholly-owned sites, colocation and leased deployments, where in some cases we do not control the underlying third-party facility. We also do not control all inputs to our platform, such as the manufacture of GPUs, power generation equipment, or other critical infrastructure components, which we instead procure through close partnerships with key suppliers. Given the long lead times associated with the development of large-scale data center campuses, as of August 31, 2026, our active capacity is primarily composed of colocation and leased deployments. These deployments enable us to provide compute capacity in the nearer term, establish and expand customer relationships, and demonstrate our ability to deploy, operate and manage GPU infrastructure and related cloud services at scale. However, the substantial majority of our contracted data center sites by contracted capacity are wholly owned by us and provide us with greater control over key aspects of the AI infrastructure stack, typifying the vertically integrated nature of our platform. Beginning in 2023, we proactively assembled a multi-gigawatt portfolio of powered land in structurally low-cost power markets, positioning the company ahead of accelerating demand for high-density, AI-optimized infrastructure. As demand for AI compute has accelerated, access to reliable, large-scale contiguous power has emerged as the primary gating factor for AI infrastructure deployment. In March 2026, we took a decisive step to secure long-term leadership in the U.S. AI infrastructure market by acquiring 100% of the share capital of American Intelligence & Power Corporation (“AIPCorp”), which includes the Monarch Compute Campus in Mason County, West Virginia, one of the largest AI-dedicated infrastructure sites globally, with a power generation capacity runway scalable to over 6.5 GW of IT load power and over 8 GW of gross power. Following this acquisition, we created a new subsidiary—Nscale Energy & Power—to internalize power origination and development capabilities and establish our Energy & Power division. This early, power-first strategy now underpins our long-term unit-economic advantage and provides the foundation on which we continue to scale our AI campuses and distributed deployments globally. 1 Table of Contents   Some of the world’s largest and most advanced technology companies choose Nscale because we deliver large-scale AI infrastructure at structurally lower cost, with exceptional delivery certainty enabled by our vertically integrated model. Our platform is engineered for stable runtime performance and high operational uptime, supported by continuous monitoring, automated alerting, and proactive remediation at the node, rack, and cluster levels. By unifying software, hardware, and data center design within a single platform, we provide customers with a consistent operating environment for AI workloads at global scale. We generate the vast majority of our revenue through long-term, multi-year take-or-pay contracts. These contracts have an industry-leading weighted average contract life of approximately 5.7 years, reflecting the value of our powered land portfolio and vertically integrated infrastructure. Contract terms commence upon successful delivery of GPU compute clusters of infrastructure. We aim to extend customer relationships beyond initial contract maturities through phased infrastructure refresh cycles and modular upgrades that allow customers to deploy successive generations of AI hardware and continue running inference workloads at scale without relocating data or re-architecting platforms. In addition, at the request of our customers, many of our contracts feature a right of first refusal that grants our customers the priority right to secure additional compute capacity, which speaks to the strength of our service and customer relationships. As of August 31, 2026, our infrastructure portfolio included approximately 25,000 active GPUs and 461,000 active and contracted GPUs, five active and twelve contracted data center sites (including seven wholly-owned sites, nine colocation sites, and one leased site) and approximately 1.37 GW of active and contracted capacity (representing 1 GW at owned sites, 200 MW at leased sites and 165 MW at colocation sites), with line of sight to approximately 10 GW of potential power capacity for development across sites under ownership or long-term control and power procurement agreements following the acquisition of the Monarch Compute Campus. Our footprint is concentrated in renewable-rich, low-cost power regions such as Norway, Portugal, Iceland, and select locations in the United States and APAC. Our global footprint, structurally advantageous cost base, and long-standing senior-level relationships across the power, infrastructure, and hardware supply chain provide a differentiated competitive position in addressing emerging demand for sovereign AI solutions and position us well to support continued enterprise adoption and expanding hyperscaler demand.   2 Table of Contents       In addition to our physical infrastructure, our proprietary full-stack software platform manages the deployment and operation of large-scale AI compute through a single, secure control plane that covers global campuses, edge, and sovereign deployments. The platform layer provides fleet management and observability, quota and identity controls, automated health checks with remediation, and unified policy enforcement, giving customers a single-pane view for scheduling, capacity and SLAs across heterogeneous hardware and geographies. The managed software and application services layers operationalize production-grade primitives, including bare metal, Slurm, NKS and virtual instances, while offering inference, fine-tuning, a curated model library and evaluation tooling. These capabilities shorten time-to-first-token (“TTFT”) and raise GPU FLOP utilization, lowering costs through predictive scheduling and autoscaling. They also enforce enterprise security, provide auditable registries and ensure data-residency controls, while supporting distributed inferencing and edge deployments. To further enhance our software platform, we entered into a definitive agreement to acquire Anyscale on July 28, 2026 (the “Anyscale Acquisition”) and the team behind Ray. Anyscale is an AI compute platform built on Ray, a leading open-source framework for distributed AI. Ray is experiencing exponential growth with 740 million cumulative downloads, including approximately 174 million downloads in the second quarter of 2026 alone. Anyscale brings in the orchestration layer, built around open-sourced Ray, that abstracts away distributed computing complexity and optimizes both training and inference workloads across AI infrastructure. Anyscale, which is already powering AI at AI-native companies and enterprises, expands our customer base and is used by both AI native and traditional enterprise customers to train and run open-sourced models on proprietary data. Approximately 200 employees focused on improving workload performance and infrastructure utilization will join us as part of the Anyscale Acquisition. We deliver enterprise-grade inference today through Nscale Cloud. To date, we have processed billions of tokens, with token usage growing rapidly. Our serverless inference supports a broad set of open models and serves as the foundation of an enterprise-ready inference platform built for reliability, performance, and scale. These production services are tightly integrated with our model and data registries and fine-tuning pipelines, enabling customers to move from prototyping to production with consistent performance and governance. The depth and breadth of our offering, clear advantages of vertical integration, and structural cost efficiencies have enabled us to attract customers and realize significant growth in our business. For the six months ended June 30, 2026 and 2025, we generated revenues of $140.6 million and $10.4 million, respectively, representing an increase of 1,252%. Revenue for the year ended December 31, 2025 increased 73%, from $19.1 million in 2024 to $33.0 million in 2025. As of August 31, 2026, we had approximately $2.6 billion of active and $103.4 billion of active and contracted TCV under long-term take-or-pay contracts with customers, compared to $0.5 billion of active and $38.0 billion of active and contracted TCV as of December 31, 2025. These contracts support the deployment of approximately 461,000 GPUs that were active or contracted as of that date. 3 Table of Contents   History of Nscale Nscale was formed on the thesis that the rapid advancement and adoption of AI throughout the economy would drive enormous demand for dedicated high-performance data center and compute infrastructure required to support compute intensive training and inference workloads at scale. We also believe that AI would drive digital infrastructure development beyond traditional Tier 1 data center markets into Tier 2 and Tier 3 locations, with larger tranches of low-cost power. With this vision, we acquired gigawatts of land and power assets in the lowest-cost power markets globally, establishing an incumbent position. In 2022 and 2023, before becoming independent from Arkon Energy, we proactively began assembling a portfolio of powered land in low-cost power markets, positioning us ahead of accelerating demand for high-density, AI-optimized infrastructure. In May 2024, Nscale was spun out from Arkon Energy. Since then, we have quickly developed relationships with leading AI customers, raised over $3.3 billion through our series financing, secured aggregate commitments of approximately $1.4 billion through our GPU Financing Facility (as defined herein), an aggregate commitment of $900.0 million with a letter of credit sublimit of $200.0 million under the Revolving Credit Facility (as defined herein), an aggregate initial-term rent of approximately $2.54 billion under the DFS Framework Agreements (as defined herein) as of September 4, 2026, an aggregated commitment of up to $790.0 million under the Kvandal South DC Facility (as defined herein), an aggregated commitment of up to $331.9 million under the Macquarie Iceland Facility (as defined herein), an aggregated commitment of up to $1.85 billion under the Ward County GPU Facility (as defined herein), and an aggregated commitment of up to $1.2 billion under the North Carolina GPU Facility (as defined herein) and expanded our data center and powered land footprint. Our early, power-first strategy now underpins our long-term unit economics advantage and provides the foundation on which we continue to scale our AI campuses and distributed deployments globally. To further enhance our software platform, we entered into a definitive agreement in relation to the Anyscale Acquisition in July 2026 with closing expected at the time of or concurrent with this offering. By pairing our global, low-cost physical infrastructure-spanning behind-the-meter power generation, modular liquid-cooled data centers, and high-performance GPU clusters with Anyscale’s enterprise-grade software layer, we are building a vertically integrated, full-stack AI hyperscaler.     Industry Background AI as a Strategic Imperative: Consumer, Commercial, Sovereign We believe that artificial intelligence represents the most significant technological shift since the advent of the internet, catalyzing a fourth industrial revolution that is rapidly transforming the global economy. AI adoption is robust, global, and growing as AI continues to flow into both everyday and mission-critical workflows. Improvements in model intelligence are enabling users and companies to work more efficiently, automate tasks, keep organized, and improve business results. These efficiency gains unlock new use cases, which in turn accelerate adoption and fuel additional demand for AI. As this flywheel strengthens, implementing AI is becoming a strategic imperative for enterprises that want to remain competitive. Control over compute, data, and AI models is also increasingly viewed as a matter of national security. Data privacy, critical infrastructure, cybersecurity, and AI-specific legislation has emerged across Europe, the Americas, and APAC. Compute needs tied to local legal frameworks and national boundaries are creating AI infrastructure demand at the sovereign level, most notably across Europe, North America, and Asia. 4 Table of Contents   Robust Demand for AI Infrastructure Every technological revolution requires a shift in the underlying infrastructure. There is a direct relationship between the compute resources available to train and run AI models and the quality and competitiveness of AI applications. Modern AI models, and the AI applications built on these models, require very large-scale, contiguous clusters of the latest generation of GPUs along with high bandwidth and low-latency networking, driving the rise of scaled, AI‑ready digital infrastructure. Legacy data centers and CPU-centric clouds were not designed for modern AI workloads and cannot economically support the step-change in rack density, power, delivery, and cooling that modern accelerated computing requires. The scale of these prior-generation facilities does not align with AI hyperscaler-level compute requirements, and prior-generation build practices do not meet current expectations for data center construction timelines and specifications. At the same time, hyperscalers’ capital expenditures have increased significantly over the prior few years and are expected to continue rising. Underpinning this is the fundamental mismatch in data center supply lagging demand, which is similarly expected to persist as the AI buildout continues. Customers are also demanding sustainable power sourcing at the same time that compute is trending towards commoditization, placing increasing pressure on unit economics. These factors necessitate an evolved approach to purpose-built, vertically integrated AI infrastructure. Power Cost and Procurement at Scale are the Critical Enablers of AI Access to power and its delivery cost have emerged as the primary gating factor for AI capacity expansion. As AI workloads scale, a differentiated power strategy focused on long-term availability, cost predictability, and delivery certainty has become increasingly essential to the intelligence ecosystem. Compute and data center providers are turning to securing alternative, longer-term, lower-cost, and lower-carbon power solutions, both utility-served and behind-the-meter. Grid build-outs or modernization to support large-scale AI deployments has extended project lead times and raised costs. This power-led sourcing approach is shifting data center development from Tier-1 markets, where major new deployments are increasingly power-constrained, toward Tier-2 and Tier-3 markets with more available, lower-cost power, and the ability to support multi-gigawatt deployments over time. This accelerates capacity build-outs in these regions and drives advantageous unit economics. As the AI infrastructure market matures and supply and demand imbalances subside, we believe winners in AI infrastructure will be defined by the ability to deliver high-performance compute at scale and speed while lowering effective costs per watt. AI Clouds Enabling the AI Revolution Purpose-built AI compute providers have emerged to meet the specialized requirements of modern AI workloads. Components such as power availability and procurement at scale, data center construction and maintenance, low-latency interconnects, GPU rack densities, support systems, and management software require rapidly evolving technical expertise. As compute consumption surges with new model releases, product launches, and use-case adoption, specialized third-party compute providers are increasingly vital to deliver compute capacity quickly, reliably, cost‑efficiently, and at scale. The impracticality associated with individual hyperscalers and other enterprises maintaining these functions drives the durable need for specialized AI compute providers. The Rise of Distributed AI and Open‑Source Compute Frameworks As AI models and datasets have grown, the computation required to train and serve them no longer fits on a single server and must be distributed across large clusters of GPUs and CPUs. Doing so efficiently and reliably has historically required specialized systems engineering that most AI teams lacked. Ray emerged to solve this: an open‑source framework that abstracts the complexity of distributed computing, letting developers scale AI workloads (training, post‑training, inference, and data processing) from a laptop to a large cluster without rewriting their applications. In part because it lowers this barrier, we believe the distributed‑compute layer for AI is increasingly standardizing on open‑source frameworks such as Ray. Enterprises running mission-critical workloads typically require open-source frameworks alongside managed orchestration, performance tuning, security and access controls, observability and reliability guarantees. Anyscale was founded by the original creators of Ray to provide that commercial layer. This dynamic and broad open-source adoption creating demand for a commercial control layer is central to how we serve AI‑native and enterprise customers. 5 Table of Contents   Solving the Limitations of Legacy Infrastructure The AI paradigm shift requires services and capabilities that legacy hardware and general‑purpose clouds were not designed to meet: • Prior-generation build practices are obsolete. AI deployments require rapid, flexible delivery of GPU and data center infrastructure at scale without sacrificing cost efficiency. • Legacy data centers are not suited for AI workloads . Modern AI clusters require advanced and future-oriented power densities, cooling technologies, and modularities. • Infrastructure stacks are fragmented and suboptimal for modern customers. Access to and management of AI infrastructure has to be seamless and catered to the various support and management needs of the customer. • GPUs are not being used to their fullest extent. Optimizing GPUs for modern AI workloads requires end-to-end orchestration that solves scheduling, performance, cluster health, and rapid failure recovery. • Existing software solutions are not tailored for AI. AI developers and enterprises demand flexible, unified tooling to build, test, fine‑tune, and deploy AI workloads at scale, spanning distributed scheduling and orchestration, autoscaling, observability, security, and governance, and increasingly built on open‑source standards. Together, these limitations underscore the need for a vertically integrated, low-cost, purpose-built cloud that can deliver large quantities of compute reliably, securely, and tailored for the modern hyperscaler, AI-native, enterprise, sovereign, and developer. Our Solution We deliver our full-stack, vertically integrated platform, spanning power to token, through two complementary products: Nscale Infrastructure and Nscale Cloud. Nscale Infrastructure is our global platform delivering hyperscale AI infrastructure services, combining behind-the-meter power generation, liquid cooled AI data centers, and high-performance compute under long-term, take-or-pay contracts with large-scale customers. Nscale Cloud is our high-performance, scalable, and secure AI cloud, delivering the full AI lifecycle under one contract, one identity layer, and one governance model, and enhanced by the Anyscale Acquisition, acquiring the team behind Ray. Together, these products optimize cost, performance, and efficiency across the full stack, from power to token. Our service offerings to customers comprises the following: Nscale Infrastructure : Our global platform delivers hyperscale AI infrastructure services under long-term, take-or-pay contracts with large-scale customers, and comprises the following: • Fleet Operations: Our Fleet Operations is the software stack that operates the fleet at scale, comprising Fleet Manager for automated workflows from day-zero provisioning through day-two remediation, Control Center for unified lifecycle automation, the Radar API for customer-facing break and fix controls, instance metadata, and topology, centralized observability for metrics, logs, and traces at global scale, and deployment tooling spanning low-level design generation, ERP integration, and host discovery. • Infrastructure Services: We deploy large-scale AI infrastructure clusters engineered specifically for high -performance AI computing workloads. These infrastructure clusters combine the latest generation computing infrastructure, including the NVIDIA Grace Blackwell “GB” 300 and Vera Rubin “VR” 200 GPUs, high-performance network fabrics, and optimized storage solutions to deliver optimal performance on AI workloads. These clusters are architected to maximize performance and support sustained, high-utilization workloads. Crucially, our infrastructure is designed to flexibly support both training and inference, allowing customers to shift their usage over time without requiring changes to the underlying hardware. • Purpose-Built AI Data Centers : We anchor our platform with a global footprint of advanced, sovereign, and sustainable data centers. Our facilities are designed from the ground up as high-performance, liquid-cooled AI data centers, purpose-built to avoid the bottlenecks of traditional infrastructure. Our use of a prefabricated, modular architecture reduces on-site complexity, accelerates deployment, and enables repeatable, industrialized scale. This design supports higher power density deployments and allows for phased infrastructure refreshes, enabling modular upgrades over time without decommissioning entire facilities. • Behind-the-Meter Power Micro-grids : We design, build, own, and operate behind-the-meter power infrastructure to deliver low-cost power at large scale for our large AI campus projects. These AI micro-grids combine on-site generation with intelligent power management and operate independently from the utility grid. 6 Table of Contents   Nscale Cloud : Our high-performance, scalable, and secure AI cloud delivers the full AI lifecycle to AI-native and enterprise customers under one contract, one identity layer, and one governance model, and comprises the following capabilities: • AI Services: Our AI Services accelerate the path from development to production, allowing customers to consume AI outputs directly rather than managing models or infrastructure themselves. Our AI Services offers integrated tooling, industry-standard application programming interfaces (“APIs”), serverless and dedicated inference, fine-tuning and evaluation as managed pipelines, a curated open-weight model library with day-zero support for new releases, a prompt workbench for prototyping, an AI gateway providing a single endpoint for routing, authentication, governance, and cost controls, and bring-your-own-model with fully managed inference on dedicated capacity for enterprise customers running proprietary weights. Our AI Services also provide distributed orchestration for scaling data processing, training, inference, and reinforcement learning workloads. • Platform Services: We abstract the operational complexity of managing large-scale GPU fleets through a single secure control plane, with security and sovereignty by design. This includes Nscale Kubernetes Service (“NKS”) for container orchestration with reservations and multi-tenancy, Managed Slurm for tightly-coupled multi-node training and batch workloads, virtual and bare-metal instances, enterprise identity and access management with sophisticated role-based access control, bring-your-own identity provider, and enterprise single sign-on, and Envir, our managed environment service that lets enterprises deploy the Nscale software stack in pre-configured, secure tenancies. The Anyscale Acquisition will enhance our ability to provide a managed platform for developing and running distributed AI workloads across large-scale GPU infrastructure. Built on Ray, Anyscale provisions computing resources, distributes workloads across GPUs, automatically scales capacity, recovers failed tasks and provides tools to monitor performance and resource usage. Our unified interface will provide customers with comprehensive visibility and control over resource utilization, system health, and workload performance, enabling them to focus on their core business while we manage the underlying infrastructure. How We Design, Deploy, and Operate We define success at the cluster level and align every design, tooling, and staffing decision to a single service-level standard across the design, deployment, and operation of each cluster. During design, we engineer each cluster as one machine built for multiple GPU generations, using digital-twin simulation to validate data center fit and power-constraint compliance and integrating our supply chain to reduce deployment delays. During deployment, our GPU fleet passes an extensive multi-stage burn-in and validation sequence: spanning single-node, multi-node, and full-cluster testing, designed to establish cluster health from day one. During operation, automated orchestration detects, isolates, migrates, and replaces failing nodes so that customer training and inference workloads continue with minimal interruption, which we believe improves effective GPU utilization and lowers the delivered cost per token. Anyscale and Ray On July 28, 2026, we entered into a definitive agreement in relation to the Anyscale Acquisition, acquiring the team behind Ray, which upon completion will add a developer platform and distributed compute runtime to Nscale Cloud, further enhancing our full-stack platform of managed services. Ray is a leading open-source framework for distributed AI, with 740 million cumulative downloads, including approximately 174 million in the second quarter of 2026 alone. Anyscale is the commercial platform built by Ray’s founders. Anyscale is a managed, enterprise-grade platform providing a managed control plane, a performance-tuned runtime, enterprise controls, and a production developer surface spanning development workspaces, jobs, and services. Anyscale is deployable across public clouds, on-premises environments, and our own infrastructure, allowing customers to run distributed AI workloads wherever their data and compute reside. Anyscale already powers AI at AI-native companies and enterprises. We expect the acquisition to expand our customer base and to add approximately 200 employees focused on improving workload performance and infrastructure utilization. Anyscale’s founders created Ray. They built both the open-source engine and its commercial control layer, which we believe gives us a differentiated ability to advance the standard and to commercialize it. Together, Nscale and Anyscale enable optimization across the full stack in a way we believe no single-layer provider can replicate. We own and operate every layer beneath the workload, spanning power generation, data centers, GPUs, networking, and platform services, and with Anyscale we gain the runtime and developer platform that direct how those resources are consumed. Because each layer is engineered for the one above it, we can tune scheduling, workload placement, and resource utilization against infrastructure we control end to end, improving effective GPU utilization and lowering the delivered cost per GPU-hour and per token. Providers that operate only the infrastructure layer cannot see or shape the workload; providers that operate only the software layer must optimize against infrastructure they neither own nor control. We do both. 7 Table of Contents   8 Table of Contents   Competitive Differentiation Our vertically integrated AI platform is designed to address the specific operational, economic, and execution challenges faced by customers deploying large-scale AI workloads. Our platform is engineered to create durable competitive advantages: power-first economics, modular scalability, full-stack control, security-by-design, and open-source-by-design. Our key competitive strengths include: Global Powered Land Portfolio: Securing land with access to a large quantum of low-cost power is a strategic imperative and competitive advantage of our business. We follow a rigorous and proven powered land portfolio acquisition strategy focused on reserving significant near-term low-cost power, with high renewable power concentration and identifiable pathways for near to medium-term expansion potential, such as available adjacent land, expandable power capacity, likely approval of required regulatory permissions, and favorable supporting infrastructure. Our approach includes sites that can support campus-scale development with onsite powered micro-grid solutions, reducing reliance on constrained utility grids. In March 2026, we further strengthened this differentiation by acquiring 100% of the share capital of AIPCorp, which includes the Monarch Compute Campus in Mason County, West Virginia. The Monarch Compute Campus expands our U.S. footprint with an approximately 2,250-acre site which can deliver up to 8GW of gross behind-the-meter power. This site is positioned for large-scale AI campus deployments. We expect to develop up to 2 GW of gross power generation capacity by the first half of 2028, with the potential for multi-gigawatt expansion scalable to over 6.5 GW of IT load power and over 8 GW of gross power. Our differentiated global platform provides opportunities for long-term expansion capacity across our hyperscale campuses and distributed deployments, supporting our hub-and-spoke model by delivering lowest-cost compute for non-ultra latency sensitive workloads at hub campuses that serve primarily as hyperscale AI hubs, while maintaining the flexibility to serve latency and jurisdiction-sensitive workloads via our in-country spoke or distributed deployment facilities. Low-Cost, Renewable Power: Our infrastructure strategy prioritizes site selection in regions with abundant renewable power and structurally low-carbon power generation, including hydroelectric and other renewable sources. The economics and scalability of AI infrastructure are increasingly influenced by access to reliable, cost-effective, and environmentally sustainable power sources. By operating AI-ready data centers powered by renewable power and designing facilities optimized for high efficiency, we seek to reduce long-term operating costs and carbon intensity while supporting customers’ sustainability and energy transition objectives. Energy and Power Generation : We secure and operate power infrastructure purpose-built to support large-scale AI deployments to ensure long-term availability, cost predictability, and delivery certainty. Our power strategy emphasizes direct access to scalable energy sources, and where appropriate, dedicated or on-site generation to improve time-to-power and reduce reliance on congested utility grids. Our in-house energy & power team coordinates power and compute deployment in parallel, enhances execution certainty, and supports multi-gigawatt scalability. By internalizing power origination and development capabilities, we improve long-term unit economics for our AI campuses. AI Data Centers : As of August 31, 2026, our platform operated across five active and twelve contracted data center sites (including seven wholly-owned sites, nine colocation sites, and one leased site), with approximately 1.37 GW of active and contracted capacity. Our modular infrastructure design and globally coordinated supply chain enable us to scale GPU deployments efficiently across campuses, manage long-lead-time components proactively, and maintain consistency in performance, quality, and commissioning across deployments. This modularity, coupled with off-site construction and factory testing, significantly reduces on-site delays and transforms on-site work primarily into rapid assembly and validation. This design facilitates phased expansion without disrupting live workloads, allowing for future infrastructure swaps one module at a time. AI Infrastructure : Our infrastructure is designed to seamlessly support both training and inference at scale, allowing customers to switch workloads without infrastructure or software changes. Our end-to-end orchestration across the value chain further enables tighter operational control and more efficient use of each megawatt. Our extensive infrastructure and platform, supported by diverse partnerships with strategic original equipment manufacturers (“OEMs”), enables us to deliver high-performance compute at scale for our customers. AI Cloud Platform : We own the stack below and Anyscale optimizes the stack above. Anyscale’s optimized AI runtime and developer platform is synergistic with our full infrastructure stack, with every layer engineered for the one above it. This allows us to offer better performance and developer experience, through the likes of uptime and model efficiency, debugging, and proprietary fast node start-up, and further opens up the non-hyperscaler enterprise market for compute. Superior Unit Economics: Our vertically integrated model delivers structurally advantaged unit economics and a lower total cost of ownership per GPU-hour. This is achieved through our power-first strategy, which secures abundant, low-cost, and reliable power —often at electricity prices approximately 70% lower than major U.S. power markets . Our purpose-built AI data centers further optimize efficiency, targeting industry-leading PUEs and maximizing GPUs per megawatt. Coupled with our proprietary full-stack software platform that improves GPU utilization and tools that will be acquired as part of the Anyscale Acquisition, reinforcing our Nscale Cloud 9 Table of Contents   opportunity, we provide an end-to-end AI platform with structurally lower dollars-per-watt cost of compute. This comprehensive optimization across power, infrastructure, and software drives attractive margin generation on a per-megawatt basis. Integrated and Modular Supply Chain Approach: Our platform is built on an integrated, global view of the AI infrastructure supply chain, enabled by a modular design philosophy. By standardizing factory‑built modules across mechanical, electrical, plumbing, and GPU halls, we can source, manufacture, test, and deploy key components in a coordinated, repeatable, and geographically interchangeable manner. This modularity allows components to move seamlessly across regions—with minimal customization—optimizing global supply, reducing bottlenecks, and improving our ability to rebalance deployment schedules across Europe and North America. This approach materially enhances procurement efficiency, reduces execution risk, and enables faster response to customer demand. We maintain centralized oversight of supplier relationships while executing deployments locally, allowing us to combine the scale benefits of a global platform with the speed and flexibility required for regional delivery. Open-by-Design Architecture : We differentiate through an open‑first architecture built on open‑source standards rather than proprietary lock‑in. With Anyscale, we will operate with both Ray and its commercial control layer, which we believe gives us a differentiated ability to advance the open‑source standard and to monetize it. An open‑by‑design approach enables customers to adopt our platform through freely available software and scale into paid, production‑grade services, enables enterprise and sovereign customers to own their models and data rather than depend on closed, frontier‑lab ecosystems, and, because Ray runs across public clouds, on‑prem, and our own capacity, we expect it to broaden our reach while lowering switching costs. Combined with our vertically integrated, power‑to‑tokens model, we believe this positioning enables us to compete for developers and enterprises that more closed platforms may not effectively serve. Security-by-Design Across the Full Stack: We architect security at every layer of the stack and deliver on a wide range of customer needs or preferences while maintaining consistent infrastructure, operational, and supply chain standards across regions. Sovereignty: As a UK-domiciled company with a broad geographic footprint, Nscale has a competitive advantage in providing sovereign compute to customers across multiple continents, in particular Europe. In the context of AI compute, sovereignty means enabling governments, enterprises and AI natives to run critical AI workloads under their own operational, legal, and security frameworks, with clear control over where infrastructure is located, who operates it, who can access it, and the applicable governing law. We believe there will be growth opportunities from sovereign AI in the coming years, from governments, enterprises and AI natives. Growth Strategies We are focused on the following key strategies to drive long-term growth and value creation: Expand Access to Powered Land: We are focused on accelerating and expanding our access to powered land, and have secured, and will continue to explore securing, differentiated sites with abundant power to deploy next-generation AI infrastructure at scale as we expand our deployments to meet customer demand. Our proprietary site selection, vertical integration, and power visibility procurement strategies help ensure rapid time-to-market and durable cost advantages. By internalizing power origination and development expertise, we can accelerate time-to-market for new AI campuses, improve cost visibility and reliability over the long-term, and better align power availability with customer demand. Capture New Workloads with Existing Customers: The partnerships formed via our multi-year contracts with customers coupled with the diversity and tailored nature of our offering positions us to grow alongside our customers. By continually supporting their existing and evolving compute needs, we aim to expand our commercial engagements through extensions, add-ons, additional sites, and referrals. Further, embedded growth arises from recurring capex refresh cycles as customers upgrade to the latest-generation GPUs to support cutting-edge workloads. Broaden Customer Base and Enter New Markets: We are extending our reach into new industries, geographies, and verticals, serving regulated and sovereign customers worldwide, and replicating our proven hub-and-spoke model in Asia and other high-growth regions. AI compute demand is broadening from hyperscaler-led large language models (“ LLM”) training to enterprise inference, fine-tuning, and deployment, as well as physical AI and other use cases. While continuing to capitalize on our offerings that are optimized for AI training at scale, we are simultaneously positioning to serve evolving needs through enterprise-ready capacity, managed orchestration, and distributed clusters aligned to regulatory and performance constraints. We are in the process of support