NASDAQ: XLAB

The Acquisition Merger

Following the Domestication Merger, the Acquisition Merger was effected. At the closing of the Acquisition Merger:

●
Each issued and outstanding Simple Agreement for Future Equity (each, a “SAFE”), by and between Legacy Exascale and the holder thereof (each, a “SAFEholder”), was canceled and converted into the right to receive a number of shares of our Class A Common Stock determined under the terms of the applicable SAFE;

●
A base camp agreement between Legacy Exascale and an investor (the “Base Camp Investment Agreement”) was cancelled and converted into the right to receive a number of shares of our Class A Common Stock determined in accordance with the terms of the Base Camp Investment Agreement;

●
Each outstanding Legacy Exascale equity incentive award was cancelled and converted into the right to receive a number of shares of our Class A Common Stock determined based on Legacy Exascale’s fully diluted capitalization at the time of the Business Combination;

1

●
Each issued and outstanding Legacy Exascale Class A common stock was cancelled and converted into the right to receive a number of shares of our Class A Common Stock determined based on Legacy Exascale’s fully diluted capitalization at the time of the Business Combination;

●
Each issued and outstanding Legacy Exascale Class B common stock was cancelled and converted into the right to receive a number of shares of our Class B common stock, par value $0.0001 per share (“Class B Common Stock,” and together with the Class A Common Stock, the “Common Stock”) determined based on Legacy Exascale’s fully diluted capitalization at the time of the Business Combination; and

●
Each share in Merger Sub issued and outstanding immediately prior to the effective time of the Acquisition Merger, automatically became an issued share of Legacy Exascale (with such shares becoming the only issued shares of Legacy Exascale immediately after the effective time of the Acquisition Merger).

No fractional shares of our Common Stock were issued in connection with the Business Combination.

In connection with the Extraordinary General Meeting and the Business Combination, holders of 26,865,211 BCAR Class A Ordinary Shares exercised their right to redeem their shares for cash.

On the Closing Date, we issued, or reserved for issuance, a total aggregate of 33,689,050 shares of Class A Common Stock and 30,645,739 shares of Class B Common Stock, of which an aggregate of 19,354,261 shares of Class A Common Stock and 30,645,739 shares of Class B Common Stock were issued to the former Legacy Exascale securityholders in exchange for their equity interests in Legacy Exascale, representing an aggregate merger consideration of $500,000,000 based on a deemed value of $10.00 per share of our Common Stock. In addition, we assumed the BCAR Warrants, which became our warrants, with the result that, as of the Closing Date, we had 14,099,992 warrants issued and outstanding, each whole warrant entitling the holder thereof to purchase one share of our Class A Common Stock at an exercise price of $11.50 per share.

Listing

Prior to the Closing Date, BCAR’s units (the “BCAR Units”), the BCAR Class A Ordinary Shares and the BCAR Warrants were listed on the Nasdaq Stock Market LLC (“Nasdaq”) under the symbols “BCARU,” “BCAR” and “BCARW,” respectively. In connection with the Business Combination, all of the BCAR Units separated into their component parts and ceased trading on Nasdaq.

On August 28, 2026, our Class A Common Stock and warrants began trading on Nasdaq under the symbols “XLAB” and XLABW,” respectively. Our Class B Common Stock are not listed on Nasdaq or any other securities exchange and are not publicly traded.

Our Business

We are a next-generation AI infrastructure provider operating an asset-light, software-defined GPU compute platform and related AI infrastructure solutions. Our core business includes GaaS, through which we provide reserved and on-demand access to high-performance GPU compute capacity sourced from third-party data centers globally, as well as GPU cluster management and optimization services for AI data center (“AIDC”) operators. In addition, we have developed certain modular data center, high-density liquid cooling, high-voltage direct current (“HVDC”) power, data center interconnectivity and energy storage solutions that are designed to address deployment bottlenecks in AI infrastructure and that we believe are ready for commercial engagement, although these capabilities have not yet generated revenue as of the date of this Annual Report. The platform is purpose-built for large-scale AI workloads, including LLM training, fine-tuning, and high-concurrency inference.

2

Our business consists of two primary product and service categories. First, we provide GPU-based compute services through our GaaS offering, which delivers scalable access to high-performance GPU capacity via bare-metal and VM configurations. These services are offered through both on-demand and reserved usage models and are designed to support a range of AI workloads, including large-scale model training, fine-tuning, and high-concurrency inference. Second, we provide Infrastructure Solutions for AI deployments, which include (i) GPU cluster management and operational services provided to AIDC operators, including planning and configuration support, monitoring, performance tuning, and ongoing operational assistance for large-scale GPU deployments, which are revenue-generating and delivered pursuant to commercial service arrangements, and (ii) certain modular data center, advanced liquid cooling, HVDC power, data center interconnectivity and energy storage solutions that our management believes are ready to support customer deployments as of the date of this Annual Report, although such offerings have not generated revenue to date. We expect to pursue these offerings on an asset-light basis, primarily through partnerships, systems integration, contract manufacturing and other collaborative structures.

Industry and Market Background

The AI industry is undergoing a generational paradigm shift, driven by the rapid adoption of Generative AI and LLMs. This shift has created an unprecedented demand for specialized, high-performance accelerated computing infrastructure that far exceeds the capabilities of traditional general-purpose cloud architectures. We believe the market is currently in the early stages of a secular transition from legacy central processing unit (“CPU”)-centric data centers to accelerated computing environments purpose-built for AI.

The proliferation of foundational models and AI-native applications has triggered a massive capital investment cycle. According to a September 2025 report by Gartner, Inc., a business and technology insights company, global spending on AI infrastructure is projected to grow to exceed $2.0 trillion by 2026. This growth is driven not only by the training of increasingly larger models which now regularly exceed trillions of parameters, but also by the exponential rise in inference workloads as enterprises integrate AI into production environments. The demand for compute capacity is outstripping supply by a significant margin. As of early 2026, despite increases in manufacturing capacity, the demand for cutting-edge GPUs, such as NVIDIA’s Blackwell architecture and subsequent generations, remained robust. Market indicators suggest that supply constraints for high-end AI processors could extend through 2027 and into 2028.

Limitations of Legacy Cloud Infrastructure

Traditional hyperscale cloud providers have historically architected their infrastructure to primarily support general-purpose web applications, such as web-hosting, e-commerce, databases, and search, and have historically relied on CPU-based, web-scale computing architectures. While hyperscale cloud providers have added GPU offerings and continue to invest in AI-related infrastructure, the operational and architectural assumptions that underlie general-purpose cloud platforms, such as multi-tenant abstractions designed for a wide set of workloads, may not be optimized for certain AI workloads that require dense GPU clusters, high-performance interconnects, and operational practices focused on maximizing effective utilization and minimizing job disruption. Industry participants have stated that certain large, diversified cloud providers are not purpose-built for the AI and accelerated compute use cases served by specialized AI infrastructure providers.

As AI adoption accelerates, the market increasingly demands infrastructure that is purpose-built to address the unique characteristics of AI workloads. This shift is driving specific requirements for platforms and service providers that combine GPU capacity with specialized software, operational tooling, and infrastructure management practices intended to (i) optimize performance, (ii) maintain stability and uptime, and (iii) reduce the complexity of operating high-performance AI infrastructure.

3

In particular, AI adoption has driven demand for infrastructure that can address several interrelated requirements, including the following:

●
Performance at scale through balanced system design. Large AI training workloads can require high-throughput data pipelines and coordinated operation across many GPUs. In these settings, performance is often influenced not only by the GPUs themselves but also by the design and operation of supporting infrastructure components (including networking, storage, and systems software).

●
Maximizing effective utilization of expensive GPU resources. Because GPU compute capacity is a significant input cost for many AI workloads, effective utilization can meaningfully affect the economics and throughput of AI development and deployment. Industry users place strong emphasis on the degree to which real-world performance approaches hardware potential, and on how that performance can be affected by software stack efficiency, data movement and bottlenecks, as well as operational factors that interrupt or degrade workload execution.

●
Reliability, stability, and operational consistency. Large-scale training runs and production inference environments can be sensitive to interruptions, failures, and performance variability. As clusters scale, operational stability and uptime become increasingly important for avoiding disruptions and managing overall compute costs and time-to-completion. We have observed an increasing demand for lifecycle management, monitoring, validation, and proactive health-checking capabilities intended to prevent failures and rapidly remediate issues in complex AI infrastructure environments.

●
Reducing operational complexity for customers and improving usability. Deploying and operating GPU clusters at scale often involves significant complexity, including provisioning, configuration management, observability, incident response workflows, and ongoing tuning of infrastructure and software environments. The industry requires monitoring and observability solutions, as well as operational services designed to support the deployment, ongoing operation, and remediation of infrastructure components throughout their full lifecycle. These capabilities are needed to shift a meaningful portion of the infrastructure management burden from customers to the platform.

●
Speed of deployment and access to current-generation GPU capability. AI demand has increased the importance of time-to-capacity, including the ability to deploy and operate GPU clusters in a timely manner and, in some cases, to adopt new GPU generations as they become commercially available. Industry participants have cited speed to market and the scale of GPU clusters as factors relevant to competitive positioning in accelerated computing markets.

The Evolution of Purpose-built AI Clouds (NeoClouds)

In response to the demand for accelerated compute capacity and the constraints associated with obtaining and deploying advanced GPU resources, a category of purpose-built AI infrastructure providers, often referred to in industry discussions as “neoclouds,” has emerged.

Neocloud providers generally offer GPU-centric infrastructure and services designed specifically for AI workloads. By combining compute capacity with managed configuration, monitoring, incident response workflows, and workload tuning practices, these platforms are intended to improve effective utilization and deliver more predictable performance for model training and large-scale inference, while supporting service stability and uptime. Neocloud offerings are also commonly structured to reduce the operational complexity associated with deploying and operating high-performance GPU clusters and to provide customers with more rapid access to scalable GPU capacity as demand changes.

Business models among neocloud providers vary. Some specialized providers have adopted capital-intensive approaches that involve significant investments in GPU fleets and data center capacity, and certain market participants have described their operations as capital-intensive and related capital market risks.

4

Capital Investment and Industry Trajectory

The AI infrastructure sector continues to attract unprecedented levels of capital investment, driven by what we believe is a secular transition in global computing architecture. However, this rapid expansion is inherently capital-intensive, often requiring substantial upfront expenditures that can translate into significant balance sheet leverage, while simultaneously exposing operators to risks associated with hardware obsolescence, accelerated depreciation, and rapid technology refresh cycles. A November 2025 report by CreditSights projects combined capital expenditures for the top five hyperscalers increasing from approximately $256.0 billion in 2024 to approximately $602.0 billion in 2026.

We believe that this investment cycle is in its early stages. According to an April 2025 article by McKinsey & Company, global data centers will require a cumulative investment of approximately $6.7 trillion by 2030, of which approximately $5.2 trillion is specifically attributed to AI-related infrastructure. This forecast implies a sustained, multi-year expansion in the addressable market for data center delivery, specialized compute services, and hardware optimization.

We believe these capital inflows underscore the strategic importance of computing power as a fundamental resource for future economic growth. At the same time, the scale and structure of these investments highlight the importance of capital-efficient models that can mitigate leverage, manage asset lifecycle risk, and optimize returns amid ongoing hardware evolution. The magnitude of the projected investment suggests durable market demand for infrastructure providers capable of delivering high-performance resources with speed and capital efficiency.

Physical Constraints: Data Center, Power, and Deployment

The scaling of AI infrastructure is increasingly constrained by physical limitations related to power availability, thermal management, and data center construction timelines. As the thermal design power (“TDP”) of next-generation AI accelerators approaches and, in some cases, exceeds 1,000 watts per GPU, legacy data centers designed for lower-density workloads, typically supporting approximately 10 to 15 kilowatts per rack, are becoming insufficient for modern AI deployments. As a result, the industry is undergoing a structural transition toward high-density computing environments capable of supporting rack densities ranging from approximately 40 kilowatts to over 100 kilowatts per rack. We believe this transition requires the adoption of advanced infrastructure technologies, including next-generation liquid cooling solutions and modular data center (“MDC”) architectures, to overcome the thermal and power-efficiency limitations of traditional air-cooled facilities.

In addition to thermal constraints, limitations on utility power availability and transmission capacity are increasingly influencing the design and deployment of AI infrastructure. These constraints have driven growing interest in HVDC power architectures, which are designed to improve power transmission efficiency and reduce energy losses within high-density AI computing environments. Collectively, these physical constraints are becoming a critical factor in determining the pace at which AI infrastructure can be deployed and scaled and are increasingly viewed as a prerequisite to sustaining continued performance improvements in next-generation AI models.

As GPU TDP continues to rise, software optimization has emerged as a critical economic lever for AI infrastructure providers and their customers. Given the high capital cost and ongoing scarcity of advanced AI hardware, the ability to improve effective compute throughput and increase GPU utilization rates through software-defined efficiency is becoming increasingly important to the economic viability of AI workloads. At the same time, the industry is mandating a shift toward liquid cooling technologies to support next-generation rack densities that exceed 100 kilowatts, which we believe requires AI infrastructure to be architected from the ground up to operate reliably and efficiently in high-density environments.

We believe that the convergence of supply constraints, increasing technical complexity, and the need for rapid deployment has created a significant and durable market opportunity for asset-light, execution-focused AI infrastructure providers, like us, that can deliver high-performance compute capacity while addressing these evolving physical and operational challenges.

5

Our Solution

We have developed an asset-light, software-defined AI infrastructure platform designed to provide customers with access to high-performance GPU compute and related infrastructure services for large-scale AI workloads. The platform supports reserved and on-demand compute services and is intended to enable customers and infrastructure operators to provision, manage, monitor, and optimize GPU environments used for large-scale model training, fine-tuning, and high-concurrency inference.

Our business is organized around two primary product and service categories:

(i)
GaaS: We provide GPU-based compute services through our GaaS offering, which delivers scalable access to high-performance GPU capacity via bare-metal and VM configurations. These services are offered through both on-demand and reserved usage models and are designed to support a range of AI workloads, including large-scale model training, fine-tuning, and high-concurrency inference.

(ii)
Infrastructure Solutions. We provide Infrastructure Solutions for AI deployments, which include (a) GPU cluster management and operational services provided to AIDC operators, including planning and configuration support, monitoring, performance tuning, and ongoing operational assistance for large-scale GPU deployments, which are revenue-generating and delivered pursuant to commercial service arrangements, and (b) certain modular data center, advanced liquid cooling, HVDC power, data center interconnectivity and energy storage solutions that management believes are ready to support customer deployments as of the date of this Annual Report, although such offerings have not generated revenue to date.

Supporting these offerings, we utilize infrastructure-level interfaces, APIs, and operational tooling to facilitate provisioning, resource management, monitoring, incident response, performance management, and service delivery across our compute and infrastructure services. We expect to pursue our broader infrastructure offerings on an asset-light basis, primarily through partnerships, systems integration, contract manufacturing, and other collaborative structures.

We believe that our combination of GPU compute services, operational capabilities, and infrastructure solutions is designed to address the performance, reliability, deployment, and operational requirements of modern AI infrastructure and to position us for long-term growth in the accelerated computing market.

Competitive Strengths

The following subsections describe certain competitive strengths that we believe differentiate us in the rapidly evolving AI infrastructure market and may may support our ability to compete as the market continues to evolve. The following competitive strengths should be balanced with, and considered in the context of, the risks we faces, as discussed in the “Risk Factors” section of this Annual Report.

Purpose-Built for Large-Scale AI Workloads

Our platform and service model are designed for large-scale AI workloads, including model training, fine-tuning and high-concurrency inference. We believe this focus allows us to align our compute services, tooling and operational processes with the performance, stability and usability requirements of GPU-intensive environments.

Proprietary Software-Defined Efficiency

In an industry constrained by the high cost and scarcity of compute resources, we view software optimization as our primary lever for value creation. We have developed a proprietary Intelligent Scheduling System designed to decouple workload performance from raw hardware availability.

6

Key elements of this approach include:

●
Process-Level Orchestration. Unlike conventional schedulers that typically manage resources at the server level, our system is engineered to provide deep, process-level observability and control. It dynamically schedules computing tasks and optimizes memory allocation to address bottlenecks inherent in large-scale cluster training and parallel computing workloads.

●
Focus on Cost-Performance Efficiency. By optimizing kernel execution and mitigating network latency, our platform is designed to achieve utilization rates that significantly exceed standard industry benchmarks for generalized clouds. This efficiency objective allows us to potentially lower the effective total cost of ownership for our customers while maximizing the revenue yield of our deployed capacity.

Asset-Light and Scalable Delivery Model

We prioritize leveraging the underlying hardware resources and operational services of third-party data centers, and integrate those resources through our proprietary software systems and service capabilities to deliver GPU-as-a-Service and related software offerings to end customers, rather than incurring heavy capital expenditures on real estate and direct hardware ownership. This model is intended to provide several operational benefits, including:

●
Agility and Risk Mitigation: Our model enables us to scale capacity in response to customer demand without the long lead times and significant balance-sheet burdens associated with building greenfield data centers or owning depreciating hardware assets. It also provides the flexibility to adapt to new hardware generations, aiming to reduce the risk of technology obsolescence.

●
Focus on Core Competencies: By partnering with top-tier data center operators for physical facilities, we focuses our resources on what we believe are our core differentiators, namely: software orchestration, supply chain integration, and customer service delivery.

Rapid Deployment

Time-to-market is a critical differentiator for our customers in the AI sector. We leverage the extensive experience of our technical leadership team to navigate complex supply chains and accelerate infrastructure delivery across a truly global footprint. Drawing on our leadership’s prior experience deploying large-scale, high-performance computing clusters, we apply specialized execution methodologies to significantly compress deployment timelines compared to industry standards.

Diversified Customer Base

We have strategically cultivated a diversified customer base to enhance commercial resilience and revenue stability. As of June 30, 2026, we served close to 50 distinct customers, with our largest single customer contributing approximately 20.1% of our total revenue. We believe this level of diversification differentiates us in the specialized AI cloud market, where high revenue concentration from a small number of anchor tenants is often a prevalent structural characteristic. We believe our broad customer distribution reduces our dependency on any single entity and validates the widespread applicability of our service offerings. Our diversified portfolio helps us mitigate counterparty risks and maintain more predictable revenue streams amid fluctuating market cycles.

Accessible Service Model

We have architected our product and service framework to democratize access to high-performance AI infrastructure, addressing a significant gap in the market for underserved segments. While many specialized infrastructure providers prioritize massive-scale engagements with high minimum spend thresholds, effectively excluding a large portion of the market, we maintain a flexible engagement model. We offer product configurations, technical support structures, and commercial terms specifically designed to be accessible to small-and-medium-sized enterprises (SMEs) and emerging AI startups. Our service delivery model includes dedicated technical support suited for organizations that may lack the massive internal engineering resources of large technology giants. By providing this level of accessibility, we are able to capture high-growth opportunities within the broader AI ecosystem that are often overlooked by other providers.

7

Infrastructure Expertise and Future-Readiness

In addition to our compute services and revenue-generating GPU cluster management offerings, we have developed certain modular data center, advanced liquid cooling, HVDC power, data center interconnectivity and energy storage solutions that we believe are ready to support customer deployments as of the date of this Annual, although such offerings have not generated revenue to date. We expect to pursue these offerings primarily through partnerships, systems integration, contract manufacturing and other collaborative structures. We believe this approach may allow us to participate in broader AI infrastructure deployments over time while maintaining an asset-light operating model.

Our Principal Products and Services

Our products and services are organized into two primary categories: (i) GPU-as-a-Service (“GaaS”), which includes software-defined AI compute services, infrastructure-level interfaces, and operational tools that enable customers and operators to provision and manage GPU resources at scale; and (ii) AI Infrastructure Solutions, which includes revenue-generating GPU cluster management and operational services for AI data center operators, as well as complementary infrastructure solution capabilities, including modular data center solutions, high-density liquid cooling systems, HVDC power architectures, data center interconnectivity and energy storage capabilities, that management believes are ready for commercial engagement but that have not yet generated revenue as of the date of this Annual Report.

AI Compute Services (GPU-as-a-Service)

Our flagship offering is GPU-as-a-Service (“GaaS”), which provides customers with reserved or on-demand, scalable access to high-performance computing resources. Delivered through our unified control plane, these services are designed to meet the performance requirements of modern AI workloads.

We offer GPU compute configurations through both bare metal instances and virtual machines (“VMs”). We provide single-tenant, bare-metal servers that offer customers direct access to hardware resources without virtualization overhead. This configuration is optimized for large-scale cluster training and performance-critical workloads that require maximum throughput and low latency. We also offer flexible, isolated VM instances suitable for development, testing, and scalable inference workloads. These instances allow for rapid provisioning and efficient resource scaling.

We offer our compute services through multiple commercial models, including on-demand offerings that allow customers to provision capacity on a pay-as-you-go basis for short-term or burst workloads, as well as reserved instance offerings that provide guaranteed capacity and pricing stability for customers with predictable, long-term production requirements. Reserved arrangements typically range from one to three years and are intended to provide customers with supply certainty while providing us with improved revenue visibility.

Our platform supports a range of AI workloads, including large-scale multi-node training, enterprise fine-tuning of pre-trained models and production inference workloads requiring optimized latency and throughput. In connection with these services, we may also provide ancillary services that support compute usage, including networking and storage configuration, operating environment setup, and support services, as required by the customer and within the scope of the service arrangement. The availability of specific configurations and services may depend on supplier arrangements, data center capacity, and operational considerations.

Infrastructure-Level Interfaces and APIs

We provide infrastructure-level interfaces designed for developers and enterprise customers that support automated creation, management, and monitoring of compute resources through APIs. These interfaces are intended to enable customers to programmatically provision and manage resources without accessing a separate management console, including within the customer’s own systems and workflows.

8

Our API capabilities are intended to support, among other things, (i) programmatic provisioning and lifecycle management of compute resources, including GPU and CPU nodes, (ii) Command Line Interfaces (CLIs) that allow developers to provision, manage, and monitor compute resources via code, which enables direct integration with customers’ internal machine learning operations pipelines and CI/CD workflows, (iii) programmatic management and monitoring of customer environments, and (iv) integration of compute resources with networking and storage configurations as supported under the applicable service offering. The scope of API functionality available to any customer depends on the customer’s service configuration, access permissions, and the terms of the applicable arrangement.

Operational and Management Tools Supporting Service Delivery

We operate internal operational and management tools used by our personnel to manage the compute infrastructure and support service delivery. These tools are used to manage server resources and underlying services and to support ongoing operations. Core functions supported by these internal tools include (i) server management and operations, including onboarding, configuration, operational control, monitoring, and inspection workflows, (ii) environment management and operations, including monitoring and management of network conditions, thermal conditions, and power-related parameters, and (iii) supporting functions, including access management, logging, analytics and reporting, and integrations with third-party tools used to support operations. These internal tools are intended to support consistent operational procedures across infrastructure deployed in third-party facilities and to enable us to provision and manage customer compute environments through its platform.

AI Infrastructure Solutions

Our Infrastructure Solutions category includes GPU cluster management and operational services for AIDC operators, as well as certain infrastructure solutions designed to support large-scale AI deployments.

GPU Cluster Management and Operational Services

We provide GPU cluster management services to AIDC operators. These services are intended to assist AIDC operators in deploying, operating, and optimizing large-scale GPU clusters and may be delivered in connection with customer deployments or ongoing operations, depending on the terms of the engagement. These services have generated revenue for us. The scope of our GPU cluster management services may include, as applicable, (i) planning and configuration support for GPU cluster deployments, (ii) operational monitoring and incident response support, (iii) performance tuning and optimization activities, and (iv) operational process support and ongoing assistance. Engagement terms, service scope, and duration may vary depending on the customer’s requirements and the nature of the deployment or operating environment.

Complementary Infrastructure Solution Capabilities

In addition, we have developed complementary infrastructure solution capabilities intended to support large-scale AI deployments, including modular data center solutions, high-density liquid cooling systems, HVDC power architectures, data center interconnectivity and energy storage capabilities. We have made progress in the development and validation of these capabilities, and management believes that certain of these solutions are ready for commercial engagement with prospective customers. However, as of the date of this Annual Report, these capabilities have not been deployed at scale under binding commercial contracts and have not generated revenue. We expect to pursue these offerings primarily on an asset-light basis through partnerships, systems integration, contract manufacturing, and other collaborative structures.

9

Sales and Marketing

Our go-to-market approach is designed to support the delivery of GPU-based compute services and related infrastructure management services to developers, enterprise customers, academic and research institutions, and certain AIDC operators. Our strategy includes a mix of direct sales to enterprise clients and collaborations with cloud service providers and value-added resellers to broaden market reach. We engage customers through a combination of (i) platform-led provisioning of compute services and (ii) direct, service-oriented engagements for certain infrastructure management services. Our go-to-market approach is implemented in conjunction with our sourcing and facility relationships, including third-party GPU capacity suppliers and data center partners.

Customer Acquisition Channels

Our customer acquisition and engagement channels generally include the following:

●
Platform-led onboarding and ordering. For compute services, customers can access our platform to provision and manage compute resources via self-service workflows and interfaces, including API-based provisioning and management.

●
Direct sales and account-driven engagements. For certain enterprise customers and service engagements, we may pursue direct customer relationships that involve structured onboarding, customized configurations, support requirements, or other service terms.

●
Service-driven engagements with AIDC operators. For GPU cluster management services, we typically engage customers through direct service arrangements, which may be structured as project-based engagements or ongoing support arrangements, depending on customer requirements.

The mix of channels utilized for a given customer may depend on the customer segment, workload characteristics, service configuration requirements, and the scope of support requested.

Partnerships and Ecosystem Relationships

Our go-to-market activities are supported by relationships with third parties, which include (i) GPU capacity suppliers, from which we source GPU hardware resources and underlying hardware operations services and integrates them through its proprietary software systems and customer-facing service capabilities to provide GPU-as-a-Service, (ii) data center and facility partners, which provide physical infrastructure inputs such as space, power, cooling, and network connectivity, (iii) technology and service providers, including vendors and tools used to support monitoring, management, security, and operations within our infrastructure environment, and (iv) systems integration, engineering, manufacturing and component partners that may support our Infrastructure Solutions offerings, including modular data center, advanced liquid cooling, HVDC power, data center interconnectivity and energy storage solutions.

These relationships are intended to support our ability to deliver our GaaS and Infrastructure Solutions offerings at scale. Our ability to expand customer engagements may be affected by supplier availability, data center capacity constraints, component availability, partner execution, permitting requirements, and operational integration and deployment timelines.

10

GPU Capacity Supplier Arrangements

Our GPU capacity supplier arrangements are primarily integrated GPU capacity and related infrastructure service arrangements. Under these arrangements, third-party suppliers provide us with access to specified GPU servers or GPU capacity, together with bundled hosting, power, rack space, data center resources, network connectivity and related infrastructure services necessary to operate and make such GPU capacity available. The primary commercial purpose of these arrangements is to obtain access to GPU compute capacity. The hosting, power, rack space, data center resources, network connectivity and similar services provided by these suppliers are bundled infrastructure components that support the operation and delivery of such GPU capacity.

We pay GPU capacity suppliers primarily through usage-based fee arrangements. Depending on the supplier and the applicable order or service arrangement, we may be required to pay deposits or prepayments, or may be invoiced periodically in arrears based on actual usage. Certain supplier arrangements involve rolling monthly usage-based payments, which may be structured as prepaid amounts or invoiced after usage. We do not enter into revenue-sharing arrangements with its GPU capacity suppliers.

We rely on multiple GPU capacity supplier relationships as part of our overall supply model. This multi-supplier approach is designed to support supply continuity, procurement flexibility and access to alternative sources of GPU capacity. We actively evaluate capacity availability across existing and prospective third-party GPU capacity suppliers and seek to diversify our sourcing relationships so that we are not operationally dependent on any single supplier arrangement. We believe that maintaining multiple supplier relationships provides flexibility to source additional or replacement GPU capacity from existing suppliers or alternative third-party providers, subject to market availability, pricing, technical configuration, location, deployment timing and other commercial and operational considerations.

Our GPU capacity supplier arrangements vary in duration. Certain arrangements have longer-term contract periods, while other arrangements may be shorter-term, order-based or subject to rolling monthly usage-based terms. Our GPU capacity supplier arrangements do not include take-or-pay obligations, minimum purchase commitments or exclusivity obligations. Certain suppliers provide non-exclusive priority allocation or preferred access to GPU capacity, which is intended to support supply availability for us. However, such arrangements do not require us to purchase a minimum amount of capacity and do not provide us with exclusive rights to a supplier’s GPU capacity.

Supplier costs under these arrangements are a principal component of our cost of revenue. Pricing under our GPU capacity supplier arrangements is generally subject to market conditions for GPU compute capacity, although pricing may be fixed for short periods or for specific usage periods, orders or capacity configurations. Under certain usage-based arrangements, our unit cost may decrease as usage volume increases. However, we may not be able to maintain or reduce unit costs if GPU supply becomes constrained, supplier pricing increases, utilization levels decline or we are unable to obtain favorable terms.

Our GPU capacity supplier arrangements do not involve revenue sharing, the purchase or lease of GPU equipment or data center equipment from such suppliers, or a long-term lease of data center facilities. Although we rely on multiple supplier relationships as part of our overall supply model, our purchases have been concentrated among a limited number of suppliers. We do not currently believe that any individual GPU capacity supplier arrangement or hosting arrangement described above represents the major part of our requirements for GPU capacity or related infrastructure services.

11

Segment-Specific Engagement Considerations

Our engagement approach may vary by customer segment. Developers and AI-native companies may primarily engage through platform-led ordering and provisioning workflows and may provision compute capacity as needed based on project cycles and workload requirements. Enterprise customers may require structured onboarding, defined service parameters, access controls, and support arrangements consistent with internal operational requirements. Academic and research institutions may have procurement and budgeting processes that differ from commercial enterprises and may require scheduling, data handling, or operational considerations tailored to research workflows and institutional requirements. AIDC operators engaging us for GPU cluster management services typically require operational support for deployment, ongoing operations, and the optimization of large-scale GPU clusters. In addition, certain enterprise customers, infrastructure operators or AIDC operators may engage us in connection with Infrastructure Solutions offerings, including modular data center, advanced liquid cooling, HVDC power, data center interconnectivity and energy storage solutions, which may require coordination with facility infrastructure, third-party partners, and project-specific operational constraints.

Implementation, Onboarding, and Account Management

For compute services, customer onboarding generally includes account setup, access provisioning, configuration of compute environments within our managed platform, and operational coordination based on the customer’s selected service configuration. Customers may expand or modify service configurations over time, subject to our available capacity, operational constraints, and the terms of the applicable arrangement. For GPU cluster management services, implementation typically includes scoping of the engagement, aligning with deployment or operational objectives, and delivering services consistent with the agreed scope and duration. For Infrastructure Solutions offerings, implementation may include solution scoping, design and deployment planning, coordination with facility, engineering, manufacturing or other third-party partners, integration activities, and commissioning or operational support, depending on the scope of the customer arrangement. Our account management activities may include operational coordination, support, and escalation processes, as well as periodic service reviews depending on the engagement structure.

Competition

The industry and markets in which we operate are highly competitive, rapidly evolving and characterized by technological change, capacity constraints and significant capital requirements. We compete in the provision of GPU-based compute services and related Infrastructure Solutions for AI deployments. Our current revenue-generating offerings include GaaS and GPU cluster management and operational services for AIDC. In addition, as we commercialize our modular data center, advanced liquid cooling, HVDC power, data center interconnectivity and energy storage offerings, we expect to compete more directly in additional segments of the AI infrastructure market. Competitive dynamics are influenced by, among other factors, the availability and configuration of GPU capacity, performance and reliability requirements, deployment constraints, including power availability and cooling capacity, service quality, customer support, pricing and commercial terms, security and compliance capabilities, systems integration and deployment capabilities, and the pace of innovation in AI hardware and software.

We face competition from a range of entities with differing business models, operating scales and strategic priorities, including the following categories:

General-Purpose Cloud Computing Providers

We compete with large, diversified cloud service providers that offer GPU-based compute as part of broader cloud platforms. These providers include Amazon Web Services (“AWS”), Microsoft Azure, Google Cloud Platform, Oracle Cloud Infrastructure and IBM Cloud. These platforms typically benefit from significant financial resources, global infrastructure footprints, established customer relationships and extensive product ecosystems. However, their platforms were generally developed to support a wide range of general-purpose workloads and may not be purpose-built for certain AI workloads that require specialized GPU-centric configurations, high-performance interconnects and optimized utilization of underlying infrastructure.

12

Specialized AI Infrastructure and GPU Compute Providers (“NeoClouds”)

We also compete with specialized providers focused on delivering GPU compute and AI-optimized infrastructure services. These providers include CoreWeave, Nebius, Lambda, Crusoe, Voltage Park, WhiteFiber and Hyperstack, among others. Certain of these competitors operate capital-intensive business models that involve owning and operating substantial GPU fleets and, in some cases, developing or controlling data center facilities. These providers may offer deep specialization for AI workloads but may face higher capital requirements and balance-sheet exposure associated with hardware ownership and facility development.

GPU Aggregators and Compute Marketplaces

We also compete with GPU aggregators and marketplaces that aggregate third-party GPU capacity and provide access to compute resources through software platforms. Examples of such providers include Vast.ai, Aethir Cloud, Hyperbolic and similar marketplace-based offerings. These platforms may offer flexible access to distributed capacity but may have more limited involvement in underlying infrastructure deployment, operational management and deep performance optimization, and may have varying levels of control over service quality, reliability and customer experience.

Data Center Operators and AIDC Service Providers

Certain data center operators and AIDC participants offer AI infrastructure services directly or through affiliated service offerings. In addition, some AIDC operators may provide managed cluster services or otherwise compete for customer workloads requiring large-scale GPU deployments. To the extent such operators internalize capabilities that overlap with our services, including GPU cluster management and operational support, they may compete with us for certain customer engagements.

AI Data Center Infrastructure and Solutions Providers

To the extent we pursue commercial engagements for our infrastructure solution capabilities, including modular data center solutions, liquid cooling systems, HVDC power architectures and data center interconnectivity, we may also face competition from established infrastructure equipment manufacturers and solutions providers. These may include providers of data center power distribution, cooling, and modular infrastructure products. These companies generally have significantly greater manufacturing scale, established supply chains, broader product portfolios, and longer track records of commercial deployment than us. Our infrastructure solution capabilities have not yet generated revenue, and there can be no assurance that we will be able to compete effectively in such market segment.

Factors Affecting Competition

We believe that competition in the markets in which we operate is generally based on a combination of factors, including, without limitation:

●
availability of GPU capacity and configuration options, including access to relevant GPU types and deployment timelines;

●
performance and efficiency, including the ability to optimize workload performance and utilization within given infrastructure constraints;

●
reliability and stability, including uptime, operational consistency, and network interconnect performance for cluster-based workloads;

●
operational capabilities and support, including deployment planning, monitoring, incident response, and customer support processes;

13

●
ease of use and integration, including tooling, automation features and API-based controls that reduce operational complexity for customers;

●
infrastructure constraints and deployment considerations, including access to suitable facilities, power, cooling, component availability and partner execution;

●
speed and efficiency of infrastructure deployment, including the ability to design, configure, and commission AI-ready compute environments within compressed timelines;

●
pricing and commercial terms, including flexibility and alignment with customer needs and requirements; and

●
security, data protection and regulatory compliance considerations.

The industry and markets in which we operate are highly competitive, and we face competition from a number of companies and other entities with varying business models, operating scales and strategic priorities. Many of our current and potential competitors have substantially greater financial, technical, marketing and other resources, broader customer relationships, longer operating histories, greater brand recognition, and more established infrastructure than us. As a result, these competitors may be able to respond more quickly to new or emerging technologies, devote greater resources to the development, promotion and sale of their offerings, withstand pricing pressures more effectively, or offer more favorable commercial terms than us. Increased competition could result in pricing pressure, reduced margins, increased customer acquisition costs or loss of market share. For additional discussion of risks related to the competition we face or could potentially face, see the section of this Annual Report captioned “Risk Factors.” Notwithstanding the foregoing, we believe our business model differs from certain capital-intensive providers by sourcing GPU capacity through third-party arrangements and delivering GaaS and Infrastructure Solutions through our managed platform, technology systems and operational processes on an asset-light basis. We intend to compete by applying our technology and operations capabilities to support performance optimization, operational stability, and usability for customers operating GPU-based workloads, including through GPU cluster management services for AIDC operators and, over time, through its complementary infrastructure solution capabilities as those offerings are commercially deployed.

Seasonality

Our business is not inherently seasonal. Demand for GaaS (including its APIs and supporting tools) is driven by ongoing AI training, inference, and production workloads, which are generally non-seasonal in nature. At present, revenue within our Infrastructure Solutions category is derived from GPU cluster management and operational services. Revenues from these services, and from other Infrastructure Solutions offerings if and when they begin to generate revenue, may exhibit period-to-period variability due to the timing of customer capital expenditure decisions, project milestones, deployment schedules, facility readiness, partner execution and delivery timelines. Such fluctuations are primarily project-based rather than seasonal, and we do not believe seasonality has a material impact on our business.

14

Patents

Information concerning our patents and copyright, as of the date of this Annual Report, is set forth below:

Patents

Title
Number
Registration Date
Jurisdiction
Status

COMPUTING POWER NETWORK SYSTEM
12,058,179
August 6, 2024
United States
Granted and in force

AIOPS SCHEDULING METHOD AND SYSTEM BASED ON MULTI-AGENT COLLABORATIVE AUTONOMY
19/440,612
January 5, 2026
United States
Pending

AN ADAPTIVE EXTERNAL SUPPLY-AND-RETURN WATER TEMPERATURE REGULATION SYSTEM AND METHOD FOR A MODULAR DATA CENTER
19/440,577
January 5, 2026
United States
Pending

A METHOD AND SYSTEM FOR MAXIMIZING THROUGHPUT OF A GPU CLUSTER
19/444,167
January 8, 2026
United States
Pending

Software Copyright

Title
Case Number
Application Date
Jurisdiction
Status

EXASCALE ARTIFICIAL INTELLIGENCE COMPUTING POWER MANAGEMENT PLATFORM
1-15021399291
October 15, 2025
United States
Pending

Regulation

We are subject to the laws and regulations of various jurisdictions and governmental agencies affecting our operations, products and services including, but not limited, laws relating to AI, intellectual property, tax, import and export requirements, anti-corruption, economic and trade sanctions, national security and foreign investment, data privacy and security requirements, competition, advertising, employment, product regulations, environment, health and safety requirements, and consumer laws. A discussion of the risks related to such laws is set forth in the “Risk Factors” section of this Annual Report, as supplemented by the discussion below. To date, costs and accruals incurred to comply with regulations have not been material to our capital expenditures and results of operations. Although there is no assurance that existing or future governmental laws and regulations applicable to our operations, products and services will not have a material adverse effect on our capital expenditures, operating results, and competitive position, we do not currently anticipate material expenditures for compliance with regulations. Nonetheless, we believe that global trade regulations could potentially have a material impact on our business.

As a global company, the import and export of our products and services are subject to laws and regulations including international treaties, U.S. export controls and sanctions laws, customs regulations, and local trade rules around the world. The scope, nature, and severity of such controls varies widely across different countries and may change frequently over time. Such laws, rules, and regulations may delay the introduction of products and services or impact our competitiveness through restricting our ability to conduct business in certain jurisdictions or with certain entities and individuals. For example, the U.S. Department of Commerce continues to tighten export controls and add firms to the “Entity List.” These export restrictions, which would require that we obtain licenses from the U.S. Department of Commerce to allow us to export infrastructure services to such listed firms, could limit or prevent us from doing business with certain potential customers or potential suppliers. These restrictive governmental actions and any similar measures that may be imposed on U.S. companies by other governments could limit our ability to conduct business globally.

15

Our operations, and the third-party data center and partner-operated facilities in which our GPU capacity is deployed, are subject to laws and regulations of various jurisdictions and governmental agencies, including local, state, and federal environment laws, health and safety requirements. These requirements may relate to, among other things, the siting, build-out and operation of data center facilities, power and cooling infrastructure, and the handling and disposal of certain equipment and materials. In addition, because a significant portion of our cost structure is driven by third-party facility inputs (including colocation, power, cooling and network connectivity), changes in environmental laws at the local, state, and federal level, regulations or permitting requirements applicable to such facilities or related utilities could increase our operating costs, delay deployments, or otherwise affect our ability to scale capacity on expected timelines.

To date, costs and accruals incurred to comply with governmental regulations (including the environment, health and safety requirements described above) have not been material to our capital expenditures and results of operations, and we do not currently anticipate material expenditures for compliance with regulations. However, there can be no assurance that existing or future governmental laws and regulations applicable to us or our products and services will not have a material adverse effect on our capital expenditures, operating results, and competitive position.

We have developed infrastructure solution capabilities, including high-density liquid cooling systems designed to lower power usage effectiveness and HVDC power architectures intended to improve overall energy efficiency and facilitate better integration with renewable energy sources and grid-scale storage. To the extent we commercially deploy these infrastructure solution capabilities, such deployments may be subject to additional regulatory, permitting and compliance requirements, including those related to electrical systems, building codes, environmental standards, and equipment safety certifications, which could vary by jurisdiction and may affect deployment timelines and costs. As of the date of this Annual Report, these capabilities have not been commercially deployed and we have not incurred material regulatory compliance costs in connection with these capabilities.

Corporate Information

We are a Delaware corporation. Exascale Labs Holdings Inc. was incorporated under the name “D. Boral ARC Merger Corporation” as a Delaware corporation on December 19, 2025, and Legacy Exascale was incorporated as a Delaware corporation on June 1, 2022. Through the Business Combination, Exascale Labs Holdings Inc. became the combined company of the Business Combination and succeeded to the business of Legacy Exascale. Our principal executive office is located at 820 Gessner Road, Suite 332, Houston, Texas 77024 and our telephone number is (650) 537-7553. Our corporate website address is www.exascalelabs.ai. Our website and the information contained on, or that can be accessed through, our website is not deemed to be incorporated by reference in, and is not considered part of, this Annual Report.

Employees

As of the date of this Annual Report, we had
12 full-time employees on a consolidated basis. Of these employees, five were directly employed by us in the United States, and seven
were employed by our wholly owned Singapore subsidiary. The subsidiary employees work remotely in support of our business operations.

Our workforce currently consists of (i) one Chief Executive Officer, (ii) one Chief Financial Officer; (iii) one data center partnerships lead; (iii) one research and development lead; (iv) one marketing lead; (v) one business development lead; (vi) four senior research and development engineers; and (vii) two operations personnel.

Our core research and development activities are led and performed by internal personnel, including our research and development lead and senior research and development engineers. Customer service and support are also handled internally by our personnel.

We also utilize certain third-party service providers to support specific functions. For example, we engage project-based outsourced development teams to assist with specific development tasks, certain third-party personnel to provide on-site data center operations support, and third-party service providers to assist with certain finance-related execution functions. These arrangements supplement our internal workforce and do not replace internal responsibility for management, core research and development, customer support, or overall business oversight.

16