NASDAQ: SHAZW

Access
to Data Center Capacity in a Capital Efficient and Speed to Market Approach: Rather than incurring large capital expenditure and
multi-year investment in building its own data centers, the Company deploys its infrastructure directly into the facilities of data center
partners such as NEXTDC, under long-term contractual arrangements. This enables us to scale capacity at speed, through rapid provisioning
of resilient, high-density compute capacity and surrounding infrastructure essential for AI workloads;

The
Sovereign AI Advantage: As governments and regulated industries increasingly mandate data sovereignty, we serve as a trusted domestic partner. We provide a secure solution for organizations that must adhere to strict Australian privacy
laws; and

Purpose-built
AI Infrastructure that drives Cost Efficiency for Customers: Our platform is engineered specifically for modern AI and HPC workloads, integrating high-bandwidth networking, low-latency
fabrics and storage optimized for large-scale model training and inference. We have demonstrated an ability to deliver higher efficiency
and performance from advanced GPU infrastructure.

The demand for our specialized model has been evidenced by our engagement with
global industry leaders. Recent commercial wins with high-growth AI native Canva and industry participant GMI Cloud US Inc., demonstrating
our ability to serve customers that have large, sophisticated AI and HPC requirements.

Company
history

SharonAI,
Inc. (“SharonAI”), the Company’s current operating and wholly-owned subsidiary, is a corporation formed in Delaware
on February 15, 2024, with the intent to act as a holding company to acquire various assets focused on or in the High Performance Computing
(“HPC”) industry and the AI field of technology, and currently one of Australia’s
leading neoclouds, a cloud infrastructure provider that focuses on specialized, high-performance compute—especially GPU-heavy workloads
for AI, machine learning, and HPC. HPC is a computing technology that uses clusters of processors or processor cores working in parallel
to solve advanced computational problems across a wide range of scientific, engineering, finance, business and other fields. SharonAI
is specifically focused on infrastructure and technology associated with the development and delivery of these HPC/AI services to users
and applications which require both large amounts of Graphic Processing Units (“GPU”) and Central Processing Units (“CPU”),
combined with data storage. CPUs are general purpose processors while GPUs are optimized for parallel processing and were originally
used for computer graphics. Data storage is used to store the large data sets common in HPC/AI and to back up information. SharonAI’s
two main business lines are an AI/HPC cloud platform, which is based in Australia, and the development of data center assets, which is
based in the U.S., each as described further below.

-6-

Key
operational and strategic milestones

In
March of 2024, we formed two new wholly owned subsidiaries in Delaware, SharonAI Operations LLC, which is intended to be used for
U.S. based operational activities as its operations expand to the U.S., and SharonAI Hosting LLC, which is intended to be used to hold
assets that are acquired in the future and based in the U.S.

In
April of 2024, we acquired 100% of the issued capital of Alternative Asset Management Pty Ltd ACN 645 215 194, an Australian company
that was renamed SharonAI Pty Ltd (“SAIPL”) and which has a business operating distributed data storage a type of cloud storage
that utilizes Web 3 technology to provide decentralized networks of independent nodes to securely store and retrieve data, ensuring redundancy,
fault tolerance, and resistance to censorship while incentivizing storage providers with blockchain-based rewards. This acquisition was
part of a transaction in which SAIPL also obtained certain assets from Digital Income Fund Pty Ltd ACN 643 155 328 as trustee for the
Digital Income Fund ABN 12 771 427 247 (“DIF”), an Australian company which had storage servers and ancillary equipment for
the operation of the distributed storage operations. In addition to these assets, SAIPL had acquired a Tier 3 designed modular data center,
although it had not fully paid for the equipment at the time of acquisition.

In
June 2024, we acquired a controlling interest in Distributed Storage Solutions Limited (‘DSS’), an Australian
company established in 2021 that specialized in distributed cloud data storage and HPC and AI infrastructure. This was a critical step
in delivering storage infrastructure, technical expertise and accelerating deep industry relationships. DSS had been providing decentralized,
enterprise-grade storage systems for customers for multiple years and brought us an operational team and strategic partnerships
with NVIDIA and Lenovo.

Between
June and December 2024, we acquired 192 NVIDIA L40S GPUs, materially expanding
the company’s GPU fleet size. Over this time, we observed and confirmed positive unit economics for the GPU fleet which led to a
decision to further expand into GPU compute. This involved planning for the next GPU form factor acquisition, which was a group of 160
NVIDIA H100s.

In December 2024, we completed testing using older models of NVIDIA GPUs to
demonstrate that it could successfully deliver HPC use cases under NVIDIA reference architecture. This milestone led to us being certified
as an NCP partner. We are one of three NCPs currently operating in Australia, part of a global network of cloud partners.

We continued to build proprietary technology beyond the provision
of hardware and storage. In February 2025, we launched the SharonAI Cloud, an orchestration and automation platform, enabling
self-provisions GPU compute for AI training, inference, Visual Effects (‘VFX’) rendering and complex HPC workloads.
This marked our transition from a purely supply hardware capacity to an AI infrastructure provider that allows customers
to deploy GPU compute autonomously and at scale.

In
March 2025, we announced the development of a supercluster (‘Supercluster’), which is an interconnected
network of a material number of specialized processors and high-speed memory constructed on NVIDIA reference architecture, built to function
as a single and powerful computer. Designed as a system of more than 1,000 GPUs hosted inside NEXTDC’s M3 data center, the Supercluster
is dedicated to large-scale AI and HPC workloads.

In
October 2025, we entered into a Memorandum of Understanding (‘MOU’) with Cisco to establish a strategic
collaboration for managed enterprise cloud AI solutions. Under the terms of the MOU, Cisco will provide technical guidance on
architecting enterprise-ready AI data centers, facilitate customer introductions, and offer solution engineering support for complex
opportunities deploying Cisco products and technologies. This strategic collaboration marked our evolution from a pure
infrastructure provider to a turnkey enterprise AI solution partner with direct channel access to Australia’s largest
corporate and government customers through Cisco’s sales network.

In
December 2025, we through our Australian subsidiary SharonAI Pty Ltd, entered
into an agreement with strategic partner NEXTDC to materially expand upon its existing data center footprint, including the Supercluster,
with up to 50MW of additional capacity in NEXTDC’s data center facility, providing us with access to data center availability that
will enable us to expand and deploy more than 20,000 NVIDIA B200, B300, or GB300 GPUs.

Over
the course of 2025, we deployed the Supercluster network and Spine at the NEXTDC M3 Data Center, and also acquired 16 NVIDIA H200s,
which were primarily used for customer latency testing and proof of concepts. we also broadened its strategic partner ecosystem
through further partnerships with VAST and Megaport to position itself as one of Australia’s leading Neoclouds.

In
January 2026, we signed its first major customer contract with lighthouse customer, Canva and industry participant GMI Cloud US Inc.
We are continuing the strong momentum and is aiming to convert its pipeline of customers into signed agreements to materially scale
the business.

We believe that we are well positioned to win new customers and achieve significant scale in the Asia-Pacific region, supported by its product
and service offering and sector tailwinds driven by anticipated structural demand for accelerated computing. This growth is underpinned by the strategic engagements which we have in
place.

-7-

Key
corporate milestones

On
January 28,2025, Roth CH Acquisition Co., a publicly traded Cayman Islands company trading on the OTC Market (Roth CH), entered
into a business combination agreement, with Roth CH Merger Sub, Inc., a Delaware corporation and a wholly-owned subsidiary of Roth CH
(Merger Sub), SharonAI Inc. and Roth CH Holdings, Inc. (Roth CH Holdings) (the BCA). Under the BCA, Roth CH merged
with and into Roth CH Holdings on 16 December 2025 and was renamed “SharonAI Holdings Inc.” and become domesticated in the
State of Delaware, and Merger Sub merged with and into SharonAI Inc. becoming the wholly owned subsidiary of the Company. The transaction
was completed in December 2025.

As
a result of the BCA transaction described above, equity holders of SharonAI Inc. received the Company’s securities. Shares of SharonAI Holdings Inc. Class A Ordinary Common Stock began trading on the OTC Markets under the ticker symbol
“SHAZ” on December 18, 2025.

To
fund our expansion, we have undertaken two material capital raisings, and a divestment of a non-core asset.

In
December 2025, we successfully completed a US$103 million pre-initial public offering (‘Pre-IPO’) capital
raising in the form of unsecured convertible notes, introducing new institutional and strategic shareholders. As part of this transaction
Digital Alpha Advisors LLC, which has a strategic collaboration agreement with Cisco, invested in us via the unsecured convertible
note and remain strategic investors in the Company.

Also
in December 2025, we announced a transition from a hybrid model of site development to a pure-play Neocloud operator for the
immediate term. We had previously formed a 50:50 joint venture, Texas Critical Data Center LLC (‘TCDC’), with New
Era Energy & Digital Inc (‘New Era’) in January 2025, to fund and develop a data center site with a natural
gas fired power plant in the Permian Basin of western Texas. SharonAI sold its 50% interest in TCDC to its joint venture partner,
New Era, for a consideration of US$70 million, paid via a combination of cash, a secured convertible promissory note and equity in
New Era. The transaction was completed in January 2026.

On January 19, 2026, we announced a potential investment from Digital Alpha of up to $200 million and a strategic
technology partnership with Cisco, subject to execution of definitive documentation which is expected to enable the Company to further
accelerate customer deployments and expand our cloud infrastructure for enterprise AI and high-performance compute in Australia and Asia
Pacific.

On
January 22, 2026, Mr. Wolfgang Schubert, resigned as our Chief Executive Officer and from our board of directors (the
“Board”). In connection with Mr. Schubert’s resignation as Chief Executive Officer of the Company, on January 22,
2026, the Board appointed Mr. James Manning, Non-Executive Chairman, director and greater than 10% stockholder of the Company, as
our Chief Executive Officer.

On
January 22, 2026, we announced that
USD.AI had approved a debt facility of up to US$500 million for SharonAI, a subsidiary of the Company, subject to execution of
definitive documentation. The facility is expected to enable us to access asset-backed, non-recourse financing through USD.AI’s
on-chain credit system, which is expected to allow approved GPU deployments to be financed with stablecoin liquidity. The structure is
designed to support capital-efficient expansion of AI infrastructure while reducing reliance on traditional bank and private credit markets.

In
February 2026, we listed on the NASDAQ Capital Markets, raising US$125 million in a concurrent underwritten public offering before
costs. This transaction was a key strategic step that is expected to enable us to access to the largest public capital market in the
world, providing us with capital raising alternatives which could lower its weighted average cost of capital while minimizing
near-term equity dilution.

Industry
Overview

Cloud
computing is at its core the delivery of compute and storage servers and software applications over the internet. It has gained popularity
in part due to the flexibility around paying only for what customers need and outsourcing the acquisition and management of the infrastructure.
Traditional cloud computing is based on CPUs but more recently, cloud GPU computing, or GPU-as-a-Service, has increased, propelled by
demand for high-performance computing across a range of data-intensive and computationally complex applications. Enterprises and researchers
increasingly rely on advanced GPUs to power AI, machine learning, data analytics, computer vision, scientific simulations, and other
tasks that require large-scale parallel processing. This has led to growth in both the demand and supply of on-demand GPU resources.
This combination of increasing workloads that demand parallel processing and evolving consumption models that make GPUs more accessible
has caused industry growth, creating specialized GPU-as-a-Service and GPU cloud platforms including the competitors listed below. The
AI industry also faces challenges, with concerns including bias and misuse as well as around the environmental impact of the growing
data center industry.

Cloud
storage is an integral part of cloud computing which increasingly deals with large data sets. One of the first modern cloud services
offered was Amazon Web Services’ Simple Storage Service (S3). There are different use cases for storage, including high-performance
storage optimized for rapid data access and backup storage for data that is accessed less frequently.

Data
centers are used to host the compute and storage servers, providing security, electricity, cooling and network connectivity. According
to a research report by CBRE, the rise of AI compute is driving demand for power-intensive infrastructure, including a premium on energy-efficiency
capabilities such as liquid cooling over air cooling. In addition, tertiary and rural markets have seen increased deal activity for powered
land. The recently announced Project Stargate highlights the strategic importance of data center infrastructure and President Trump announced
plans by the government and private industry to invest up to $500 billion over the next four years in the U.S. as part of the project.
The demand for data center infrastructure has also created challenges around the supply chain and the procurement of critical components,
including on the power generation side.

-8-

Products
and Services

We provide access to a range of accelerated computing infrastructure solutions specifically targeted towards AI and HPC applications for
enterprises, hyperscalers, government and research institutions.

We offer three key solutions designed for AI and HPC:

1.GPU-as-a-Service
(‘GPUaaS’);

2.AI
Studio (Platform-as-a-Service (‘PaaS’)); and

3.Cloud
Storage Solutions.

GPU-as-a-Service

Our
flagship infrastructure service provides scalable, on-demand access to high performance GPU cloud compute. For customers, this is more
efficient for performance and affordability as the solution is designed to accelerate complex AI workloads and can be accessed on-demand
without needing to own or maintain any physical hardware and infrastructure. The service is engineered to accelerate complex AI workloads
across multiple use cases including:

Model
Training: Large language model (‘LLM’) pre-training, fine-tuning and model adaptation;

Inference:
LLM inference optimization, batch processing and real time inference;

Research
Computing: Scientific simulations, genomics analysis, climate modelling; and

Visual
Computing: 3D rendering, video processing and VFX.

Our
current GPU offering includes NVIDIA L40s, A40, H100 NVL, H200, B200, and B300 processors, with additional AMD MI300X capabilities with
each GPU type optimized for specific use cases.

AI
Studio (Platform-as-a-service)

The
SharonAI Studio is our proprietary PaaS combining cloud infrastructure
with expert AI, ML and HPC operational support. The platform represents a key point of differentiation for us versus other domestic peers,
providing leading platform capability spanning the complete AI infrastructure stack.

SharonAI
Studio delivers end-to-end AI development and deployment capabilities through a unified interface, including:

Unified
Development Environment: Interactive development frameworks (Jupyter, RStudio), multiple programming languages (Python, R, Julia);

Pre-configured
AI Frameworks: CUDA, TensorRT, cuDNN, PyTorch, TensorFlow, ONNX which are required to efficiently build, train, optimize and run advanced
AI workloads;

Specialized
Capabilities: Inference optimization, LLM deployment, neural inference models, fine tuning frameworks;

Bare
Metal, Virtual Machines, Containers, and Kubernetes: Full range of deployment options from bare metal GPU servers to containerized Kubernetes
clusters;

Expert
Support: Access to SharonAI’s AI systems administrators and cloud infrastructure engineers; and

-9-

NVIDIA
AI Enterprise Software Stack: Deep integration with NVIDIA’s proprietary AI platform providing enterprise grade security
and tooling.

SharonAI
Studio has a deep integration of proprietary and partner features on a number of releases, such as the Megaport AI Exchange,
enabling easier customer interaction with the platform and driving anticipated sales velocity. The platform enables customers to
focus on AI outcomes rather than management of infrastructure complexity.

Cloud
Storage Solutions

We provide highly scalable and cost-effective cloud storage designed for large scale AI and HPC datasets. Services include:

S3
Compatible Cloud Storage: High-capacity object storage with S3 API compatibility for seamless integration with existing workflows;

High
Performance SSD Storage: Low latency storage optimized for training and inference workloads; and

Archive
and Backup: Cost-effective long-term storage for historical data and compliance requirements.

Storage
integrates with Our compute infrastructure through partnerships with leading storage providers including VAST, whose AI
Operating System (InsightEngine) unifies storage, database, and runtime, bringing together all services needed to run AI pipelines at
scale, including retrieval augmented generation capabilities. We are early to market in Australia with a locally hosted enterprise
grade agentic AI and inference engine, courtesy of the partnership with VAST.

Target
Customers

We target organizations undertaking advanced AI and HPC computing workloads, including enterprises, hyperscalers, government and research
institutions. These customers operate at computational scales that demand deterministic performance, guaranteed uptime, sovereign data
residency and the ability to scale rapidly. The SharonAI Cloud platform is engineered to meet these requirements through non-contended
GPU resources and enterprise-grade orchestration tools to support mission-critical applications including complex AI training, real-time
inference, scientific modelling and industry-specific analytics.

As
we continue to broaden its commercial footprint, we are actively expanding relationships across a diverse set of
potential customers. Its service offerings are designed to serve the full spectrum of AI users – from individual developers
leveraging pre-configured ML environments, to enterprise and government customers that require secure, sovereign and
production-ready GPU infrastructure.

The
primary target customer segments are outlined below.

Types of Customers

Customer
Type

Description

Use
Cases

Enterprises

-

Corporations
are integrating AI into their workflows and require scalable infrastructure for model training and inference

ML
model training, inference, data analytics, generative AI applications

Hyperscalers

-

Large
global cloud and internet platform companies that procure GPU and data center capacity in massive contiguous blocks to run and scale
their own cloud services and AI workloads.

Building
dedicated AI clusters, inference at scale, model serving

AI
Labs

-

AI research firms dedicated to researching, developing
and applying AI, and require scalable infrastructure for model training and inference

ML
model training, inference, data analytics, generative AI applications

Research
Institutes & Universities

-

Academic and scientific organizations conducting complex
simulations and data-intensive research that depend on high-performance parallel processing capabilities

LLM
development, model research, compute intensive proof of concepts

Governmental
Authorities

-

Public
sector bodies seeking sovereign cloud capabilities and secure infrastructure for sensitive computational tasks

High
performance computing, climate modelling, genomics, scientific simulation

AI
Start-ups and Developers

-

Early-stage companies and individual developers who need
flexible, on-demand access to powerful GPU resources to build and test new applications without large capital outlays

Sovereign
AI infrastructure, secure computing, defense applications

GPU
Aggregators & Marketplaces

-

Platforms that aggregate GPU capacity from various providers
to serve a broad user base. These customers provide SharonAI with immediate access to a wide demand pool

Immediate
access to broad customer pools, spot and short-term market participants

-10-

Customer
Value Proposition

We deliver a differentiated value proposition centered on high-performance accelerated compute, sovereign Australian infrastructure, a
frictionless and developer-ready cloud experience, and enterprise-grade operational support. Together, these capabilities position the
Company as a trusted GPUaaS provider for organizations seeking secure, scalable and high-throughput AI infrastructure.

Our
customer value proposition:

-High-performance
compute: the Company’s GPU Supercluster, engineered to NVIDIA reference architecture
provides advanced sovereign AI compute platforms, enabling large-scale training with high-bandwidth,
low-latency interconnects. This is optimized for non-contended, dedicated GPU performance
to enable mission-critical enterprise and research workloads;

-Organizations
sovereign proposition: sovereign data residency within NEXTDC Tier IV-certified data centers
ensures physical security, interconnectivity and zero downtime. Our infrastructure
is purpose-built for secure, compliant AI compute at scale for organizations that must adhere
to strict Australian privacy laws;

-Ease
of use and speed: the SharonAI Cloud provides an API-first orchestration layer that enables
customers to self-provision bare metal GPU nodes with minimal setup time. This dramatically
reduces the time to train AI workloads. Pre-configured environments with popular frameworks
in addition to NEXTDC’s AXON interconnection fabric provides fast, low-latency connectivity
for customers looking to scale compute quickly; and

-Specialized
customer support capabilities: we provide customers with operational certainty, 100%
uptime expectations and enterprise-grade support as they transition from experimentation
to production, particularly in sectors requiring fine-grained permissions, strict compliance
and specialized infrastructure guidance.

Customer
Acquisition Strategy

We employ a multi-channel customer acquisition approach:

1.NVIDIA
Consumption Desk Referrals: NVIDIA actively refers customers with NCP capacity based on
geographic requirements and workload fit;

2.Cisco
Channel Integration: Cisco’s ANZ enterprise sales team are intended sell SharonAI powered cloud
services into the enterprise customer base by leveraging Cisco’s products and technologies,
with strength in government contracts;

3.Lenovo
Partnership: Lenovo’s technical teams and sales organization support customer development;

4.Direct
Enterprise Sales: Our sales team targeting hyperscalers, large enterprises,
government and AI labs alike; and

5.Research
& Education Partnerships: Academic partnerships provide reference customers and ecosystem
credibility.

-11-

Compute
Infrastructure

We work closely with NVIDIA to ensure compute infrastructure meets their stringent NVIDIA Cloud Partner reference architecture. This architectural
design delivers maximum throughput and minimum latency across the entire cluster, ensuring no bottlenecks across compute, storage or
network IO.

Critical
AI, ML and HPC tasks rely on end-to-end latency and throughput guarantees. NVIDIA’s reference architecture ensures that valuable
GPU compute time is not wasted over the course of large, complex workloads, particularly where large data transfers would typically slow
down traditional cloud or enterprise networks.

NVIDIA’s
non-blocking leaf and spine network design is architected specifically for the desired cluster size, and is delivered in entirety as
the first step of a cluster build. This ensures that growth of the cluster from the first GPU to the last meet the required performance
levels no matter what stage of the rollout, and compute resources are available as they come online without delay at maximum performance.

We collaborate with a number of GPU, technology and digital infrastructure providers to construct and deliver its compute infrastructure.

As
a certified NCP, we are technically aligned with NVIDIA’s accelerated computing roadmap and is capable of deploying and
operating high-performance GPU infrastructure at scale. We build our GPU cloud platform around NVIDIA’s latest generation
of training and inference accelerators and leverages NVIDIA’s reference architectures to ensure optimal performance, efficiency
and interoperability for enterprise and research customers.

Central
to this partnership is the SharonAI Supercluster, a 1,016-GPU deployment built to NVIDIA reference architecture specifications and
designed to support complex AI training workloads with high-bandwidth interconnects, low-latency communication and sovereign hosting
requirements. This is co-located at NEXTDC’s M3 data center in Melbourne, Australia and currently features NVIDIA H200 GPUs
interconnected with NVIDIA Quantum-2 InfiniBand networking for high-speed, low-latency performance required for large-scale model
training and inference. The balance of this cluster is expected to consist of NVIDIA B200 GPUs expected to come online in the
first half of 2026.

By utilizing NVIDIA-accelerated platforms and software frameworks, We integrate
pre-configured ML environments, orchestration tooling and enterprise-grade APIs into its cloud, reducing time-to-deployment for customers
and ensuring compatibility with the rapidly evolving NVIDIA AI software stack. This strategic alignment is designed to allow us to scale
its GPU cloud footprint quickly and reliably, providing customers with the performance and flexibility required to operationalize generative
AI pipelines, HPC applications and other latency-sensitive workloads.

We are also deploying a 1,024-unit NVIDIA B300 GPU cluster which
will be located at NEXTDC’s S3 data center and expected to be coming online in the first half of 2026. This cluster is also built
in partnership with Cisco and is expected to be Australia’s first Cisco Secure AI Factory.

GPU-as-a-Service
Architecture

We operate a focused GPU-as-a-Service platform spanning the core layers of the AI infrastructure stack required to deliver compute capacity
to end customers.

Unlike
traditional hyperscalers that typically offer broad Infrastructure-as-a-Service (‘IaaS’) with AI as one component, Neoclouds
like SharonAI are purpose built from inception for high density, low latency GPU compute. This specialization manifests in:

Reference
Architecture Deployment: Building to NVIDIA’s proprietary specifications for optimal performance;

Thermal
Design Optimization: Advanced cooling solutions (liquid to chip) supporting high density GPU clusters;

Network
Architecture: Spine and leaf network topology with high bandwidth, low latency interconnects optimized for distributed training; and

AI
Centric Software Stack: Orchestration layers, inference engines, and developer tools designed specifically for AI/ML workloads.

-12-

Hardware
& Infrastructure Layer

We
acquire, configures, and deploys specialized GPU and CPU hardware optimized for AI and HPC workloads.
We partner with leading colocation and data center providers (for example NEXTDC, GreenSquare and DigiCo) to gain access to power, space,
and network infrastructure. We currently operate 411 GPUs housed in 51 servers with over 59 petabytes of storage capacity across co-location
data centers in Australia. This operational fleet is distinct from the larger, future deployment of B-Series and GB-Series NVIDIA GPUs
outlined below which is expected to take the total number of GPUs by the Company deployed to approximately 2,435, which is almost 6 times
the amount of GPUs since the second half of 2025.

SharonAI’s
– Infrastructure Platform

At December 31, 2025, we had 411 GPUs deployed
and generating revenue, with this operational fleet being distinct for the larger, future deployment of B-Series and GB-Series
NVIDIA GPUs outlined below.

Existing
GPU Fleet

GPU
Model

Quantity
Deployed

Release
Year

NVIDIA
A40

43

2020

NVIDIA
L40

192

2022

NVIDIA
H100

160

2022

NVIDIA
H200 (SXM)

16

2024

Total

411

From
the proceeds raised under the December 2025 Pre-IPO Convertible Note offering, we purchased the following GPU fleet:

GPU’s
to be Deployed

GPU
Model

Quantity

Release Year

Expected
Delivery

NVIDIA
B200

1,000

2024

1H
2026

NVIDIA
B300

1,024

2025

1H
2026

Total

2,024

The NVIDIA GPUs are expected to be deployed as
contiguous ‘Superclusters’ using the over 54MWs of available capacity secured by us across their Colocation Data
Center portfolio in Australia.

Summary
of GPUs and CPUs used in our platform

Model

Description

Key
Features

Release
Year

Image

NVIDIA
L40s

-
The NVIDIA L40s GPU provides high-performance visual computing capabilities within data center
environments, facilitating the execution of diverse and demanding computational workloads.

-
Utilizing the NVIDIA Ada Lovelace architecture, the L40s is engineered to support a broad spectrum of applications, including 3D
design, complex simulation, AI-enhanced graphics, and advanced data science.

-
Built on the NVIDIA Ada Lovelace architecture

-
Delivers up to 1,466 TOPS (FP8 Tensore Core, with sparsity)

-
48 gigabytes (“GB”) of GDDR6 memory

-
Use cases include generative AI, LLM inference, LLM fine-tuning and small-model training, NVIDIA omniverse enterprise, rendering,
3D graphics, streamlining, and video content

2022

-13-

NVIDIA
A40

-
The NVIDIA A40 is a high-performance data center GPU engineered on the NVIDIA Ampere architecture,
integrating RT Cores, Tensor Cores, and CUDA Cores.

-
The unit features 48 GB of graphics memory, designed to facilitate the execution of complex visual computing and data-intensive workloads.

-
The hardware incorporates advanced NVIDIA RTX technology, providing professionals with the capacity for high-fidelity visualization
and industrial innovation.

-
Built on the NVIDIA Ampere architecture

-
48GB of GDDR6 memory

-
Up to 299.4 TFLOPS (FP16 Tensor Core)

-
Used for virtual workstations, 3D rendering, AI training, data science, visual computing

2020

NVIDIA
H100

-
Engineered on the NVIDIA Hopper architecture, the H100 NVL is specifically optimized to facilitate
the high-density computational scaling required for LLM’s and advanced artificial intelligence
applications.

-
The hardware provides up to a 30-fold increase in AI inference performance compared to prior-generation architectures.

-
Built on the NVIDIA Hopper architecture

-
Features 94GB of HBM3 memory

-
Up to 3,341 TFLOPS FP8 performance

-
Used for LLM inference, AI training, HPC, scientific computing

2022

-

NVIDIA
H200

-
The NVIDIA H200 features a significant onboard memory capacity of 141 GB, providing the hardware
foundation necessary to store and process the vast data sets required for modern digital
operations

-
The hardware is specifically engineered for professional AI and HPC environments, supporting reduced energy consumption and improved
infrastructure sustainability.

-
Built on NVIDIA Hopper architecture

-
141GB of HBM3e memory

-
Up to 4 PetaFLOPs of FP8 performance

-
Used for AI inference, LLMs, scientific computing, HPC workloads

2024

-

AMD
MI300X

-
Engineered on the AMD CDNA 3 architecture, the AMD MI300X is designed to facilitate the high-density
computational requirements of generative AI, LLM’s and HPC workloads.

-
Each unit features an onboard memory capacity of 192 GB, providing the hardware foundation necessary to process expansive datasets
within a single accelerator environment while supporting optimized total cost of ownership for enterprise-grade AI deployments.

-
Built on next-gen AMD CDNA 3 architecture

-
192GB of HBM3 memory

-
Used for AI / ML training, generative AI, large language models, HPC

2023

-

-14-

NVIDIA
B200

-
Engineered on the NVIDIA Blackwell architecture, the B200 GPU incorporates 180 GB of HBM3e
memory and advanced fifth generation Tensor Cores to facilitate large-scale artificial intelligence
training, scientific simulations, and complex ML tasks.

-
The unit is designed for deployment within enterprise GPU clusters and AI infrastructure, providing the computational density required
for high-speed AI inference and compute-intensive industrial applications.

-
Built on next generation NVIDIA Blackwell architecture ‘Base GPU’

-
180GB of HBM3e memory

-
Delivers up to 9 PetaFLOPs of FP8 performance

-
Engineered for demanding AI training and inference workloads

2024

NVIDIA
B300

-
The NVIDIA B300 is a next-generation data center GPU in the Blackwell Ultra series, engineered
to accelerate demanding AI and large-scale compute workloads.

-
It features an expanded 288GB high-capacity HBM3e memory configuration and enhanced architecture optimized for AI inference and training
at large scale. The B300 is designed for deployment in enterprise GPU clusters and hyperscale AI infrastructure.

-
Built on next generation NVIDIA Blackwell architecture ‘High-Spec GPU’

-
288GB of HBM3e memory

-
Delivers up to 72 PetaFLOPs of FP8 performance

-
Engineered for extreme-scale AI inference and reasoning

- 2025

NVIDIA
GB300 (Anticipated)

-
Engineered on the NVIDIA Blackwell Ultra architecture, the GB300 utilise a dual-reticle design
incorporating over 20,000 CUDA and Tensor cores and 288 GB of HBM3e memory to facilitate
high-capacity AI training and inference workloads

-
The hardware delivers a peak memory bandwidth of approximately 8 TB/s and achieves up to a 50% increase in performance compared to
the predecessor GB200.

-
Built on next generation NVIDIA Blackwell architecture Super Chip

-
System-level “AI Factory” deployments

2025

Network
& Connectivity Layer

Network architecture is engineered
to NVIDIA reference design specifications to ensure optimal performance of GPU clusters. The Company collaborates with NVIDIA, Cisco and
Lenovo on network spine and leaf architecture design to maximize cluster efficiency and maintain low latency interconnects capable of
supporting distributed training across thousands of GPUs.

Secure external access into the GPU clusters is handled by our orchestration and customer portal. Customers are identified by their chosen enterprise single sign-on (SSO) offering, and strict network isolation, multi
tenancy controls and encrypted communications ensure connectivity is seamlessly secure.

Software
& Orchestration Layer

Our
orchestration and automation platform allow customers to deploy GPU compute autonomously and at scale. It provides us with the
foundation for repeatable onboarding of enterprise customers, offering scaling up or down with guaranteed, non-contended resources and
migration of workloads from simulation into production.

Proprietary
inference engine and orchestration platform allow customers to consume AI capabilities via tokens or models, abstracted from the underlying
GPU technology. This design enables:

●Interoperability
across different GPU types and design architectures (i.e. NVIDIA, AMD)

●Ease
of integration or migration of customer workloads from on-prem or cloud offerings into Sharon
AI

●Extension
of asset useful life through inference optimization, enabling older generation GPUs to remain
economically viable for inference workloads even as newer chips are used to optimize training
workloads.

●Customer
focus on AI outcomes (tokens, models, inference calls) rather than infrastructure management

-15-

Customer
Access & Services Layer

Multiple
customer access models including bare metal GPU servers, virtual servers, containers, and managed Kubernetes, as well as a host of preconfigured
applications, environments, blueprints and tools. Customers may choose from either a simple “one-click deploy” interface
or opt to use APIs to completely automate and rapidly scale up their utilisation easily.

Combined
with expert AI/ML/HPC operational support and technical expertise through Sharon AI Studio platform, a simple to use and evolving set
of tools means customers can take their AI journey from early research and development through to mature production workloads across
any inference or training requirement.

Integrated
into the platform and secured by enterprise SSO is a documentation portal and service desk system, allowing customers to consult best-practice
guides or reach out directly to our technical support staff for further assistance.

Funding
for Compute Infrastructure

Our growth and investment in future
GPU hardware acquisition and infrastructure development requires substantial capital. To fund this expansion strategy necessitates significant
capital, the Company is pursuing a diversified financing strategy aimed at aligning hardware procurement with confirmed customer demand.

We intend to utilize a range of funding sources, including:

●
Customer prepayments;

●
Revenue share agreements;

●
Traditional debt facilities;

●
Equipment financing; and

●
Proceeds raised from the sale of equity and/or debt securities.

Our capital deployment model is designed to mitigate upfront
cash requirements by utilizing customer commitments to partially fund hardware procurement where possible. The objective of this approach
is to enhance capital efficiency and shorten the payback period for individual deployments.

Data
Center Partnerships

We have agreements with data center
operators to host our AI cloud platform. Our current operational footprint is in high-density, Tier III and IV facilities with
NEXTDC, DigiCo and GreenSquareDC., which support sovereign AI requirements and next-generation hardware deployments, and which
offers industry-leading uptime, power density and interconnection.

This model is designed to enable us to focus on scaling core neocloud
services rapidly without the capital intensity and long lead times which are typically associated with building large-scale data centers.

The data center agreements are long dated in
nature and ensure that we have the energy and data center capacity to materially expand its GPU fleet and cloud offering to its customers.

NEXTDC

NEXTDC
is Australia’s leading independent data center operator, listed on the ASX with a market capitalization of approximately A$8
billion (at 18 January 2026). NEXTDC is an operator of Tier III and Tier IV data centers across the Asia-Pacific region, providing
high-density infrastructure engineered for resilience, security, interconnection and 100% uptime across major cities. Its
NVIDIA-certified ‘AI Factory’ facilities make NEXTDC a foundational enabler for sovereign AI compute, offering the
high-power, liquid-cooled, ultra-low-latency environments required by advanced GPU clouds to deploy Superclusters and serve
mission-critical enterprise and government AI workloads.

NEXTDC is our largest, non-exclusive co-location provider with an expansion
agreement in place for up to 40MW at its M2 facility in Melbourne, and 10.75MW at its S6 facility in Sydney. NEXTDC’s M3 facility
currently hosts Sour NVIDIA B200, H200 and L40s GPU’s, and combined with the M2, S3 and S6 capacity is also engineered to accommodate
the thermal and power loads of our future expected GPU deployments with capacity for 20,000+ NVIDIA B200/B300/GB300 GPUs.

-16-

We have agreed to lease up to 54MW of capacity from NEXTDC, comprised of the following:

-Current
Operations: 1.6MW already operational and ready for deployment at NEXTDC M3 Melbourne, hosting
the SharonAI Supercluster with more than 500 NVIDIA B200 GPUs;

-Approximately
13MWs of Distributed Capacity: Additional distributed capacity across Sydney and Melbourne
NEXTDC facilities (S6, S3) for a mix of B200 and B300 GPU deployments, with flexibility for
precinct deployments and single cluster customers; and

-40MWs
of Contiguous Capacity: Contiguous high-density deployment at NEXTDC’s M2 Melbourne
facility, designed for NVIDIA B300 and GB300 reference architecture with liquid to chip cooling,
targeting 4 x 10MW halls capable of supporting latest generation GPU deployments.

Other
Data Center Operators

DigiCo

DigiCo Infrastructure REIT (ASX: DGT) is a diversified
owner, operator, and developer of data center infrastructure, listed on the Australian Securities Exchange and externally managed by HMC
Capital. The REIT maintains a portfolio of 13 data centers across key Australian markets — including Sydney, Brisbane, Adelaide,
and Townsville — as well as North American locations in Dallas, Kansas City, Chicago, and Los Angeles.

GreenSquareDC

GreenSquareDC (GSDC) is an Australian owner, developer, and operator of large-scale sustainable
data centers, focused on delivering high-density, energy-efficient infrastructure optimized for AI, cloud, and hyperscale workloads. The
company is backed by Partners Group, with GreenSquareDC’s flagship facility is the SYD1 campus (also known as SYDGPU1), a brownfield
redevelopment located in Sydney’s Norwest Business Park, designed to deliver up to 110MW of capacity upon full build-out. The facility
supports high-density, liquid-cooled configurations.

Revenue
Model

Target
Customer Contract Framework

We generate revenue from fees paid by customers for access to its
AI and HPC platform. Pricing is determined by the hardware resources allocated, including GPUs, CPU capacity, high-performance storage
and interconnect bandwidth, as well as the customer’s selected consumption model. The Company offers both usage-based, on-demand
services and term-based ‘take-or-pay’ contracts, allowing customers to procure compute through either elastic consumption
or contracted capacity commitments. This structure reflects the operational characteristics of the SharonAI Cloud, a platform that provides
self-provisioned bare-metal GPU servers, virtualized compute environments, and containerized workloads with non-contended performance
and pre-configured AI frameworks.

Historically, revenue has been derived largely from usage-based, on-demand
workloads, including customers accessed through marketplace aggregators, as the Company optimized and performance validated its infrastructure.
During this phase, We deployed earlier-generation NVIDIA GPUs to demonstrate performance consistency under NVIDIA reference architectures
as it sought admission to the NCP partner program. These early deployments enabled us to validate cluster stability, software orchestration,
and workload isolation while building a track record with enterprise AI and research users.

As
we prepares for scaled commercial operations, we have shifted its primary commercial contracting framework toward term-based
offtake agreements on a ‘take or pay’ basis, which aligns customer commitments directly with the capital required to procure
and deploy GPU capacity. This structure supports predictable, multi-year revenue and aligns infrastructure investment with visible demand.
It also enables SharonAI to expand its GPU fleet, including next-generation NVIDIA H100 and planned H200, B200, B300 GPUs, as well as
large-scale deployments such as its 1,016-GPU Supercluster at NEXTDC, while managing working capital requirements.

A
key advantage of this term contract model is it leverages customer prepayments where available, to de-risk our
procurement strategy. In an environment where high-density GPUs are supply-constrained, contracted customer offtake agreements provide
the financial assurance necessary to secure GPUs at scale and to support underwriting of significant debt facilities, with the ‘take
or pay’ structure providing minimum levels of revenues. For customers, term agreements ensure guaranteed access to scarce GPU capacity,
dedicated infrastructure without performance degradation, sovereign hosting within Tier III or IV Australian data centers and accelerated
deployment timelines compared to self-deployment. These capabilities are increasingly critical for organizations training LLMs, executing
inference at scale or operating regulated workloads.

The
typical contract lifecycle is described:

Illustrative
Customer Contract Lifecyle

-17-

Contract
Signing and Prepayment (Month 1): Upon execution of a term contract, the customer provides an initial prepayment or deposit. This financial
commitment allows us to submit a purchase order for the specific GPU and network infrastructure required to service the contract.
We target 10% to 20% of contract value in the form of a customer prepayment;

Procurement
and Installation (Months 1-4): The hardware is typically delivered within approximately two to three months of the contract signing.
Final payment for the equipment is made upon delivery, and the infrastructure is installed and configured over the subsequent two-to-four
weeks;

Operational
Phase and Revenue Recognition (Months 5-60): Once the infrastructure is operational (typically beginning in the fifth month), the Company
generates monthly revenue for the duration of the contract term (e.g., 36 to 60 months). The customer makes corresponding monthly payments
for the reserved capacity with the ‘take or pay’ provision ensuring 100% utilization rate recognition once fully deployed;

Post
Contract Lifecycle (Years 3-6): The useful economic life of the GPU and network infrastructure is expected to extend for two or more
years beyond the initial contract term to approximately 6 years. Upon the conclusion of the initial contract, the Company has multiple
options to continue monetizing the fully-paid hardware including:

○Re-contracting
to same or new customer under subsequent multi-year agreement at potentially lower rates
to reflect the age of the depreciated asset;

○Capacity
sold on spot or on-demand market through GPU aggregator platforms, capturing higher hourly
rates for variable demand customers; or Older
generation GPUs redeployed through our inference engine for token-based consumption.

Our
software layer is a key differentiator with respect to extending the useful
life of our GPU fleet post this initial contract lifecycle, with our inference engine enables token-based consumption abstracted from
underlying GPU hardware (i.e. interoperability), supporting asset redeployment as technology evolves.

Strategic
Partnerships

We have partnered with global leaders in AI and digital infrastructure, including NVIDIA, NEXTDC, Cisco, WWT, Lenovo, VAST and Megaport to
ensure successful best practice and on time delivery of SharonAI’s products and services.

-18-

Strategic
Delivery Partners

Partner

Nature
of the Partnership

-

-

NVIDIA
is a manufacturer of the Company’s GPUs and is a key operational partner under NVIDIA’s NCP program.

NVIDIA is a strategic shareholder
in SharonAI.

-
NEXTDC serves as the Company’s
primary co-location data center provider, hosting the our hardware infrastructure.

-

-

Cisco
provides the Company with AI-ready networking infrastructure, including access to Cisco Nexus HyperFabric AI and NVIDIA Spectrum-X-aligned
architectures.

Digital Alpha, a digital
infrastructure investment firm with an exclusive technology partnership with Cisco is our strategic shareholder.

-

WWT is a global technology solutions provider and a primary architect of the physical and networking infrastructure required for generative AI.

-
Lenovo provides access to hardware
procurement and lifecycle services as part of the TruScale program. This is through a financing facility to fund GPU hardware acquisition.

-
VAST provides us with access
to its InsightEngine and underlying VAST AI Operating System.

-
Megaport provides us with
access to its network, including over 1,000 data centers across 26 countries, to build secure private links between our workloads
and their own locations.

NVIDIA

NVIDIA is the world leader in accelerated computing, having pioneered the
GPU and built a full-stack computing platform that spans chips, systems, software, and services. Listed on the NASDAQ with a market capitalization
of US$4.5 trillion (as of January 18 2026), its technologies power the global AI ecosystem that underpin the world’s AI factories.

For us, NVIDIA is a manufacturer of its GPUs, key operational and
industry partner and also a strategic referral source, which reflecting the ongoing collaboration between the two companies.

GPU
hardware

We are deploying several of NVIDIA’s GPUs (L40, A40, H100, H200, B200, B300, GB300) which are purpose-built for AI and HPC applications.
We also deploys NVIDIA’s AI Enterprise software library of LLM’s, tools, and resources, which is designed to streamline
the development and deployment of generative AI applications. Our proprietary orchestration platform complements this by
automating the management of compute workloads and storage across the Company’s infrastructure.

The Company as a NVIDIA Cloud Partner

In December 2024, we were appointed as an NCP. We are one of
three NCPs currently operating in Australia, part of a global network of partners. The NCP program comprises a
network of partners authorized to offer certified hosted hardware and software solutions utilizing NVIDIA products in a cloud or managed
services model. To maintain NCP status, We and other neoclouds are required to demonstrate the capability to build to NVIDIA’s
reference designs and deploy proprietary architectures benchmarked for performance.

-19-

NVIDIA
manages the total number of NCPs globally to ensure quality control and adherence to reference architecture standards. This rigorous
certification process creates a barrier to entry for potential competitors. This status was granted following NVIDIA’s assessment
of our position as a specialized provider of GPU compute in the Australian public cloud environment.

The
NCP designation provides us with three key operational advantages:

●
Accessibility: we receive access to NVIDIA’s product roadmap, facilitating
the evaluation and deployment of emerging GPU technologies;

●
Tailored Network Infrastructure: our engineering team collaborates with
NVIDIA to align its proprietary orchestration software and physical infrastructure with NVIDIA’s technical specifications; and

●
Validation (Implicit and Explicit): The NCP designation indicates to enterprise
customers that our platform performance meets NVIDIA’s standards for performance, reliability, and security in deploying AI workloads.

This
translates into specific operational benefits, including:

●
Forecasting Integration: Visibility into NVIDIA’s 12-to-18-month rolling supply chain forecast, facilitating improved infrastructure planning and allocation decisions;

●
Preferential GPU Access: Facilitated access to new GPU generations, including the B200 and B300 processors based on NVIDIA’s Blackwell architecture, which is optimized for AI workloads;

●
Customer Referrals: Access to the NVIDIA consumption desk, a mechanism through which NVIDIA refers qualified customers to partners with available capacity;

●
Technical Support: Ongoing support to ensure deployment aligns with NVIDIA’s reference architecture, including specific requirements regarding file systems, storage, networking, and cooling solutions; and

●
OEM Relationships: Facilitated engagement with OEMs such as Cisco, Lenovo,
Dell, and Super Micro to support server procurement and integration. Leveraging its arrangements with NVIDIA and NEXTDC, We have
commenced the deployment of a GPU cluster connected via InfiniBand and based on the NVIDIA Reference Architecture. This cluster is designed
to support HPC and AI workloads.

NEXTDC

NEXTDC
is a leading operator of Tier III and Tier IV data centers across the Asia-Pacific, providing high-density infrastructure engineered
for resilience, security, interconnection and 100% uptime across major cities. Its NVIDIA-certified ‘AI Factory’ facilities
make NEXTDC a foundational enabler for sovereign AI compute, offering the high-power, liquid-cooled, ultra-low-latency environments required
by advanced GPU clouds to deploy Superclusters and serve mission-critical enterprise and government AI workloads.

NEXTDC
is our primary, non-exclusive host with an agreement of up to 52MW at sites
located across Melbourne and Sydney. Under this arrangement, NEXTDC hosts our hardware infrastructure and provides critical operational
support, including network connectivity, cooling systems, and access to electricity at wholesale commercial rates.

We consider access to large-scale data center capacity to be a key strategic competitive advantage. We have secured access
to up to 54MW of capacity within NEXTDC’s facilities. This capacity is intended to be deployed as follows:

M2
(40MW): A single contiguous load deployment consisting of 4 x 10MW halls utilizing the NVIDIA B300 and GB300 reference
architecture;

M3
(2MW): Supercluster of 125 servers, comprising a total of 1,000 NVIDIA B200 GPUs;

S6
(10.75MW): A planned distributed deployment of NVIDIA B300 and potentially NVIDIA GB300 clusters; and

S3
(2MW): A planned single NVIDIA B300 cluster.

-20-

The
NEXTDC partnership provides us with access to:

Tier IV Data Center Infrastructure: World class facilities with industry
leading Power Usage Effectiveness (‘PUE’) of 1.10 to 1.5, meaning that approximately 10% to 50% of power is used for cooling/auxiliary,
whilst approximately 72% to 90% of power goes directly to compute, lowering operating costs and providing us with a strong competitive
advantage as its pays significantly less for electricity, which is typically one of the largest operating expenses for a neocloud provider;

DTA
Certified Strategic Sites: Australian Federal Government Digital Transformation Agency certification for sovereign government workloads;

Global
Connectivity: Low latency subsea cable network providing connectivity to Singapore (81ms), Hong Kong (130ms), US West (165ms), and Europe
(265ms); and

Proven
Hyperscale Support: Track record supporting major cloud providers with mission critical infrastructure.

The Company is currently in discussions with NEXTDC regarding potential
future partnerships, including the option to participate in NEXTDC’s publicly disclosed development pipeline which targets 3GW+
of capacity by 2028 (NEXTDC Limited, FY25 Full Year Results Presentation (ASX Announcement, 28 August 2025), page 34 https://www.nextdc.com/investor-center/).
NEXTDC is an important relationship for the Company, and the Company has additionally established strong relationships with other data
center operators.

Cisco

Cisco is a global technology leader that supplies networking hardware, software,
telecoms equipment, and security and collaboration solutions to enterprises, service providers and data center environments. Listed on
the NASDAQ with a market capitalizations of US$297 billion (as at 18 January 2026), its unified architecture for AI makes Cisco a foundational
infrastructure partner for AI providers as it delivers the high-bandwidth, low-latency, and resilient interconnect fabric required to
support modern GPU clusters and AI workloads.

We are a strategic partner of Cisco. Digital Alpha, a digital infrastructure investment firm focused on building premium digital
platforms through an exclusive technology partnership with Cisco, also became one of our investors in
December 2025/January 2026.

The
objective of this three-way partnership is to combine our AI capabilities with Cisco’s market reach to generate
revenue opportunities across the enterprise and government sectors.

Under
the terms of the partnership, the parties have agreed to collaborate on the following initiatives:

Technical
Guidance: Cisco will provide insights and guidance on architecting and deploying enterprise-grade cloud and AI solutions.

Pipeline
Targets: We intend to build a sales pipeline leveraging Cisco’s products and technologies.

Market
Access: Cisco will support our go-to-market strategy, facilitating engagement with large-scale enterprise and government
customers.

In addition to the collaboration, we are a member of the Cisco 360
Partner Program. This designation authorizes us to build, sell, and manage integrated Cisco solutions.

The
program includes a performance-based incentive structure. Subject to meeting specific eligibility criteria, we are entitled to
receive a fixed percentage rebate calculated on the revenue generated from the sale of eligible Cisco solutions. This structure is designed
to align the Company’s incentives with the adoption of Cisco’s network technologies.

-21-

WWT

WWT
is a global technology solutions provider that offers consulting, supply chain and IT services in areas like AI, Security and Data in
over 60 locations globally and is one of the largest private companies in the US. WWT is a primary architect of the physical and networking
infrastructure required for generative AI and helps clients and partners conceptualize, test and validate innovative technology solutions
for business outcomes and then deploys them at scale through its global warehousing, distribution and integration capabilities.

In
January 2026, we signed an agreement for WWT to be an important delivery partner for execution of its AI Cloud Services.

Lenovo

Lenovo is a Hong-Kong-based multinational corporation that designs, builds,
and delivers a broad portfolio of hardware, technology and infrastructure solutions across client, edge, cloud, network and intelligence
segments. Listed on the HKSE at a market cap of HK$100 billion (as at January 18 2026), its expanding Infrastructure Solutions Group provides
GPU-dense servers and data-center platforms such as Lenovo TruScale.

Lenovo has been a primary supplier of compute and storage equipment for us via
our subsidiary Distributed Storage Solutions since 2021. We are also a participant of the Lenovo TruScale program, a strategic engagement
that provides access to hardware procurement and lifecycle services.

Through
the TruScale program, the Company has established a financing facility to fund GPU hardware acquisitions. This facility is structured
to reduce upfront capital expenditure and improve working capital efficiency by matching payment obligations with the asset’s useful
life.

Utilizing
this facility, we have executed an agreement to acquire 125 servers, comprising a total of 1,000 NVIDIA B200 GPUs. This hardware
is slated for the Company’s next cluster deployment in NEXTDC’s M3 facility in the first half of 2026.

We consider this to be a key enabler of our growth strategy, allowing compute equipment to scale rapidly
in alignment with the availability of data center capacity and customer demand.

VAST

VAST
is a New York headquartered, privately held technology company that provides an AI-focused data infrastructure platform. Built around
its AI Operating System, VAST consolidates high-performance storage, data management, and compute orchestration into a single coherent
architecture to meet the demands of large-scale AI and deep-learning workloads.

We have established a relationship with VAST to integrate its AI Operating System to serve enterprise and government customers with inference
at any scale. Our technology will be able to leverage VAST’s InsightEngine, an end-to-end ingestion, embedding, indexing
and retrieval system that lets organizations continuously ingest structured, unstructured and streaming data in real time, feeding inference
systems by delivering low-latency search for workflows at scale.

This
collaboration will allow our customers to ingest, process and analyze massive volumes of throughput with low-latency inference to
assist organizations in moving from AI experimentation to production with repeatable, enterprise-grade workflows.

Megaport

Megaport is a leading Australian global Network-as-a-Service (‘NaaS’)
provider listed on the ASX with a market capitalization of approximately A$2.2 billion (as at January 18 2026). Megaport’s NaaS
underpins the connectivity layer of the AI ecosystem by enabling fast, private and scalable links between data centers, GPU cloud providers
and major hyperscale clouds. Through its software-defined networking platform, Megaport offers on-demand, high-performance connectivity
across data centers globally, supporting low-latency AI workloads and multi-cloud architectures for enterprises. Its infrastructure plays
a critical role as AI deployments increasingly require rapid data movement between GPUaaS platforms, cloud storage, and distributed compute
environments.

-22-

In
October 2025, we agreed to a collaboration with Megaport under a reseller arrangement to enhance its go-to-market strategy. This
arrangement enables Megaport’s enterprise, government and research customers in over 1,000 data centers across 26 countries to
privately connect to the SharonAI Cloud.

This
collaboration allows our customers to utilize connectivity options including
Amazon’s AWS Direct Connect, Microsoft’s Azure ExpressRoute and Australia’s own AARNet to reduce costs and increase
reliability to enterprise, academic and AI workloads. Customers can also directly utilize Megaport’s expansive network to build
secure private links between our workloads and their own locations, which can be used for expanding global connectivity or ensuring domestic
sovereignty and data security for compliance and legal requirements.

Growth
Strategy

Dependencies
Underpinning Demand for Our Products and Services

The
Company’s success will be determined by four critical dependencies that define the AI/HPC infrastructure ecosystem which will generate
strong demand for our solutions into the future:

Pillar
1 – Demand for GPU computing;

Pillar
2 – Access to the supply of chips;

Pillar
3 – Access to data center capacity and the supply of power; and

Pillar
4 – Our proprietary cloud platform.

These
pillars position us to capture substantial value from the exponential growth in AI computing workloads globally.

Key Pillars Underpinning Our Growth

Feature

Detail

Pillar
1: Demand for GPU Computing

-
The
demand for GPU compute capacity is experiencing a significant structural surge. This rapid acceleration is driven by the adoption
of Generative AI, large language models, and advanced machine learning being increasingly integrated into workloads across a wide
range of customer segments, including enterprise and government.

-
NVIDIA
noted in their Q3 November 2025 result that its GPU installed base of new and previous generations are fully utilized as demand for
AI infrastructure continues to exceed expectations, with compute growing 56% year-on-year. Supply of GPUs globally is currently backlogged,
as NVIDIA has announced its cloud GPUs, including the Blackwell series, are sold out, with massive orders booked through 2026.

-
As demand for GPU compute continues to accelerate across training and inference to outstrip supply, neoclouds are
meeting an unmet need by providing unmet GPU-as-a-Service. Neocloud revenue is expected to grow at a compound annual growth rate (CAGR)
of 23% over the next six years from 2024 to 2030.

Pillar
2: Supply of Chips

-
We are a NVIDIA Cloud Partner, which provides it with preferential access to NVIDIA’s GPUs. This includes new
generations of chips such as the B200 and B300 processors based on NVIDIA’s Blackwell architecture, which are optimized for AI workloads.

-
NVIDIA
manages the total number of NCPs globally to ensure quality control and adherence to reference architecture standards in GPU construction.
The rigorous certification process required to be inducted as a NCP restricts the supply of neocloud distribution channels and provides
us with the benefit of barriers to entry.

-23-

Pillar
3: Data Center Capacity & Access to Power

-
Power capacity through its data center
partnerships with top-tier facilities in Australia, which are critical to neocloud deployment especially for sovereign AI purposes. Currently,
we partner with NEXTDC, GreenSquare and DigiCo to host its GPU Supercluster and GPU Fleets.

-
Notably, we have
entered into a strategic engagement with NEXTDC, Australia’s leading independent data center operator, to provide access to
up to 54MW of capacity. This arrangement is designed to support our expansion pipeline through 2026.

-
We retain the flexibility to enter into agreements with other data center operators. We are currently in negotiations
with multiple parties to secure additional sources of power and co-location capacity to support future growth.

Pillar
4: SharonAI’s Proprietary Platform

-
SharonAI’s AI Cloud
platform for orchestration and automation is engineered to extract higher efficiency and performance from advanced GPU infrastructure
than traditional cloud environments. The platform integrates dedicated, non-contended GPU orchestration, high-throughput data pipelines,
and pre-configured AI frameworks to minimize idle time and compress deployment cycles. This allows deterministic performance so that
customers can achieve more output per GPU and materially improve their total cost of computation.

-
This proprietary approach creates a structural advantage that is difficult for competitors to replicate as it originates
from our proprietary, integrated software stack and ecosystem alignment with key operating partners including NVIDIA, NEXTDC, Cisco, WWT,
Lenovo, VAST and Megaport.

-
This high performance platform which is more accessible for broader customer segments, such as enterprise and government
organizations, will allow SharonAI to capture market share and value from the growth in computing workloads globally.

Approach
to Capture Demand for Our Products and Services

Our growth strategy focuses on scaling its specialized HPC infrastructure to address the increasing demand for AI capabilities.
The Company’s objective is to establish itself as a leading HPC and AI infrastructure providers by executing a strategy built on
four key pillars:

1.Rapid Capacity Expansion — Rapidly scale specialized AI and high-performance computing infrastructure by deploying
advanced GPU clusters in strategic data center locations across Australia and Asia Pacific.;

2.Securing Customer Demand — Establish long-term contracted revenue streams by partnering with enterprise, government,
and hyperscale clients seeking sovereign and low-latency AI compute solutions across the region.;

3.Capital-Efficient Deployment — Maximize return on invested capital by leveraging phased build-out strategies, and strategic co-location
partnerships to reduce upfront expenditure and accelerate time to revenue.; and

4.Technological Leadership — Maintain a competitive edge by integrating next-generation GPU architectures, advanced
liquid cooling systems, and energy-efficient infrastructure to deliver best-in-class AI compute performance at scale.

We are undertaking a significant expansion of its compute capacity utilizing funds obtained in recent financings. This phase involves
the development of a specialized AI supercluster utilizing advanced GPU architectures.

-24-

We intend to continue deploying next-generation
chip clusters to meet evolving customer demand, with a focus on capturing high-value AI workloads.

The
strategy is designed to sustain a competitive advantage through strategic partnerships and operational efficiency:

●
NCP Status: Our certified NCP status provides access to NVIDIA’s latest technologies, technical support, and go-to-market collaboration. In a supply-constrained environment, this relationship facilitates access to the hardware required to meet customer demand;

●
Infrastructure Efficiency: Relationships with leading data center operators provide access to facilities with favorable power and cooling cost structures. This contributes to a cost-efficient operating model; and

●
Continuous Innovation: We intend to maintain market leadership by continuously evaluating and deploying the next generation of AI accelerators, ensuring customers have access to high-performance infrastructure.

Intellectual
Property

We
rely on trademark and trade secret laws, as well as employee and third-party non-disclosure, confidentiality and other types of contractual
arrangements to establish, maintain and enforce our intellectual property rights, including with respect to our proprietary rights related
to our products.

As
of the date of this report, we have the following trademark:

Trademark

Country

Date
of registration

Registration
No.

SharonAI

United
States of America

September
16, 2025

7943687

SharonAI

Australia

April 3,
2025

2481068

We
believe that the trademarks that we use in our business are important for building our brand image and brand recognition. Therefore,
we will develop marketing strategies, including advertising and branding campaigns, accordingly.

Competition

The
HPC/AI cloud industry is dynamic and global. Many of the industry participants are larger operators of facilities, with access to both
large energy infrastructure and supply of the appropriate compute and storage equipment required to operate cloud platforms at scale.
Some of the larger operators of cloud platforms are also developers or owners of data centers besides specialized data center operators.

We
will compete with them directly for the acquisition of new compute and storage equipment, access to energy infrastructure and raising
capital. Digital infrastructure providers, including us, also compete with more traditional industries, for example, when obtaining the
lowest cost of electricity, or access to sites with reliable sources of power. Many digital infrastructure operators are not publicly
operated, and therefore data is not readily available.

Public
reporting companies operating cloud platforms and/or data centers used by cloud platforms include:

●
Bit
Digital, Inc.

●
Hive
Digital Technologies Ltd.

●
IREN
Limited

●
Applied
Digital Corporation

●
Core
Scientific Inc

●
Hut
8 Corp.

●
BitDeer
Technologies Group

●
DigitalOcean
Holdings, Inc.

●
Nebius
Group N.V.

-25-

●
Amazon.com
Inc.

●
Alphabet
Inc.

●
Microsoft
Corp.

●
Coreweave,
Inc.

Government
Regulations and Environment

Our
business is subject to regulation by various federal, state, local, and foreign governmental agencies, including agencies responsible
for monitoring and enforcing employment and labor laws, workplace safety and environmental laws, including those related to energy usage
and energy efficiency requirements, privacy and data protection laws, AI, financial services laws, anti-bribery laws, sanctions, national
security, import and export controls, anti-boycott, federal securities laws, and tax laws and regulations.

For
example, governmental authorities have in the past sought to restrict data center development based on environmental considerations and
have imposed moratoria on data center development, citing concerns about energy usage, requiring new data centers to meet energy efficiency
requirements. We may face higher costs from any laws requiring enhanced energy efficiency measures, changes to cooling systems, caps
on energy usage, land use restrictions, limitations on back-up power sources, or other environmental requirements.

In
certain foreign jurisdictions, these regulatory requirements may be more stringent than those in the United States. These laws and regulations
are subject to change over time and thus we must continue to monitor and dedicate resources to ensure continued compliance. In particular,
the global AI regulatory environment continues to evolve as regulators and lawmakers have started proposing and adopting, or are currently
considering, regulations and guidance specifically on the use of AI. Non-compliance with applicable regulations or requirements could
subject us to investigations, sanctions, mandatory product recalls, enforcement actions, disgorgement of profits, fines, damages, civil
and criminal penalties, or injunctions and jail time for responsible employees and managers. If any governmental sanctions are imposed,
or if we do not prevail in any possible civil or criminal litigation, our business, operating results, financial condition, and future
prospects could be materially adversely affected. In addition, responding to any action will likely result in a significant diversion
of management’s attention and resources and an increase in professional fees. Enforcement actions and sanctions could harm our
business, operating results, financial condition, and future prospects.

Our
sustainability initiatives, goals, or commitments could be difficult to achieve or costly to implement. Moreover, compliance with recently
adopted and potential upcoming ESG requirements, including California legislation that requires various climate-related disclosures,
the European Union’s Corporate Sustainability Reporting Directive and Corporate Sustainability Due Diligence Directive, and the
United Kingdom’s Streamlined Energy and Carbon Reporting framework will require the dedication of significant time and resources.
In addition, we may also be required to comply with the SEC’s comprehensive climate change disclosure rules, which have been stayed
pending judicial review. Additionally, if our competitors’ corporate social responsibility performance is perceived to be better
than ours, potential, or current investors may elect to invest with our competitors instead. Our business may face increased scrutiny
related to these activities and our related disclosures, including from the investment community, and our failure to achieve progress
or manage the dynamic public sentiment and legal landscape in these areas on a timely basis, or at all, could adversely affect our reputation,
business, and financial performance.

Employees

Our
employees are critical to our success. As of March 18, 2026, we had 25 employees, board members, advisors and contractors based in Australia
and the United States. We further rely on the extensive expertise of our external advisers, including legal, audit, financial and compliance
consultants, who may be engaged on an hourly basis, or on a project basis.

The
table below breaks down our full-time personnel by function as of March 18, 2026:

Function
Number of

Employees
% of

Total

Executive
6
20%

General Operations
24
80%

Total
30
100%

-26-

Corporate
Information

We
were incorporated in the state of Delaware on February 15, 2024 under the name SharonAI, and now includes the businesses of
DSS, which is 99% owned and AAM (now SharonAI Pty Ltd) which is 100% owned, which date back to 2021. On December 17, 2025, we
completed the Business Combination with Roth CH Acquisition Co, in accordance with the terms of the Business Combination Agreement
dated January 28, 2025, as amended, by and among us, Roth CH Acquisition Co., and Roth CH Holdings, Inc., Roth CH Merger Sub, Inc.
pursuant to which Roth CH Acquisition Co. merged with and into The Company with the Company as the surviving corporation (the
“Domestication Merger”) and Merger Sub has merged with and into SharonAI with the SharonAI as the surviving corporation (the
“Acquisition Merger” and collectively with the Domestication Merger the “Business Combination”).

Our
principal executive offices are located at 745 Fifth Avenue, Suite 500 New York, NY 10151. Our contact email is info@sharonai.com, and
our website is www.sharonai.com. The information contained on our website is not included in, nor incorporated by reference into, this
Annual Report on Form 10-K, and our website address is included in this document as an inactive textual reference only.

Available
Information

Our
website address is www.sharonai.com. The contents of, or information accessible through, our website is not part of this Annual
Report on Form 10-K, and our website address is included in this document as an inactive textual reference only. We make our filings
with the U.S. Securities and Exchange Commission (“SEC”), including our Annual Report on Form 10-K, Quarterly Reports on
Form 10-Q, Current Reports on Form 8-K and all amendments to those reports, available free of charge on our website as soon as reasonably
practicable after we file such reports with, or furnish such reports to, the SEC. The public may read and copy the materials we file
with the SEC at the SEC’s Public Reference Room at 100 F Street, NE, Washington, DC 20549. The public may obtain information on
the operation of the Public Reference Room by calling the SEC at 1-800-SEC-0330. Additionally, the SEC maintains an internet site that
contains reports, proxy and information statements and other information. The address of the SEC’s website is www.sec.gov.
The information contained in the SEC’s website is not intended to be a part of this filing.