What SCX.ai does

SCX.ai Holdings Limited is an Australian private company founded in 2025 that operates a sovereign AI infrastructure platform built on SambaNova SN40L ASIC chips purpose-built for transformer inference. The company deploys AI inference capacity onshore at Australian data centres — initially the Equinix SY5 facility in Sydney — and delivers it via OpenAI and Anthropic-compatible APIs, enabling drop-in migration of existing AI applications. Its product portfolio includes a proprietary Australian-aligned large language model family (MAGPiE, available in 20B and 120B variants), a sovereign coding model (SCX-coder), managed vector storage for enterprise RAG, container deployment for custom models, and embedded AI enablement services. The platform supports leading open-weight and commercial foundation models including from DeepSeek, OpenAI (gpt-oss-120b), Meta (Llama), and Google (Gemma), and exposes a proprietary STU (Service Token Unit) billing abstraction across LLM, speech, and embedding workloads.

The company serves three primary customer segments: Australian government and public sector agencies requiring strict data sovereignty and IRAP compliance, regulated enterprises in financial services (APRA-aligned banks, insurers, superannuation funds) and healthcare (TGA-aligned medical technology), and AI-native software builders seeking API-compatible sovereign inference. Named customers include Australis Scientific (medical technology), a major Australian bank (compliance automation use case), and a Fortune 500 financial services company (enterprise RAG processing 10M+ documents at sub-100ms latency). Distribution combines direct enterprise field sales with self-serve developer access via platform.scx.ai and consumer access via chat.scx.ai, plus an international channel partnership with New Zealand consortium Kereru.ai.

SCX.ai generates revenue primarily through monthly subscriptions (Starter at A$500/month for 400 STU, Growth at A$4,000/month for 5,000 STU, Enterprise at A$9,000/month for 15,000 STU) with usage-based throttling rather than overage charges, supplemented by professional services (embedded AI engineering, fine-tuning via LoRA/PEFT, integration with LangChain/AutoGen/CRM/ERP) and dedicated infrastructure arrangements including BYOM hosting. Pricing differentiates from hyperscalers through STU-banded token pricing (1.00–8.00 STU per 1M tokens across light/standard/premium model bands) and a throttling-first philosophy. The company has raised approximately A$30 million in undisclosed funding for first-phase compute deployment, holds SOC 2 Type 2 and ISO 27001 certifications, and is expanding from NSW to all Australian states throughout 2026 with a parallel international push into New Zealand through Project Kererū.

Name

SCX.ai

Legal name

SCX.ai Holdings Limited

Website

https://scx.ai

Company type

Private

Founded year

2025

Operating status

Operating

Headcount range

11–50 employees

Short description

SCX.ai is an Australian sovereign AI infrastructure company founded in 2025 that delivers ASIC-accelerated AI inference via SambaNova SN40L chips at onshore data centres to Australian government, regulated enterprises in financial services and healthcare, and AI-native builders.

Ownership category

akta.pro rank

Product category

Sovereign AI Infrastructure

NAICS

Computing Infrastructure Providers, Data Processing, Web Hosting, and Related Services (5182), Computer Systems Design and Related Services (5415)

SIC

Services-Prepackaged Software (7372), Services-Computer Programming, Data Processing, Etc. (7370)

akta.pro primary industry

Model Hosting, Serving & Inference Platforms (HDAAACAB)

akta.pro secondary industries

Model Serving, Inference & Deployment Platforms (APIs, Edge/On-Prem) (HDAEANAC), Model Deployment, Serving & Inference Platforms (HDAAABAF), End-to-End Enterprise AI Platforms (MLOps & Model Lifecycle Management) (HDAEANAA), AI Compiler, Runtime & Kernel Optimization Software (CUDA/ROCm/XLA, graph compilers) (HDAAAAAI)

Keywords

• Sovereign AI infrastructure
• AI inference as a service
• Foundation model APIs
• Enterprise AI deployment
• Australian large language models

Headquarters

HQ city

Southern Cross

HQ country

Australia

HQ region

Oceania

Offices2 records

Markets served

GTM type

B2B

Offering type

Software

Cost components

Technology or R&D, Infrastructure, Personnel, Marketing or Sales, Operations

Revenue model

1. AI Inference as a Service (STU-based): SCX.ai sells AI inference using Service Token Units (STUs) — a unified billing unit across LLM tokens, speech-to-text, text-to-speech, and embeddings. Customers purchase monthly plans (Starter, Growth, Enterprise) with included STU pools. Usage beyond plan limits results in throttling rather than overage charges. Additional STU top-ups (100/1,000/3,000 sizes) and Reserved Throughput add-ons are available. Band-L (light models) costs 1.00 STU per 1M tokens, Band-S (standard/70B-class) costs 2.00 STU per 1M tokens, and Band-P (premium/MoE) costs 8.00 STU per 1M tokens. This creates a hybrid subscription + usage-based model with predictable unit economics.
2. Model Fine-tuning and Customisation: Enterprise plan customers have access to model fine-tuning services, enabling domain-adapted models with controlled rollouts and adaptive context engineering (ACE). LoRA and PEFT fine-tuning are offered as part of the platform's AI Enablement Services.
3. AI Enablement and Support Services: SCX.ai embeds AI engineers and architects alongside customer delivery teams for solution architecture, integration and agentic AI, and model optimisation. This professional services layer includes embedded AI engineering, solution architecture planning, integration with LangChain/AutoGen/CRM/ERP systems, and model fine-tuning.
4. Dedicated Infrastructure: Enterprise plan includes access to dedicated inference endpoints and 'Bring Your Own Model' hosting, providing full isolation and IP control for sensitive workloads.

Pricing tiers

Go-to-market motion3 records

Distribution channels4 records

Marketing channels8 records

Core offering

SCX.ai delivers sovereign AI inference infrastructure that lets Australian enterprises, government agencies, and AI builders run large language models and other AI workloads entirely onshore. The platform exposes OpenAI and Anthropic-compatible APIs over ASIC-accelerated SambaNova hardware, hosts proprietary Australian-aligned models (MAGPiE, SCX-coder) alongside leading open-weight models, and wraps this with managed vector stores, RAG, guardrails, function calling, and AI Enablement engineering services. Customers consume compute through monthly plans priced in Service Token Units (STUs), with optional dedicated endpoints and Bring-Your-Own-Model hosting.

Product overview

SCX.ai is a sovereign AI platform company offering a full-stack AI infrastructure service built on ASIC-accelerated technology through a partnership with SambaNova Systems. The core platform provides OpenAI and Anthropic-compatible APIs for deploying open-weight models (MiniMax-M2.7, DeepSeek-V3.1, GPT-OSS-120B) on sovereign Australian infrastructure. Key products include MAGPiE (Australian-aligned 117B MoE LLM), SCX-coder (coding assistant), Inference as a Service with STU-based billing, Container Deployment for custom models, Vector Storage for enterprise RAG, and AI Enablement & Support services. The platform emphasizes data sovereignty with no prompt caching, no training on customer data, and full onshore processing at Equinix SY5 in Sydney, with planned expansion across all Australian states.

Differentiator

Problem solved

Functional benefit

Brands

• Project MAGPiE: Australia's first culturally aligned large language model (LLM) that thinks like an Australian, fine-tuned on Australian data including legislation, regulations, case law, and cultural context. Available in Small (20B) and Medium (120B) versions.
• MAGPiE
• Project Kererū
• SCX Labs

Products and services

• SCX.ai Platform (Inference as a Service) Core sovereign AI inference platform exposing OpenAI and Anthropic-compatible APIs on ASIC-accelerated SambaNova SN40L infrastructure. Customers access open-weight models (MiniMax-M2.7, DeepSeek-V3.1, GPT-OSS-120B) and SCX.ai proprietary models on Australian onshore infrastructure with predictable STU-based pricing, no prompt caching, no training on customer data, and isolated inference.
• MAGPiE (Australian LLM) Australia's first culturally aligned 117B MoE large language model fine-tuned on Australian legislation, regulations, case law, and cultural corpora, available in MAGPiE-Small (20B) and MAGPiE-Medium (120B) variants. Supports 128k context, tools, JSON output, and reasoning, with sovereign data handling for Australian enterprise, government, and AI builder customers.
• SCX-coder Sovereign coding model purpose-built for Australian engineering teams, optimised for algorithms, debugging, code review, and structured engineering workflows. Runs entirely on Australian onshore infrastructure so source code never leaves the country.
• Container Deployment Allows enterprise customers to deploy their own custom models and applications in containers on SCX.ai's sovereign infrastructure with dedicated endpoints and enterprise-grade isolation, supporting Bring-Your-Own-Model hosting.
• Vector Storage Managed vector stores with built-in embedding models (E5-Mistral-7B-Instruct), configurable distance metrics, and zero infrastructure overhead. Powers enterprise RAG with tenant-isolated onshore knowledge vaults, document classification, PII scrubbing, and retention policies.
• AI Enablement & Support Services Embedded AI engineering and solution architecture services where SCX.ai specialists work alongside customer delivery teams, covering solution architecture, integration with LangChain, AutoGen, CRM, and ERP systems, RAG pipeline design, model fine-tuning (LoRA/PEFT), and production operations support.
• SCX Labs SCX Labs is the global R&D engine behind SCX.ai's sovereign AI factory, publishing open research, vendor-agnostic benchmarks (SouthernCross-AussieQA-1k), and environmentally responsible model adaptation work anchored by SCX ACE, Project MAGPiE, and the Australia 1k benchmark.

Quantifiable outcome

• Up to 45% reduction in compliance costs for major banks through automated monitoring, onshore processing, and AI-powered compliance management
• +10 more outcomes

Named customers3 records

Segments6 records

Ideal customer profiles6 records

Technology focussed Yes

API detail

Has API

Yes

API docs

API detail

Core technology

AI maturity

App detail

Integration2 records

AI capability11 records

Feature19 records

Partnerships

Seven partnerships are on record, tiered minor, core and flagship / core.

• Wordcab and Knowledgator (Opir)minorStrategic or Co-development Partner · 5 June 2026SCX.ai is proud sponsor of Opir, a new family of highly performant guardrail models from Wordcab and Knowledgator. SCX.ai helped build the 2M+ training dataset for Opir encoder guardrails, which match frontier-class safety at a fraction of the cost and latency — outperforming models 20× their size.
• Servers AustraliacoreImplementation/ SI/ Consulting Partner · 30 January 2026Servers Australia provides local technical support for SCX.ai's sovereign AI node at Equinix SY5 in Sydney, enabling Australian enterprises and government agencies to run advanced AI workloads onshore.
• Kereru.aiflagship / coreStrategic or Co-development Partner · 9 January 2026New Zealand consortium Kereru.ai partnered with SCX.ai to build sovereign AI infrastructure for New Zealand. The partnership combines SCX's AI factory architecture with Kererū.ai's local ownership and governance. The key initiative is Project Kererū, a sovereign large language model program focused on New Zealand data, institutional language, local terminology, and Te Reo Māori support. Production-grade AI services expected from Q1 2026.
• SambaNova Systemsflagship / coreTechnology or IntegrationSCX.ai partnered with SambaNova Systems to launch Australia's first ASIC-based sovereign AI cloud. SambaNova provides its SN40L chips — purpose-built for AI inference — delivering approximately 3× higher throughput with significant power savings versus GPU clouds. The partnership enables SCX.ai's sovereign AI factory with all models, data pipelines, and system logs processed exclusively within Australian jurisdiction. First node deployed at Equinix SY5 in Sydney in late 2025.
• EquinixcoreTechnology or IntegrationEquinix provides the SY5 data centre in Sydney where SCX.ai deployed Australia's first sovereign AI infrastructure node. The Equinix Fabric private interconnect enables sovereign inference on Australian soil. Infrastructure is deployed within existing data centre facilities eliminating the need for extensive new build-out.
• OpenAIcoreTechnology or IntegrationSCX.ai's platform is OpenAI API-compatible, enabling drop-in replacement in just two lines of code. Supports OpenAI's gpt-oss-120b open-weight MoE model alongside other leading commercial and open-source foundation models.
• AnthropiccoreTechnology or IntegrationSCX.ai's platform is Anthropic API-compatible alongside OpenAI compatibility, providing flexibility for customers migrating from either provider.

Scale indicators11 records

Recent moves9 records

Expansion highlights6 records

Direct peers

• Replicate: Cloud platform for running open-source and custom AI models via API, targeting developers integrating AI into applications. Direct peer in the developer/API-first inference market, with similar open-weight model hosting and pay-per-use billing model.
• Anyscale: AI compute platform built on Ray, offering managed inference and training infrastructure for production AI workloads. Comparable as a developer-focused AI infrastructure platform targeting enterprises deploying open-weight models at scale with managed runtime services.
• CoreWeave: GPU-cloud AI infrastructure provider offering high-throughput inference and training compute on dedicated hardware. Direct competitor to SCX.ai in the specialised AI infrastructure layer, with a similar focus on performance-per-dollar and developer-friendly APIs, though CoreWeave operates globally and is Nvidia-centric rather than ASIC-based.
• Fireworks AI: Inference platform for open-weight and fine-tuned models with focus on low latency and high throughput, offering OpenAI-compatible APIs. Direct competitor explicitly benchmarked by SCX.ai and operating in the same inference-optimisation category with similar function-calling and structured-output features.
• Lambda Labs: AI cloud provider offering GPU clusters for training and inference with on-demand and reserved pricing. Comparable to SCX.ai as a dedicated AI compute vendor competing against hyperscalers on price-performance and developer ergonomics for inference workloads.
• Together AI: AI inference cloud platform hosting open-weight foundation models (Llama, DeepSeek, Qwen) with OpenAI-compatible APIs. Explicitly named as a benchmark competitor to SCX.ai in function-calling accuracy comparisons, making it the most direct peer in the open-model inference-as-a-service category.

Broad incumbents

• Hugging Face: Platform for hosting, evaluating, and deploying open-weight models with inference endpoints and an enterprise tier. Comparable as an open-model hub and inference API provider, though broader in scope (model repository, community, datasets) and not focused on ASIC-accelerated sovereign infrastructure.
• Microsoft Azure AI Foundry: Hyperscaler AI platform offering model hosting, fine-tuning, RAG, and agent services with Australian data residency and protected government cloud. Direct competitor for regulated enterprise and government AI workloads, with bundled compute, identity, and integration advantages that SCX.ai must offset via sovereignty and cost.
• AWS Bedrock: Hyperscaler-managed service offering foundation models (Anthropic, Meta, Mistral, Cohere) with enterprise security and regional data residency. Competes with SCX.ai in the enterprise inference market, with strong Australian region presence and government accreditation (IRAP), though as part of a much broader cloud portfolio.

Emerging players

• Modal Labs: Serverless AI compute platform for running inference and batch workloads on GPUs with developer-friendly APIs. Comparable to SCX.ai's container deployment and inference-as-a-service offerings, particularly for AI-native builders, though focused on CPU/GPU rather than ASIC.

Market position

Strengths5 records

Weaknesses5 records

Competitive moat6 records

Key risks6 records

Key highlights7 records

Customer concentration

Compliance6 records

Revenue estimate

Valuation estimate

Number of profiles

Profiles2 records

Funding overview

Funding rounds

Investors

Funding detail is available on the Subscription and Enterprise plan.Contact sales →

M&A

Investments

M&A and investment is available on the Subscription and Enterprise plan.Contact sales →

Frequently asked questions about SCX.ai