Upstage

업스테이지 / 주식회사 업스테이지

Develops the Solar family of language models and document-intelligence software that helps Asian enterprises turn unstructured records into usable data for AI-driven workflows.

Private company
Profile as of September 14, 2026
#123 in the AsiaAI Tech Index
How the ranking works →

Upstage has a credible position in South Korea's enterprise AI market by combining proprietary Asian-language models with document-intelligence tools. The central tension is that it operates in a highly contestable market, remaining loss-making, privately disclosed, and dependent on external cloud ecosystems while facing far larger global competitors.

Key figures

Revenue FY2025: 24.8 billion Korean won
Revenue growth FY2025: 78.4% Estimate
Operating margin FY2025: Negative 123.0% Estimate
Net income FY2025: Net loss of 28.4 billion Korean won
R&D spend: Not disclosed
Capital expenditure: Not disclosed
Latest private valuation April 2026: More than 1 trillion Korean won
Total capital raised April 2026: Approximately 400 billion Korean won
Operating cash flow FY2025: Negative 34.1 billion Korean won
Cash holdings End of FY2025: 100.4 billion Korean won

Overview

Upstage is a privately held South Korean enterprise AI company founded in 2020. Its core business centers on the Solar family of language models and a document-intelligence software suite designed to convert unstructured business documents into structured data for downstream AI and workflow automation. The company does not manufacture semiconductors or operate its own AI infrastructure, relying instead on third-party cloud providers like Amazon Web Services for compute capacity. Its products target enterprises that require document automation, regulated-industry workflows, and language-model APIs. By focusing on the application and model layers, Upstage positions itself as a software provider that enables enterprise AI adoption rather than a foundational infrastructure operator. Its direct buyers include businesses integrating AI into their document ingestion and retrieval pipelines, and it distributes its tools through direct sales and major cloud marketplaces.

The AI angle

Upstage generates revenue through enterprise software consumption, API usage, cloud offerings, and deployment services rather than semiconductor sales. Its core products include the Solar family of proprietary language models, such as Solar Pro, and a document-intelligence suite featuring Document Parse, Information Extract, and a Universal Information Extractor agent. These tools extract text, document structure, and business data from PDFs, scans, forms, invoices, and emails. Upstage distributes its models through direct enterprise channels and cloud marketplaces, including Amazon Bedrock Marketplace and Oracle Cloud Infrastructure. Higher demand for AI-ready document ingestion, retrieval-augmented generation, and locally optimized Asian-language models directly increases usage of Upstage's APIs and enterprise subscriptions. The company does not disclose its revenue split between Solar, Document Intelligence, implementation services, or marketplace channels. Its infrastructure exposure depends heavily on third-party cloud and accelerator suppliers, utilizing AWS Trainium and Inferentia chips for training and inference workloads.

Technology and moat

Upstage's technical differentiation lies in its optimization for Korean and other Asian languages, combined with document-specific extraction tooling. Its Document Intelligence suite addresses an enterprise problem that general-purpose language models do not fully solve: converting heterogeneous business documents into clean, structured information. The technical challenge includes optical character recognition, page-layout interpretation, tables, forms, multilingual text, and integration with business rules. Product capability alone is not a durable barrier, as major cloud providers and specialized software firms offer adjacent capabilities. The strongest defensibility comes from accumulated document-model tuning, workflow integrations, language-specific evaluation data, and enterprise deployment experience in regulated workflows. These advantages can reduce switching costs for existing customers, but they do not create a fabrication-scale or standards-based moat comparable with leading chipmakers, cloud platforms, or foundational-model companies with materially larger compute budgets. Upstage's intellectual property position cannot be quantified due to a lack of public patent disclosures.

Five-pillar assessment

Scale and market position: A growing private enterprise AI software provider with reported FY2025 revenue of 24.8 billion Korean won and a valuation exceeding 1 trillion Korean won.
Technology and R&D: Differentiated by Asian-language model optimization and proprietary document-parsing tools, though reliant on external cloud infrastructure for compute.
Supply-chain centrality: Strong strategic partnerships with AWS, Oracle, and AMD for infrastructure, alongside backing from major Korean corporate and institutional investors.
Financial momentum: Rapid revenue growth and substantial Series C fundraising, offset by significant operating losses and negative cash flow as it funds expansion.
Governance and quality: Independent professional management with notable institutional backing, but limited transparency due to its private status and lack of public financial disclosures.

Supply chain and relationships

Upstage depends heavily on hyperscale infrastructure and external semiconductor ecosystems rather than owning significant AI-compute capacity. Amazon Web Services is its preferred cloud provider under a strategic collaboration agreement. Upstage uses Amazon SageMaker for model development and AWS Trainium and Inferentia chips for training and inference workloads. It also utilizes Oracle Cloud Infrastructure to develop models and collaborates with AMD for sovereign AI infrastructure in Korea. On the customer side, Upstage sells to enterprises using document automation and distributes through AWS and Oracle marketplaces. The company does not publicly name its largest end customers or disclose revenue concentration. If Upstage stopped shipping, enterprise customers would need to migrate applications and document extraction pipelines, though the company is technically replaceable by larger providers like Naver Cloud, Google, Microsoft, and specialized document-processing vendors such as ABBYY and UiPath.

Suppliers

Amazon Web ServicesPreferred cloud provider, SageMaker, Trainium, Inferentia
OracleOracle Cloud Infrastructure for model development

Partners

AMDSovereign AI infrastructure collaboration and investor
IntelOptimization for Intel Core Ultra processors
Polaris OfficeOn-device AI document software

Competitors

Naver CloudInferredKorean-language AI and enterprise AI services
LG AI ResearchInferredEnterprise-focused Korean AI models
KTInferredKorean enterprise and sovereign-AI opportunities
KakaoInferredConsumer-scale digital platforms and domestic applications
Google CloudInferredDocument AI, language models, and cloud compute
MicrosoftInferredAzure AI, document processing, and Copilot products
ABBYYInferredDocument understanding and workflow automation
UiPathInferredDocument understanding and workflow automation

Geopolitics and risk

Upstage's primary geopolitical exposure is its reliance on US cloud and AI-chip ecosystems. Its model training and deployment depend heavily on AWS services and US-designed processors like Trainium and Inferentia. This dependence means that changes in US export controls, cloud-service policies, chip availability, or pricing can directly affect its model development and unit economics. The company also collaborates with AMD on sovereign AI infrastructure, further tying its operations to US technology providers. Domestically, Upstage benefits from South Korean sovereign-AI policy interests and national efforts to build local AI capabilities. However, policy-linked opportunities can be affected by procurement cycles, government priorities, and competition from larger Korean corporate groups like Naver and KT. As it expands cross-border into the United States and Japan, Upstage must navigate localization, channel costs, and varying data-residency and privacy regulations for sensitive enterprise documents.

Risk matrix
Risk	Severity	Why it matters
Dependence on US cloud ecosystems	High	Model training and deployment depend on AWS services and US-designed chips.
Large-model capital intensity	High	Continued model competition requires significant external capital amid operating losses.
Competition from hyperscalers	High	Larger rivals can invest more in foundation models, infrastructure, and distribution.
Private-company disclosure limits	High	Lack of public financial statements limits assessment of durability and concentration risk.
Customer concentration	Medium	Document-AI companies can become exposed to a small number of regulated-enterprise accounts.
South Korean sovereign-AI policy	Medium	Policy-linked opportunities can be affected by procurement cycles and larger competitors.
Data-residency and privacy regulation	Medium	Customers in regulated sectors may require private deployment and strict residency controls.
Cross-border expansion execution	Medium	Expansion in the US and Japan faces localization, channel costs, and security requirements.

Governance and ownership

Upstage is an independent private South Korean corporation led by co-founder and CEO Sung Kim. It is not a listed chaebol affiliate, state-owned enterprise, or publicly identified state-controlled company. The company has attracted investment from the state-owned Korea Development Bank and IBK Industrial Bank of Korea, alongside corporate investors like Hyundai Motor Company, Kia Corporation, and SK Networks, which reportedly holds an 11.7 percent stake. As a private company, Upstage does not publish a conventional investor-relations financial package, audited public financial statements, or a complete named board roster. Its regulatory posture is shaped principally by South Korean corporate law, data-protection requirements, and enterprise security obligations.

What to watch

Whether Upstage completes a pre-IPO financing or files for a South Korean listing.
Whether FY2026 revenue growth remains strong enough to narrow operating losses.
Whether the company publishes audited financial statements and detailed R&D spending ahead of any IPO.
Whether AWS marketplace distribution produces disclosed enterprise deployments and recurring revenue.
Whether AMD collaboration yields material commercial commitments in South Korea.
Whether Solar retains competitive Korean and Japanese language-model performance against larger rivals.

Recent developments

September 7, 2026SK Networks reportedly holds an 11.7 percent stake after exercising a call option and participating in Series C financing.
June 16, 2026Announced expansion of AI services through Upstage Company and reported acquisition-related plans involving Timely and Daum.
April 15, 2026Announced the first closing of a 180 billion Korean won Series C round led by Sazze Partners.
August 20, 2025Announced a US$45 million Series B bridge round and signed a strategic collaboration agreement with AWS.

Coverage on AsiaAI.FYI

$2T IPO Prep: Anthropic Reports Two Consecutive Quarters of Profitability September 14, 2026

Guides that cover Upstage

Korea AI Industry Map

About this profile

Compiled with AI-assisted research from company filings, market data, and published reporting as of September 14, 2026, then reviewed by AsiaAI.FYI. Figures marked Estimate are not company-reported. Check primary filings before relying on any number.

Confidence: C. Identity, funding rounds, and reported FY2025 financials have reasonable support from press and corporate notices, but full financial statements, historical revenue, and R&D metrics are lacking because the company is private.

Main sources: Company product pages; Company news releases; Corporate capital-increase notices; Cloud-partner materials; Korean and international business press; Private-company data platforms.

All 35 sources

Position in the AI stack

Develops proprietary language models and document-intelligence software for enterprise workflow automation.

Legal name: Upstage Co., Ltd.
Headquarters: Yongin, South Korea
Founded: 2020
Founders: Sung Kim / 김성훈
Park Eun-jung
Lee Hwal-seok
Leadership: Sung Kim / 김성훈, Co-Founder and CEO / Representative Director
Employees: 141 (external estimate) (2026)
Ownership: Private company
Business: Enterprise AI, language models, and document intelligence
Sector: AI Software & Models
Website: upstage.ai