TLDR: South Korean AI startup Upstage has announced a significant partnership with Amazon Web Services (AWS) to accelerate its global expansion into the Asia-Pacific and US markets. The collaboration aims to enhance Upstage’s Solar large language model (LLM) and make high-performance, cost-efficient generative AI more accessible worldwide. As part of this strategic alliance, Amazon has also made a minority investment in Upstage.
South Korean artificial intelligence innovator, Upstage, has officially deepened its strategic collaboration with Amazon Web Services (AWS), marking a pivotal step towards global expansion and the advancement of its generative AI capabilities. The partnership, announced on August 19, 2025, will see Upstage designating AWS as its preferred cloud provider, leveraging AWS’s robust infrastructure to build, train, and deploy its cutting-edge foundation models.
Under the terms of the agreement, Upstage will utilize AWS’s comprehensive machine learning stack, including Amazon SageMaker, as well as AWS Trainium and Inferentia chips. This powerful technological backbone will be instrumental in powering Upstage’s flagship Solar large language model (LLM) and its AI-based document processing suite. The Solar lineup is already available through prominent AWS platforms such as Amazon Bedrock Marketplace, SageMaker JumpStart, and AWS Marketplace, facilitating broader access for enterprises.
Kim Sung-hoon, CEO of Upstage, expressed his enthusiasm for the renewed partnership, stating, “Upstage and AWS have built a strong foundation through years of collaboration, and this agreement marks a new chapter in our shared vision to accelerate global AI innovation.” He further emphasized the strategic importance, adding, “Through this strategic partnership, we will expand our reach and deliver safe, intelligent and high-performance AI solutions to more public institutions.”
Echoing this sentiment, Abhijeet Muzumdar, Vice President of Corporate Development at Amazon, commented, “We are pleased to support the next stage of Upstage’s growth as it expands its Solar foundation model and document intelligence solutions. By backing visionary companies like Upstage, we aim to foster continued innovation in generative AI.”
This collaboration builds upon a multi-year relationship between the two companies. Upstage has historically trained its Solar models on Amazon SageMaker, significantly reducing training times. Furthermore, the company has developed and implemented its proprietary Depth-Up Scaling (DUS) method on AWS infrastructure, which has proven effective in enhancing GPU efficiency and improving model performance.
Key initiatives stemming from this partnership include joint go-to-market strategies and sales initiatives targeting both the Asia-Pacific (APAC) region and the United States. The goal is to deliver generative AI solutions optimized for performance, security, and compliance, specifically catering to government agencies, public institutions, and regulated industries.
An integral part of this expanded collaboration is the ‘AI Initiative,’ originally launched in April 2025. This program provides free access to Upstage’s full suite of AI models, including Solar and its OCR-powered Document AI solution, along with AWS infrastructure support, to qualified educational institutions and non-profit organizations. Over 200 institutions, including prestigious ones like Seoul National University and Korea Advanced Institute of Science and Technology (KAIST), are already participating, fostering generative AI adoption beyond the commercial sector.
Also Read:
- AWS and Code.org Expand AI Literacy Initiatives, Aiming to Empower Millions Globally
- Megazone Cloud and AWS Partner for AI Connect Seminar on Generative AI in Customer Experience
Upstage’s latest offering, Solar Pro 2, a 31-billion-parameter model, is designed for advanced reasoning, featuring dual modes for rapid responses and structured logic, and is capable of autonomously executing complex workflows. Complementing its LLMs, Upstage’s Document Intelligence solution is crucial for organizations seeking to efficiently extract and structure data from diverse documents.
