MangoBoost Expands AMD Partnership to Compete With Nvidia in AI Inference
South Korean startup MangoBoost advances AI inference platform with AMD collaboration, aiming to supply domestic and regional AI infrastructure and reduce reliance on external cloud providers.
South Korean AI infrastructure startup MangoBoost has strengthened its collaboration with Advanced Micro Devices (AMD) to challenge Nvidia’s position in the AI inference market through integrated hardware, software and inference services. MangoBoost unveiled its new AI inference platform, Mango Inference, which enables users to access AI models serverlessly via a single API key, aiming to simplify AI deployment for businesses and developers.
Quick Skim
MangoBoost Develops Serverless AI Inference Platform
The Mango Inference service allows customers to select from open-weight AI models such as DeepSeek and GLM and pay based on usage. MangoBoost’s proprietary inference optimisation software, LLMBooost, reportedly reduces data transfer bottlenecks between storage and GPUs, lowering costs by at least 20% compared to current high-performance alternatives.
CEO Kim Jang-woo described the company’s transformation from a component supplier to an ‘AI factory operator’ integrating hardware and software to deliver AI services directly to users.
AMD Collaboration Secures GPU Supply and Infrastructure Growth
MangoBoost’s partnership with AMD is central to its strategy, enabling access to tens of thousands of AMD GPUs, including phased deployment of the latest Instinct-series at a new 1-megawatt AI facility in Seoul’s Gangnam district. Full completion of this infrastructure is expected by the first quarter of 2027.
The company plans to increase the share of internally designed systems to cut dependence on external cloud providers and target domestic ‘sovereign AI’ applications that require secure data processing within South Korea. MangoBoost intends to expand on-premises AI system supply to security-sensitive sectors such as finance and defence.
The startup also aims to extend its infrastructure across the Asia-Pacific region and the United States with a long-term expansion goal of up to 500 megawatts capacity, initially funded from own resources with potential joint ventures to follow.
Implications for AI Infrastructure Deployment in Asia-Pacific
The collaboration between MangoBoost and AMD exemplifies a growing trend where local providers integrate hardware and software to deliver AI services tailored for regional data sovereignty and operational efficiency. This approach can reduce costs and improve reliability by limiting dependence on international cloud providers.
For manufacturing, maritime, and utilities industries in Asia-Pacific, the emerging ability to deploy AI inference services locally may accelerate adoption of AI-powered operations and data analytics with enhanced security and regulatory compliance.
Looking ahead, MangoBoost’s roadmap to scale GPU supply and infrastructure capacity will be a critical factor in how effectively it can compete with Nvidia’s established AI ecosystem.
