SK Telecom is investing in California-based startup Together AI to co-develop AI platforms and expand into North America. Together AI recently raised $305 million in Series B funding, reflecting strong investor confidence in its optimization technologies. The collaboration will focus on optimizing SK Telecom’s personal AI agent, large language model (LLM), and AI infrastructure. Together AI’s proprietary technologies, including ‘QTIP’ and ‘Together Kernel Collection’, will help reduce GPU costs and improve efficiency. This marks a critical step in SK Telecom’s long-term AI strategy, targeting a competitive edge in global AI ecosystems.Key Takeaways
SK Telecom has announced a strategic partnership with California-based Together AI, a rapidly scaling artificial intelligence company specializing in cloud-based AI services and open-source tools.
The move reflects SK Telecom’s ambition to accelerate the development of its AI platform and gain a foothold in the North American market.
Together AI, founded in 2022, has built its reputation by offering AI inference, fine-tuning services, and efficient GPU cluster infrastructure in collaboration with NVIDIA.
The startup serves over 500,000 customers globally, including high-profile clients like Zoom, The Washington Post, and Salesforce.
The collaboration focuses on SK Telecom’s core AI assets, including its personal AI agent Aster (A), its in-house large language model (LLM) known as adot, and the broader AI computing infrastructure through its AI Data Center (AI DC) initiative.
Financial and Technological Backing
The investment is being made shortly after Together AI raised $305 million in Series B funding, a round that underscores market confidence in the startup’s technical capabilities and its potential in the competitive AI infrastructure space.
As part of the collaboration, Together AI will contribute two critical proprietary technologies:
-
Together Kernel Collection: This suite is designed to optimize GPU computation, enhancing the throughput of large-scale AI model training and deployment.
-
QTIP (Quantized Transformer Inference Pipeline): A compression and quantization framework that significantly reduces AI model size without degrading performance, enabling more cost-effective deployment on GPU infrastructure.
By integrating these technologies, SK Telecom expects to reduce both initial GPU investments and operational costs, which are major barriers in scaling AI services globally.
Focus Areas: From Personal AI to Global Infrastructure
At the center of this initiative is SK Telecom’s next-generation AI agent, Aster (A). The company envisions Aster as a global personal AI platform, equipped with advanced language understanding and adaptive personalization.
In tandem, SK Telecom is expanding its adot LLM and data center capabilities, which Together AI’s infrastructure will directly support.
Broader Industry Implications
This partnership comes amid a global race for scalable, efficient AI infrastructure. Cloud AI providers, including well-capitalized startups like Together AI, are competing to offer differentiated services such as low-latency inference, cost-efficient model training, and open-source accessibility.
Together AI’s integration of NVIDIA-backed hardware capabilities positions it competitively within this evolving landscape. For SK Telecom, the move signals a pivot from being a regional telecom provider to becoming a tech-forward player in the global AI ecosystem.
By directly investing in cutting-edge AI infrastructure firms, SK Telecom aims to insulate itself from the high costs of AI compute, while also retaining full control over AI model development and deployment.
What Comes Next
With the partnership finalized, both companies will now begin joint development on SK Telecom’s AI stack. Early-stage priorities include:
-
Integrating QTIP into SK Telecom’s model training pipeline
-
Optimizing inference engines across Aster (A) and adot
-
Exploring expansion opportunities for SK Telecom’s AI services in North America
This marks a rare alignment between a telecom incumbent and a Silicon Valley AI infrastructure startup, potentially setting a precedent for similar cross-border collaborations in the future.
For more news and insights, visit AI News on our website.
