Korea–U.S. ties boost AI infra
The AI infrastructure landscape is tightening links between U.S. tech firms and Korea’s industrial base. In Seoul, Samsung, SK, and OpenAI signed a series of agreements under the Stargate initiative, pledging to expand DRAM wafer production to 900,000 starts per month and study next-generation AI data centers across Korea. In Italy, Microsoft’s CTO said the company intends to prioritize its chips over Nvidia and AMD, pairing in-house silicon with innovations such as microfluidic cooling to meet surging demand. Meanwhile, Korean operator KT unveiled SOTA K, a GPT-4o-based system developed with Microsoft and trained on domestic datasets, delivering cultural and linguistic performance benchmarks that surpass global peers. Taken together, these moves underscore how AI infra is evolving on three fronts — strategic international partnerships, sovereign model development, and a race toward silicon independence — setting the stage for the next wave of capacity growth. More below!
Juan Pedro Tomas
Editor
RCRTech
AI Infrastructure Top 3
Samsung, SKT join Stargate: Samsung SK and OpenAI expand partnerships under the Stargate initiative scaling memory chip supply to 900,000 wafers monthly and evaluating new AI data centers across Korea.
Microsoft bets on own chips: Microsoft CTO Kevin Scott said the company ultimately wants its own chips powering AI data centers reducing reliance on Nvidia and AMD as global compute shortages continue.
KT debuts Korea-tailored AI: KT unveiled SOTA K, an AI model built with Microsoft on GPT-4o optimized for Korean grammar culture and context with benchmarks showing it surpasses global models in comprehension reasoning.
AI-Powered Telecom Infrastructure
Supermicro, in collaboration with NVIDIA, delivers AI-powered infrastructure tailored for telcos, enhancing operational efficiency, network management, and customer experiences. Explore now
AI Today: What You Need to Know
Anthropic taps new CTO: Anthropic appointed former Stripe exec Rahul Patil as CTO, strengthening compute, inference and reliability strategy as the company faces surging Claude demand and intensifying rivalry with OpenAI and Meta.
SuperX plants U.S. roots: Singapore-based SuperX launched a Silicon Valley arm to co-develop full-stack AI systems positioning itself closer to partners and talent pools driving hyperscale innovation.
Switch market set to surge: AI data center switch revenues are forecast to rise from $4 bn in 2024 to $19bn by 2030 with hyperscalers’ insatiable networking needs fueling demand for high-throughput low-latency gear.
DOE eyes Savannah River hub: The U.S. Department of Energy opened bids for a 250–450 acre AI data center in South Carolina requiring new power generation capacity with project awards expected in December.
Dashmote raises $6M: Amsterdam’s Dashmote secured $6 million to launch a mobile-first AI sales agent bringing enterprise-level capabilities to SMBs that lack the resources for large-scale AI adoption.
Adobe launches enterprise AI agents: Adobe rolled out a family of AI agents integrated with its Experience Platform designed to help enterprises automate customer journeys while maintaining brand governance and security.
Google commits $4B in Arkansas: Google unveiled plans for a $4 billion AI campus spanning 1,100 acres in Arkansas positioning the site as one of the largest U.S. AI data center projects to date.
Rolls-Royce unveils fast-start gas gen: Rolls-Royce introduced generators that deliver 2.8MW in just 45 seconds offering AI data centers enhanced resilience amid soaring power demand and grid reliability concerns.
AI hyperscale centers explained: Hyperscale facilities power the AI era by combining massive chip clusters with liquid cooling and ultra-fast networking enabling the real-time processing needed for next-generation workloads.