Nvidia Expands Territory with AI Model Development... Server Prices Set to Rise by 15%

Licensing Poolside’s Technology for $6 Billion
Targeting DeepSeek and Kimi
Passing Cost Burden Amid Soaring Memory Prices
Direct Competition With OpenAI and Anthropic Begins
Expanding the AI Ecosystem on All Fronts

NVIDIA, which dominates the artificial intelligence (AI) semiconductor market, is expanding its business scope to include AI model development. In response to the rise of open-weight AI models such as China's DeepSeek and Kimi, the company aims to enhance its own model competitiveness. Meanwhile, due to the sharp increase in memory semiconductor prices, AI server prices are expected to rise by more than 15%.

Jensen Huang, CEO of NVIDIA, finished his visit to Korea last June and was departing through the Seoul Gimpo Business Aviation Center when he raised his thumb while listening to questions from the press. Photo by Yonhap News Agency

According to the Wall Street Journal (WSJ) on August 23 (local time), NVIDIA plans to utilize a $6 billion technology licensing deal signed with AI startup Poolside to develop world-class open-weight AI models. The goal is to create models that can compete with China’s DeepSeek and Moonshot AI’s Kimi.

NVIDIA will invest $1 billion in Poolside, based on a $12 billion valuation, and separately pay $6 billion to license Poolside’s technology and recruit most of its engineers. Around 100 Poolside staff, including engineers, are expected to join NVIDIA to work on developing the open-weight AI model “Nemotron.”

Poolside is an AI startup co-founded in 2023 by software developer Eiso Kant and former GitHub Chief Technology Officer (CTO) Jason Warner. Its recently released open-weight model “Laguna S” has drawn attention in Western markets, reportedly accelerating the contract with NVIDIA.

WSJ reports that NVIDIA has been increasing its investment in developing related models in response to the rapid growth of Chinese open-weight models. In March, the company launched the “Nemotron Alliance,” which includes Mistral, Thinking Machines Lab, and Perplexity, agreeing to share data, expertise, and computing resources.

Jensen Huang, CEO of NVIDIA, emphasized in an article published last month titled “Open-Weights and America’s AI Leadership” that the strength of US AI lies not in a single cutting-edge model, but rather in “building a strong and open ecosystem that spreads across all industries.”

Such moves are placing NVIDIA in direct competition with its existing customers. OpenAI and Anthropic are key customers of NVIDIA’s AI accelerators, but are also leading companies in the closed AI model market. As NVIDIA enhances its own model competitiveness, competition with these former customers and partners becomes inevitable.

However, major big tech companies such as OpenAI, Google, Microsoft (MS), and Meta are also developing their own AI semiconductors in an attempt to reduce their dependence on NVIDIA, prompting some to argue that the competition between both sides is already underway.

AI Server Prices to Rise by More Than 15% Due to Surge in Memory Costs

Samsung Electronics and SK hynix.

As NVIDIA widens its business scope to AI models, cost pressures are mounting as well. According to Bloomberg, some of NVIDIA’s major customers have received notice that server prices equipped with NVIDIA’s AI semiconductors will rise by more than 15% for many products. The extent of the increase will depend on the generation of NVIDIA chips and memory configuration applied to each product.

This price hike will apply to products shipped from early next year and will also include systems equipped with NVIDIA’s next-generation Vera Rubin and Grace Blackwell chips.

The main driver behind rising server prices is memory semiconductors. While Samsung Electronics, SK hynix, and Micron account for most of the world’s DRAM supply, the speed of production expansion has not kept pace with the explosive increase in AI infrastructure demand, leading to a sharp price surge.

Bloomberg noted that even NVIDIA, one of the most profitable companies in the semiconductor sector, now finds it difficult to maintain prices or absorb additional costs internally.

This means that the bargaining power of memory suppliers such as Samsung Electronics, SK hynix, and Micron has increased significantly as AI investments have expanded. Bloomberg analyzed that the mounting influence of memory suppliers is unprecedented, as memory supply is unable to keep up with soaring AI infrastructure demand.

NVIDIA’s server price hikes are likely to drive up the cost of building large-scale AI data centers. Although major big tech companies such as Amazon, MS, Google, and Meta are increasing their efforts to develop their own AI semiconductors, they still rely heavily on NVIDIA products for large-scale data center construction.

Bloomberg reported that whether the server price hikes open opportunities for rival companies will depend on how reliably customers can secure memory supplies.

This content was produced with the assistance of AI translation services.