[AI Brief] Naver Cloud to Pursue AI Factory Business With Vietnam's FPT
IT DAILY ·
✦ AI Summary
IT Daily introduced its "AI Brief" serialized roundup of major AI news from Korea and abroad.
Naver Cloud and FPT are jointly pursuing the AI factory business and will work to expand their GPUaaS business and jointly identify demand in the APAC market.
Vessl AI provided global inference infrastructure for Upstage's "Solar Pro 4," and ElevenLabs said demand for conversational agents is rising and its valuation reached USD 22 billion.
IT Daily introduced its serialized "AI Brief" roundup of major AI news from Korea and abroad.
The people in the photo were identified as Kim Yoo-won, CEO of Naver Cloud; Le Hong Bi, CEO of FPT Corporation's Smart Cloud; and Lee Soo-an, senior vice president of FPT Korea.
Naver Cloud and FPT will jointly pursue the AI factory business. Naver Cloud recently signed a memorandum of understanding (MOU) with FPT at Naver 1784 and announced the agreement's details on the 12th.
The two companies plan to jointly discuss the planning, development and operation of the AI factory business. The core of the partnership is linking both companies' GPU infrastructure, cloud technology, business capabilities and customer networks.
Based on this, the two companies will push ahead with expanding their GPUaaS business in APAC markets, including Korea, Japan and Vietnam. Their goal is to jointly broaden opportunities for AI factory and GPUaaS businesses in the APAC region.
The two companies agreed to jointly identify demand in regions that are difficult to target individually. The targets are companies that need GPU computing resources for large-scale AI training and inference, and AI service providers. The two companies plan to provide these customers with infrastructure, AI models and AI services, and support AI adoption without the burden of building their own infrastructure.
The two companies will also cooperate to strengthen engineering capabilities for applying AI technology in the field. To that end, they will operate a regular technical exchange program involving Naver Cloud's Vietnam development center, headquarters technical staff and FPT engineers. They will also jointly work on tasks such as optimizing GPU infrastructure at customer sites, improving AI inference performance and building AI agent service environments.
Kim Yoo-won, CEO of Naver Cloud, said the company would be able to combine Naver Cloud's AI technology and experience with FPT's AI factory and enterprise business capabilities. He added that the company would pursue cooperation to respond to diverse AI demand in the APAC market beyond Korea and Vietnam, and build cooperation that leads to actual business results.
Vessl AI announced on the 12th that it will provide global inference infrastructure for Upstage's LLM "Solar Pro 4." Vessl AI and Upstage shared the launch schedule, promotion schedule and expected demand, and jointly designed the necessary inference capacity accordingly.
To support the global service operation of "Solar Pro 4," the two companies prepared a global serving system using GPU clusters in multiple regions in Korea, Europe and the United States. GPU resources were configured based on model characteristics and actual traffic patterns.
Vessl AI applied its own inference engine optimization and proprietary optimization technologies in the process. This optimized the serving infrastructure to increase throughput, sped up user inference request processing and reduced token-level costs and overall costs.
Under this operation, "Solar Pro 4" ranked third in September usage on OpenRouter, a global LLM marketplace. The usage ranking was based on actual usage tallied by the platform, and Vessl AI handled the global serving of "Solar Pro 4" during that period.
The company said it applied improvements to internal cache hit rates, model quantization and inference speed. As a result, processing speed averaged 2 times faster and reached up to 3 times faster before and after endpoint deployment, and token-level costs were cut by up to 2 times based on the higher cache hit rate.
Meanwhile, ElevenLabs said on the 12th that enterprise demand for conversational agents is increasing. Against this backdrop, it said its valuation had risen to USD 22 billion.
ElevenLabs recently closed a USD 300 million employee tender offer based on a valuation of USD 22 billion. It said this is about 2 times the valuation at the time of its Series D round in February.
The company secured its first investment in 2022 at a valuation of USD 9 million. In the same year, it released its first text-to-speech (TTS) model, "Eleven v1."
It has now released two new TTS models, "Eleven v4" and "v4 Turbo," and built an AI interaction platform alongside the models. The company said this allows enterprises to deploy conversational agents linked to their own knowledge and systems in line with industry, region and use-case context.
Mati Staniszewski, cofounder and CEO of ElevenLabs, said regarding AI's direction toward human interaction that it should interact in ways aligned with how people communicate with one another. He added that ElevenLabs is seeing growing adoption of expressive voice agents by companies and governments, and that the services targeted for adoption are consumer- and citizen-facing services. He also said it is important for employees to share the value they create for customers and the wider world.
Source: IT DAILY · Yang Seung-gap
Original: https://www.itdaily.kr/news/articleView.html?idxno=242127
References
This article was produced with the help of an automated content generation algorithm.
Source: IT DAILY
View originalThis article was summarized and organized by BizCrush based on the original article from IT DAILY. For exact quotations and full details, please refer to the original article.