Meta, the parent company of Facebook, is reportedly in discussions to acquire FuriosaAI, a South Korean artificial intelligence (AI) chip startup, according to foreign media reports. This move signals Meta's push to accelerate the development of its own AI chips to achieve "low-cost, high-efficiency" in the increasingly competitive AI development landscape.

Forbes, an American business magazine, reported on February 11 (local time), citing sources, that the acquisition process could be completed within this month.

FuriosaAI, which designs AI inference-specialized semiconductors, was founded in 2017 by Baek Jun-ho, a former Samsung Electronics and AMD engineer, along with engineers from major tech companies. The company has demonstrated its competitiveness by introducing its first AI chip, 'Warboy,' in 2021, which consumes less power than comparable products. In August of last year, FuriosaAI launched its second-generation chip, 'Renegade,' which boasts up to 60% higher power efficiency than Nvidia's general-purpose chip (L40S).

Forbes reports that several companies are showing interest in FuriosaAI, and Meta is one of them.

Meta's pursuit of acquiring FuriosaAI is interpreted as an effort to develop its own semiconductor chips, which are essential for AI development. Big tech companies, including Meta, are currently spending huge amounts of money on purchasing Nvidia's high-performance AI chips. To reduce these costs, they are seeking alternative solutions, such as producing their own chips. Nvidia dominates more than 80% of the global AI chip market, and related companies are concerned about their dependence on Nvidia in terms of supply and pricing.

Previously, Meta partnered with Broadcom, a leading application-specific integrated circuit (ASIC) company, to develop its own custom AI chips. Broadcom is also working with OpenAI, the developer of ChatGPT, to design AI chips.

Meanwhile, Meta announced that it plans to invest up to $65 billion (about 94 trillion won) this year in building AI infrastructure, including the construction of new data centers.

[Copyright (c) Global Economic Times. All Rights Reserved.]