Tech giant unveils new processor for text-to-image AI generation on smartphones

By Park Jae-hyuk

Samsung Electronics made a series of announcements demonstrating its competitiveness in semiconductor production, as the chip industry expands alongside artificial intelligence (AI) and the emergence of ChatGPT.

Aiming to lead the global AI chip industry, the Korean tech giant has come up with a new mobile processor for advanced AI performance, while joining hands with startups here and overseas that specialize semiconductor design.

During the Samsung System LSI Tech Day event in Silicon Valley, Thursday (local time), the company unveiled the Exynos 2400, its next-generation flagship mobile processor that enables a 14.7 times boost in AI performance compared to the previous Exynos 2200.

Samsung demonstrated text-to-image AI generation using its Exynos 2400 reference board. The new AI tool is expected to be available on the Samsung Galaxy S24 smartphone, which will be released next year.

Samsung’s Zoom Anyplace technology, unveiled for the first time, uses AI-based tracking to capture close-ups of moving objects up to four times closer without image degradation.

In addition, Samsung had an in-depth discussion with academic experts at the event about the recent trends in generative AI and large language models, according to the company.

“Generative AI has quickly emerged as perhaps the most significant trend of the year, demanding more powerful foundational technologies to process data and bring AI to life,” Samsung Electronics President Park Yong-in said. “We are paving the path toward a new era of proactive AI.”

Amid TSMC’s dominance of AI chip manufacturing, Samsung also won orders from AI chip startups in Korea, Canada and the U.S.

Partnership with fabless companies

Rebellions, a Korean fabless company specializing in AI chips, said Thursday that it will jointly develop a new AI chip, Rebel, by the end of next year with Samsung. The partnership is intended to gain the upper hand in the rapidly growing generative AI market.

“Rebellions and Samsung are teaming up to develop generative AI chips featuring cutting-edge 4-nanometer technology, the most advanced HBM3E memory, and chiplet packaging,” a Rebellions official said.

Chiplets are are tiny integrated circuits (ICs) that contain a well-defined subset of functionality.

Jeong Ki-bong, vice president of Samsung Foundry, said the company regards its system semiconductors, especially those for AI, as the key to its business in the future.

On Oct. 2, Tenstorrent announced its next-generation AI chiplets will be manufactured at Samsung’s U.S. factory. The Canadian firm is led by CEO Jim Keller, a legendary American microprocessor engineer best known for his work at AMD and Apple.

“Samsung Foundry’s commitment to advancing semiconductor technology aligns with our vision for advancing RISC-V and AI and makes them an ideal partner to bring our AI chiplets to market,” Keller said.

The two companies signed the partnership after Keller hinted at Tenstorrent’s collaboration with Samsung, during his attendance at the Samsung Foundry Forum in June.

“Samsung Foundry is expanding in the U.S., and we are committed to serving our customers with the best available semiconductor technology,” Samsung’s U.S. foundry business head Marco Chisari said. “Samsung’s advanced silicon manufacturing nodes will accelerate Tenstorrent’s innovations in RISC-V and AI for data center and automotive solutions.”

In August, Groq, a U.S. AI solutions company founded by engineers from Google, selected Samsung as the manufacturer of its semiconductor for AI acceleration.

“Our partnership with Samsung will allow us to continue that leap forward, by tapping into the most advanced semiconductor manufacturing technologies available,” Groq CEO Jonathan Ross said at that time.