Nvidia

In a new partnership with chipmaker Nvidia, Universal Music Group plans to introduce what it calls “responsible AI” that could change music discovery and creation.

The companies will begin research on how to advance human music creation and compensation for rights holders in the age of AI, as revealed in the deal announced Tuesday. With this technology, the new partners say they hope to leverage AI-powered tools to aid and protect artists’ work, instead of using hands-off generative AI.

“We look forward to working closely with NVIDIA to direct AI’s unprecedented transformational potential towards the service of artists and their fans as we work together to set new standards for innovation within the industry, while protecting and respecting copyright and human creativity,” said Sir Lucian Grainge, Universal Music Group’s chief executive, in a press release.

Universal Music Group will use Nvidia’s Music Flamingo program, a large audio-language model designed to understand music in-depth. It was launched in November and can understand musical elements like structure, harmony, instrumentation and lyrics. The program can process songs up to 15 minutes long and will also be able to capture the historical and cultural context, as well as various emotional arcs.

With the program’s ability to process songs thoroughly, Universal Music Group aims to use this tool to help connect artists and fans. Instead of relying on typical genres or tags, Music Flamingo allows listeners to discover new music in a more automated fashion.

There are also plans for Nvidia and Universal Music Group to begin developing an incubator in which artists, songwriters and producers will help design and experiment with new AI tools. The company said it hopes this process will help AI tools fit into creative processes with greater ease.

Universal Music Group, with its corporate headquarters in the Netherlands and another office in Santa Monica, was founded in 1996. The music giant behind artists like Taylor Swift and Billie Eilish is valued at roughly $40 billion on the U.S. stock market, with shares selling around $25.35 each. The deal with Nvidia follows several other AI agreements that Universal has inked with companies like Klay and Stability AI.

Nvidia was founded in 1993, with the original goal of bringing 3D graphics to video games and multimedia projects. As the tech industry has continued to evolve, Nvidia has emerged as the leader in computer chips designed to power AI tools and applications.

“By extending NVIDIA’s Music Flamingo with UMG’s unmatched catalog and creative ecosystem, we’re going to change how fans discover, understand, and engage with music on a global scale,” said Richard Kerris, the general manager for media and entertainment at Nvidia, in a statement.

“And we’ll do it the right way: responsibly, with safeguards that protect artists’ work, ensure attribution, and respect copyright.”

Dec. 16 (Asia Today) — Rebellion Chief Executive Park Sung-hyun said Tuesday the South Korean AI semiconductor startup wants to “compete head-to-head” with Nvidia as the company marked its fifth anniversary and said its valuation has reached about 2 trillion won (about $1.5 billion), meeting the threshold commonly used for “unicorn” status.

“Even if it kills me, I want to step into the same ring as NVIDIA and face them head-on,” Park said at a media day at Rebellion’s headquarters in Seongnam, south of Seoul.

Founded in 2020, Rebellion has positioned itself as an AI chipmaker focused on inference – the computing used to run AI services – rather than large-scale model training. The company said it has built “real-world” usage references by deploying its chips in services with live traffic in telecommunications, the public sector and enterprise markets.

Executives said competition in AI semiconductors is shifting as AI services spread and inference becomes a key battleground, where power efficiency and operating costs can matter as much as raw performance. The company pointed to moves such as Google‘s expansion of its Tensor Processing Unit into large-scale cloud offerings as evidence that specialized AI chips developed for internal use can be adapted for commercial services.

Rebellion said it is seeking to differentiate itself in a market not fully centered on Nvidia by focusing from the outset on inference-optimized designs. Park said the company expects measures such as cost per token and throughput per watt to become increasingly important as AI services scale.

Park also criticized what he described as the practical challenges facing domestic AI chip companies, arguing that government support for AI infrastructure – particularly around graphics processing units – has largely benefited large companies and established cloud providers. “This is disappointing for AI semiconductor companies targeting the inference market,” he said, while adding the company plans to pursue competition through chips and systems rather than policy-driven, software-centric approaches.

Rebellion said its merger with Sapion Korea, finalized last year, strengthened its global expansion efforts. Through the deal, SK Telecom and SK Hynix became major shareholders, providing capital and boosting credibility, Park said. He added that SK Hynix’s brand reduces the burden on Korean startups seeking recognition abroad.

The company said it raised 92 billion won (about$70 million) from KT in a 2022 Series A round and 165 billion won (about $125 million) in a 2024 Series B round from overseas investors including Saudi Aramco and Singapore’s Pavilion Capital. Rebellion said a Series C round this year included investment from Arm, which it described as a milestone for an Asian startup.

Rebellion said it mass-produced its first neural processing unit, ATOM, in 2023 and later introduced a higher-performance inference chip, REBEL-Quad. It said it has established overseas subsidiaries in Japan, Saudi Arabia and the United States as it expands international business.

Rebellion said it has selected Samsung Securities as lead underwriter for an initial public offering and has begun listing preparations. The company plans to pursue a Korean listing first, while also targeting a longer-term U.S. listing, it said.

Park said the company now sees itself as part of South Korea’s “deep tech” push and aims to become a key player in global AI infrastructure.

-Reported by Asia Today; translated by UPI

© Asia Today. Unauthorized reproduction or redistribution prohibited.