Yazılar

Nebius Signs $17.4 Billion AI Infrastructure Deal With Microsoft

/in /tarafından

Nebius Group (NBIS.O) announced on Monday a five-year agreement with Microsoft (MSFT.O) to provide GPU infrastructure capacity, a deal valued at $17.4 billion that could expand to $19.4 billion if additional services are acquired. The news sent Nebius shares surging more than 47% in after-hours trading.

The partnership highlights the escalating demand for high-performance AI compute as tech giants race to secure infrastructure for training and running advanced models. Under the deal, Microsoft will gain access to Nebius’ dedicated GPU infrastructure from a new Vineland, New Jersey data center starting later this year.

Nebius specializes in offering AI cloud services powered by Nvidia GPUs, combining computing, storage, management tools, and in-house designed hardware to support AI developers. CEO Arkady Volozh said the deal is not only financially significant but also positions Nebius for accelerated AI cloud growth from 2026 onwards.

Microsoft already stands as the largest customer of CoreWeave (CRWV.O), another AI infrastructure provider. The Nebius agreement suggests the company is broadening its supply chain to mitigate risks as hyperscaler demand grows.

Amsterdam-based Nebius was formed after the split of Russian tech giant Yandex, and has been expanding rapidly into the U.S. and European AI infrastructure markets.

Apple Hit With Lawsuit Over Use of Books in AI Training

/in /tarafından

Apple was sued Friday in federal court in Northern California by authors who accuse the company of illegally using copyrighted books to train its “OpenELM” large language models. The proposed class action, filed by writers Grady Hendrix and Jennifer Roberson, claims Apple copied protected works without consent, credit, or compensation.

“Apple has not attempted to pay these authors for their contributions to this potentially lucrative venture,” the lawsuit alleges. Neither Apple nor the plaintiffs’ lawyers immediately commented.

The case adds Apple to the growing list of tech giants—Microsoft, Meta, and OpenAI among them—facing litigation over whether training AI on copyrighted material constitutes infringement or fair use. On the same day, Anthropic agreed to a $1.5 billion settlement with authors who accused it of training its Claude chatbot on pirated books, a deal hailed as the largest copyright recovery in history.

According to the lawsuit, Apple’s models were trained on a known dataset of pirated books, allegedly including works by Hendrix and Roberson. The case seeks damages and legal recognition that Apple must compensate authors when their intellectual property is used to build AI systems.

The dispute underscores the escalating clash between AI developers and creators, as courts weigh how copyright law applies to massive datasets powering generative AI. With multiple cases now moving forward in U.S. courts, the outcome could reshape both the AI industry and protections for authors in the digital era.

Microsoft Brings Native Audio Creation to Copilot With New Expressive Voice Options

/in /tarafından

Microsoft has introduced a new native audio generation feature to its Copilot platform, expanding its AI capabilities beyond text and images. With this update, users can now provide Copilot with a written script and have it converted into a natural-sounding AI voiceover in different expressive styles. Unlike traditional text-to-speech tools, Microsoft claims its system delivers audio that feels more authentic and less robotic. This breakthrough is powered by the company’s in-house MAI-Voice-1 AI model, first unveiled in late August.

The announcement was made by Mustafa Suleyman, CEO of Microsoft AI, through a post on X (formerly Twitter). Suleyman highlighted that the audio generation feature is currently available through Copilot Labs, but only to users signing in with a personal Microsoft account. The move signals Microsoft’s intention to test the feature on a smaller scale before rolling it out more broadly across its ecosystem of apps and services.

At launch, Copilot offers users three distinct voice modes. The first, Scripted mode, delivers a straightforward and literal read of the input text, making it well-suited for use cases such as formal announcements, e-learning, and document narration. This ensures a professional and clear tone without unnecessary dramatization.

The second mode, called Emotive, is designed to add flair and expression. By varying pitch, tone, and pacing, it creates a more dynamic and engaging delivery that feels closer to human storytelling. Microsoft says this style is best for marketing, advertising, or entertainment contexts where dramatic impact matters. A third style, which Suleyman has hinted at but not fully detailed yet, is expected to further broaden Copilot’s voice versatility, giving users additional creative options.