British investment group Petershill Partners announced on Thursday that it will delist from the London Stock Exchange (LSE) and return capital to its shareholders, citing persistent disappointment with its share price performance and valuation.

The decision, which values the Goldman Sachs-backed firm at $4.5 billion, marks another high-profile setback for London’s stock market, which has struggled to keep companies from leaving amid concerns about low valuations and dwindling investor interest.

Petershill, majority-owned by Goldman Sachs, was established in 2007 with a strategy centered on acquiring minority stakes in hedge funds and other alternative asset managers. The group went public in London in September 2021, touting the listing as a gateway for investors to access the fast-growing alternative investment industry. Chairman Naguib Kheraj described it at the time as a “unique opportunity” for public market investors to gain diversified exposure to private markets.

Since then, Petershill has expanded significantly, with assets under management rising from $187 billion at the time of its IPO to $351 billion as of June 2025. Yet its shares have consistently underperformed, trading at steep discounts to both book value and comparable listed peers. Even strong operational and financial results failed to lift sentiment, leaving the company’s valuation lagging well behind the industry it sought to showcase.

Under the delisting proposal, free-float shareholders will receive $4.15 per share in cash, alongside an interim dividend of $0.052 per share, totaling $4.202. The offer represents a 35% premium to the stock’s last closing price, prompting shares to jump as much as 34% on Thursday. Goldman Sachs-managed private funds, which control 79.49% of Petershill’s stock, will not participate in the capital return but have agreed to support the move. Approval from free-float shareholders representing at least 75% of votes cast at court and general meetings scheduled for November is required.

Analysts at Jefferies described the decision as “perfectly reasonable,” noting that Petershill had undertaken numerous initiatives over the past 18 months to close the valuation gap with little success. The delisting, they said, offers clarity for investors and acknowledges that London’s market has failed to reward Petershill’s scale and performance.

The move is part of a broader pattern that has deepened worries about London’s status as a global financial hub. Ireland-based building materials giant CRH shifted its primary listing to New York in 2023, citing the deeper capital pools and higher valuations available there. British semiconductor design champion Arm Holdings opted for a Nasdaq IPO the same year, despite heavy lobbying by UK officials to keep it in London. Flutter Entertainment, owner of Paddy Power and FanDuel, also moved its primary listing to New York this year to tap U.S. investors.

These exits, combined with a thinning pipeline of new listings, have left the LSE grappling with what many see as an identity crisis. Even after reforms to simplify listing rules, London continues to suffer from a reputation for depressed valuations compared with New York or even European markets such as Amsterdam. Petershill’s decision, despite its growth and ties to Goldman Sachs, underscores that London is struggling to retain firms even in sectors where it once held an advantage.

For Petershill, the delisting is a way to reconcile strong operating performance with a market that never seemed to buy into its promise, while it underlines the continuing exodus that risks eroding the role of LSE as Europe’s premier financial center.

Nvidia is throwing open the doors on one of its most intriguing AI tools. The company announced it is open-sourcing Audio2Face, a system that generates lifelike 3D facial animations from voice recordings.

Nvidia is giving developers unprecedented freedom to adapt and extend the technology for their own games, livestreaming platforms, and immersive applications by making the models, software development kits (SDKs), and even the training framework publicly available.

Audio2Face works by analyzing the acoustic features of speech — such as pitch, tone, and timing — and then mapping that data to lip movements and facial expressions in a 3D avatar. The result is a realistic synchronization of voice and animation that can be applied to both pre-scripted content and live interactions. Developers can drop the tool into an animation pipeline to reduce manual rigging and motion capture work, or use it in real time for interactive experiences like multiplayer games and virtual events.

Studios have already started experimenting with the tool. Farm51, the developer of Chernobylite 2: Exclusion Zone, has integrated Audio2Face into its production process, while Alien: Rogue Incursion Evolved Edition also leverages the technology. Nvidia has emphasized that by providing the training framework itself, it’s enabling developers to fine-tune the models — whether for specific languages, unique art styles, or performance-optimized avatars for virtual and augmented reality.

The decision to open-source marks a notable shift for Nvidia, which has historically kept its most important technologies proprietary. Its CUDA programming model, launched in 2006, remains closed-source and has effectively locked developers into the Nvidia GPU ecosystem, ensuring its dominance in AI computing. In contrast, making Audio2Face open and modifiable suggests a new approach: using open-source software as a way to drive adoption and entrench Nvidia’s broader AI ecosystem.

Industry analysts see it as part of a dual strategy. On one hand, Nvidia continues to dominate the hardware market, with its GPUs powering virtually every major AI model. On the other, it is increasingly positioning itself as a software and platform company, offering tools that make developers reliant not just on Nvidia’s chips but on its frameworks.

Audio2Face has long been a cornerstone of Nvidia’s Omniverse platform — the company’s 3D collaboration and simulation environment. By releasing it freely, Nvidia could accelerate Omniverse adoption while creating a pipeline of developers who will eventually run their workloads on Nvidia GPUs.

This contrasts with rival approaches. Epic Games, for instance, has pursued a partly open path with its MetaHuman Creator, offering advanced avatar creation tools but keeping them tied closely to Unreal Engine. Startups in the “digital human” space, meanwhile, often market their products as cloud services, limiting developer control. Nvidia’s open-source release, by comparison, hands developers full control over model training and deployment, potentially giving it an edge with studios and enterprises that want customization without lock-in.

As generative AI expands beyond text and images into video and 3D, demand for realistic avatars is rising — from gaming and film to virtual assistants and live commerce. For Nvidia, ensuring that its tool becomes the standard for AI-driven facial animation could cement its position not just in hardware but also in the emerging layer of generative AI applications.

Audio2Face is believed to have always been one of Nvidia’s most practical demonstrations of generative AI. Thus, by open-sourcing it, the chipmaker isn’t just giving developers a free tool — it’s believed to be building mindshare and ensuring that when people think about AI avatars, they think Nvidia first.

The open-source release could also accelerate experimentation across industries. Developers might adapt Audio2Face for multilingual lip-syncing, accessibility tools such as real-time sign language avatars, or even customer-service bots with more natural human expressions. With the training framework in the open, the potential use cases extend far beyond entertainment.

For Nvidia, the bet is to open up Audio2Face, which could help entrench its ecosystem just as competitors race to define the future of digital humans. While the company’s core business remains firmly rooted in selling GPUs, the open-source release signals a willingness to use software not just as a lock-in mechanism, but as a bridge to wider adoption.

Elon Musk’s artificial intelligence startup, xAI, has filed a new lawsuit against OpenAI in the U.S. District Court for the Northern District of California, escalating an already bitter feud between the two companies.

The complaint, lodged on Wednesday, accuses OpenAI of deliberately poaching xAI employees and stealing proprietary technology underpinning Grok, xAI’s flagship chatbot. Musk’s lawyers argued that OpenAI has engaged in a “deeply troubling pattern” of raiding talent to gain inside knowledge of xAI’s systems.

“OpenAI is not merely soliciting or hiring a competitor’s employees,” the filing states. “OpenAI is waging a coordinated, unfair, and unlawful campaign: OpenAI is targeting those individuals with knowledge of xAI’s key technologies and business plans.”

The lawsuit further alleges that OpenAI encouraged recruits to violate confidentiality agreements signed with xAI, effectively breaching non-disclosure obligations to secure a competitive advantage.

OpenAI rejected the allegations outright. “This new lawsuit is the latest chapter in Mr. Musk’s ongoing harassment,” the company told Business Insider. “We have no tolerance for any breaches of confidentiality, nor any interest in trade secrets from other labs.”

The case names ex-xAI engineer Xuechen Li, who is already facing a separate trade secrets lawsuit Musk’s company filed in August. Earlier this month, a judge granted xAI’s request for a temporary order barring Li from working on or communicating about AI technology with OpenAI. Still, a source familiar with the matter said Li had never worked for OpenAI, raising questions about the scope of Musk’s claims.

The complaint also references other departures, including “early xAI engineer” Jimmy Fraiture and a senior finance executive, both of whom have since joined OpenAI. While Fraiture is not a defendant, xAI alleges the hires form part of OpenAI’s broader strategy to extract sensitive information.

This lawsuit is the latest escalation in a widening legal war between Musk and OpenAI, the company he cofounded in 2015 but later abandoned amid strategic disagreements. Musk has since become one of OpenAI’s most vocal critics, accusing the company of betraying its nonprofit origins by morphing into a multibillion-dollar for-profit venture aligned with Microsoft.

In addition to suing OpenAI, Musk has also taken aim at Apple. In August, xAI filed suit against the iPhone maker, alleging it conspired with OpenAI to stifle competition in the AI market. OpenAI, for its part, has countersued Musk, arguing that his barrage of lawsuits constitutes harassment rather than legitimate grievances.

The xAI–OpenAI clash is not unfolding in isolation. Talent raids and trade secrets lawsuits have become a defining feature of the AI industry’s rapid growth. Google’s DeepMind, for instance, has faced poaching pressure from rival labs, while Anthropic—founded by former OpenAI employees—was itself born out of a rift over how safely and profitably AI should be developed. Earlier this year, Meta sued a former research scientist, alleging he had misappropriated proprietary algorithms, while smaller AI startups have accused Big Tech players of using their recruiting pipelines to siphon off both staff and intellectual property.

The legal battles underscore the extraordinary value of human capital in AI. Unlike other tech sectors where patents or hardware dominate, much of AI’s competitive edge lies in research talent and access to training data. This has fueled aggressive recruitment wars, leading to disputes like xAI’s claim that OpenAI is systematically targeting employees with insider knowledge of Grok’s architecture and roadmap.

Analysts say the outcome of Musk’s lawsuits could carry broad implications. If xAI were to succeed in its trade secrets claims, OpenAI might face restrictions on how it builds and trains its models, potentially slowing its ability to compete with rivals like Google DeepMind, Anthropic, and Amazon-backed labs. But if the cases are dismissed, Musk risks being seen as weaponizing litigation in a corporate grudge match.

The courtroom battles also highlight a deeper rift over the direction of artificial intelligence development. Musk has long argued that AI must be developed safely and transparently, a vision he claims OpenAI abandoned when it pivoted toward commercial partnerships. OpenAI, by contrast, maintains it remains committed to responsible innovation and sees Musk’s legal onslaught as an attempt to discredit a competitor rather than safeguard the industry.

As filings stack up, the dispute underlines how the race for AI supremacy is unfolding not just in research labs and cloud servers but also in courtrooms. With Musk’s xAI and OpenAI at the center of one of the tech industry’s most personal and high-stakes legal clashes, the outcome could set important precedents for how talent, intellectual property, and competitive boundaries are managed in the artificial intelligence industry.