Good morning and welcome back to another edition of The AI Rundown by Lightscape Partners.

• DeepSeek continues to pressure U.S. players with the release of its upgraded V3 model, posting strong benchmarks in reasoning and code generation. The China-based firm is positioning itself as a serious challenger to OpenAI and Anthropic, offering performance at lower cost.

• OpenAI and Meta are in talks with Reliance Industries to bring their AI models to India, including local deployment and distribution via Jio. The partnership could anchor AI services in one of the fastest-growing digital markets—and potentially house models in what's being billed as the world’s largest data center.

• Cerebras Systems' IPO is on hold as U.S. regulators review a $335M investment from Abu Dhabi’s G42. The scrutiny reflects growing concern over foreign ties—particularly with China—when it comes to strategic AI chipmakers.

Stay tuned as we explore these stories and their implications for the future of AI, technology, and innovation.

If you haven’t yet, please support the newsletter by subscribing!

Hardware + Software

Cerebras Systems faces IPO delays due to national security review of G42 investment. Link.

• The $335 million investment by Abu Dhabi's G42 into Cerebras is under review by the Committee on Foreign Investment in the United States (CFIUS), delaying the AI chipmaker's IPO.

• G42's past ties to China's Huawei have led to intense scrutiny, despite the company's national security agreement with the U.S. to sever these ties.

• The Biden administration had ramped up CFIUS enforcement, and such focus on foreign investments, especially those connected with China, continues under Trump.

• Cerebras remains optimistic about eventually moving forward with their IPO, even as the review process continues to unfold.

Oracle introduces AI Agent Studio to enhance business process automation. Link.

• Oracle's AI Agent Studio is designed to orchestrate AI agents within its Fusion Applications, aiming to streamline business processes across various disciplines on Oracle Cloud Infrastructure.

• The AI Agent Studio offers LLM customizations, allowing integration of LLMs like Llama and Cohere, or external industry-specific models for specialised use cases.

• Oracle reports the ability to modify AI agents to meet business-specific needs, exemplified by extending tasks in recruiting processes, demonstrating higher flexibility and customization.

• Executive insights suggest agentic architectures are expected to become mainstream by 2025, with Oracle’s step-by-step AI approach positioned to fast-track AI integration and automation for enterprises.

Models

DeepSeek intensifies AI competition with major V3 model upgrade. Link.

• DeepSeek's V3-0324 model, available on Hugging Face, demonstrates advancements in reasoning and coding compared to its earlier version.

• Release marks DeepSeek's strategic push in the competitive AI market, positioning it against OpenAI and Anthropic.

• Benchmark tests reveal significant performance improvements across multiple technical metrics.

• Since December, DeepSeek has consistently introduced models that challenge Western offerings while maintaining lower operational costs.

Enterprise AI Applications

Kyndryl partners with Google Cloud for AI-powered mainframe modernization. Link.

• Kyndryl introduces the Mainframe Modernization with Gen AI Accelerator Program with Google Cloud to facilitate data migration from mainframes to the cloud using generative AI.

• Google's Gemini LLMs, integrated with Google Cloud, will assist in mainframe data and cloud resources integration, enhancing application development capabilities.

• The initiative offers AI-driven code analysis and conversion from COBOL to Java, aiming to reduce modernization costs and time while integrating with modern DevOps practices.

• Research indicates a growing trend of enterprises leveraging generative AI for mainframe integration, with 86% planning to deploy Gen AI tools in mainframe environments.

Content Creation

Publicis Groupe teams up with Adobe to enhance creative and personalized content at scale. Link.

• Publicis Groupe and Adobe expand their partnership by integrating Adobe Firefly generative AI into Publicis’ CoreAI system.

• This collaboration aims to scale personalized content production and improve engagement through Publicis Groupe’s identity solutions.

• By combining Adobe's creative capabilities with Publicis Sapient expertise, the partnership will enhance digital business transformation for clients.

• Publicis will utilize Firefly's generative AI models for image, vector, and video creation to offer speed and creative control, driving efficiency and innovation.

Startup Funding & Valuations.

CoreWeave reduces IPO amidst market skepticism and Microsoft dependency. Link.

• CoreWeave plans to cut its U.S. IPO size by 23.5%, aiming to sell 37.5 million shares at $40 each, down from the initial $47-$55 range.

• Investor concerns center around CoreWeave's reliance on Microsoft and the capital-intensive nature of its business model.

• Nvidia's $250 million anchor order supports the IPO, valuing CoreWeave at approximately $23 billion.

• The offering's reception signals a reevaluation of AI infrastructure valuations amid broader market uncertainties.

Nexthop AI secures $110M to advance AI infrastructure for hyperscalers. Link.

• Nexthop AI emerges from stealth with $110 million funding led by Lightspeed Venture Partners, alongside Kleiner Perkins and others.

• The company aims to meet the extensive scale requirements of AI by enhancing networking solutions for cloud and hyperscale AI clusters.

• Customized solutions include networking hardware and a hardened network operating system, integrating smoothly into hyperscalers' existing infrastructure.

• Nexthop AI positions itself as a critical partner for major cloud providers, addressing the $35 billion market for AI infrastructure.

Navina secures $55M Series C funding to advance AI in healthcare. Link.

• Navina, an AI clinical intelligence leader, raised $55 million in a Series C round led by Goldman Sachs, totaling their funds to $100 million.

• The investment will drive the U.S. expansion and technology enhancement, aiming to improve proactive patient care outcomes.

• Navina's platform is utilized by over 10,000 professionals across 1,300 clinics, demonstrating significant integration into value-based care systems.

• The funding supports Navina’s goal to automate workflows, optimize healthcare delivery, and solidify its standing as a critical healthcare AI provider.

Former Meta executives launch AI assistant startup with $15 million backing. Link.

• Yutori, founded by ex-Meta AI executives, raised $15 million to develop AI personal assistants, funding led by Rob Toews at Radical Ventures.

• Key investors include Felicis, Fei-Fei Li, and Google DeepMind's Jeff Dean, highlighting strong backing from notable industry figures.

• Yutori aims to redefine user interactions with autonomous AI, focusing on tasks like online orders and travel logistics without human oversight.

• The startup specializes in post-training models, an emerging trend important to AI reasoning models, enhancing web navigation and task efficiency.

Regulation and Legal

New York court allows New York Times copyright lawsuit against OpenAI to move forward. Link.

• A New York judge rejected OpenAI's attempt to dismiss a lawsuit by The New York Times, which accuses the company of using its content to train AI models without permission.

• Despite narrowing the scope, the core copyright infringement claims will proceed, potentially impacting OpenAI and other AI firms relying on similar data-scraping practices.

• OpenAI argues the data usage falls under 'fair use,' a critical legal concept that allows the use of copyrighted material without consent, though interpretations of its limits are still debated.

• Should OpenAI's claims of fair use fail, they could face hefty fines, and the case may set precedents affecting the use of copyrighted content in AI training.

OpenAI

OpenAI and Meta explore AI partnerships with India's Reliance Industries. Link.

• OpenAI and Meta are in talks with Reliance Industries to expand AI offerings in India, focusing on partnerships with Reliance Jio and API distribution.

• Discussions involve Reliance selling OpenAI's models to enterprise customers and locally hosting these models to retain data within India.

• A proposed partnership also includes a plan to operate AI models in a new, massive data center in Jamnagar, touted as the world's largest.

• OpenAI is considering reducing ChatGPT subscription prices significantly, though it's unclear if this has been discussed with Reliance.

Thank you for reading the AI Rundown by Lightscape Partners. Please send any questions, comments, or suggestions to [email protected].