Jen-Hsun Huang personally names! A look at the 6 promising AI startups to watch in 2025.

NVIDIA CEO Jen-Hsun Huang (Jensen Haung) during an interview with Citadel Securities ( on 10/15 personally named six highly potential AI startups, believing they will become the key force in the era of “human-machine collaboration”. He emphasized that future corporate work teams will be composed of humans and AI assistants working together, which is not only a technological trend but has already been implemented internally at NVIDIA.

The six AI startups personally named by Jen-Hsun Huang?

Jen-Hsun Huang was the first to present his vision of the future workplace and personally named six AI startups with significant development potential. He stated:

“The future work teams of enterprises will be a combination of humans and AI assistants. Some will be based on OpenAI, while others will come from Harvey, OpenEvidence, Cursor, Replit, or Lovable.”

These six startups each represent different areas of AI applications:

OpenAI: Claimed to be the most comprehensive large language model.

Harvey: An AI startup in the legal tech field.

OpenEvidence: Focused on AI medical assistance analysis.

Replit, Cursor, Lovable: Focused on AI collaborative programming and “Vibe Coding” ) tools.

Six major characteristics of AI startups and complementary observations

The most comprehensive large language model leader: OpenAI

Founded in 2015, OpenAI was established by Sam Altman, Elon Musk, and others, with the goal of creating artificial general intelligence (AGI) that is “beneficial to all of humanity” (. The company operates under a dual structure of non-profit and “capped-profit subsidiary,” balancing mission with funding flexibility. Microsoft is its main partner and investor, providing support through Azure cloud computing.

The core product is the GPT series of language models, updated from GPT-3 to the currently promoted GPT-5. In recent years, OpenAI has been actively developing multimodal technology, allowing the models to understand text, speech, and images. At the same time, ChatGPT has also added “Agent” and “Deep Research” functionalities, supporting segmented reasoning and data integration.

OpenAI's mission is to ensure that the development of AGI is beneficial to humanity, but it currently faces challenges related to safety, ethics, and commercialization. Recently, it has been reported that they are developing their own AI chips to reduce reliance on a single hardware supplier and enhance long-term computational autonomy.

Focus on the potential of legal field AI startups: Harvey

The AI startup Harvey, focused on the legal field, was established in 2022 and launched by Counsel AI Corporation, an AI legal technology company based in San Francisco, USA. The founders include talents from OpenAI, DeepMind, and top law firms like Allen & Overy ). Its main clients include law firms, corporate legal departments, and professional legal advisory teams.

It applies large language models to legal workflows, assisting lawyers in case searches, contract analysis and drafting, clause risk alerts, and automatically generating legal opinions ( still requiring professional review ). Harvey can also integrate with commonly used tools such as Word, document libraries, and internal databases, significantly enhancing the automation level of legal operations.

Since its launch, Harvey has completed multiple rounds of financing and rapidly grown to become one of the representative AI platforms in the legal tech field. Although it initially relied mainly on OpenAI models for operation, it has now actively incorporated various foundational models from Anthropic, Google, and others to enhance flexibility and stability.

Its greatest advantage lies in its deep cultivation of legal expertise and focus on practical applications. However, it also faces competitive pressure from the rapid advancements of general large models, as well as the high demands for accuracy and compliance responsibilities in the legal industry.

AI Medical Assisted Analysis: OpenEvidence

Founded in 2022, OpenEvidence was established by entrepreneur Daniel Nadler, who comes from the financial AI sector, and natural language processing expert Zachary Ziegler. It is an AI company focused on healthcare and clinical decision support. Its core product is a medical search and knowledge engine that can quickly organize key points from a vast number of medical journals, clinical studies, and papers, assisting physicians in finding reliable research and making judgments.

The platform can structure and categorize data, and respond to physicians' questions in natural language. Its model even scored over 90% in the United States Medical Licensing Examination (USMLE), demonstrating powerful medical understanding. The advanced feature “DeepConsult” can integrate and reason across multiple papers, providing a more comprehensive clinical analysis.

By 2025, OpenEvidence will have hundreds of thousands of registered physician users, millions of queries per month, and has completed multiple rounds of financing, with a valuation reaching billions of dollars. Its greatest advantage is the combination of high-precision AI and professional medical knowledge, but it also faces challenges such as data credibility, physician trust, system integration, and medical privacy regulation.

A cloud AI development platform that emphasizes “Anyone can code”: Replit

A browser-based online integrated development environment (IDE) Replit, founded in 2016 by former Meta engineer Amjad Masad, is based on the core idea of enabling anyone to write code, run it, and collaborate in real-time on any device. Users do not need to install a development environment; they can start coding just by opening a browser.

In terms of AI capabilities, Replit launched the built-in AI assistant “Ghostwriter” in the IDE in 2022, supporting code generation, automatic fixing, and explanation functions. In 2024, Replit Agent v1 was released, capable of automatically constructing applications based on natural language descriptions of requirements. By 2025, it upgraded to the latest version, Agent 3, which features cross-file understanding, automatic error handling, and strategy adjustment capabilities. The platform also introduced the concept of “Vibe coding,” allowing developers to collaborate with AI in programming using natural language.

The advantages of Replit lie in its zero-configuration and quick start, making it very suitable for prototyping, teaching, and rapidly validating market ideas. However, it may be limited in large projects or high-performance demands, and AI-generated code still requires manual review to maintain quality. Replit also completed a round of financing of $250 million in September this year, raising its valuation to approximately $3 billion.

The AI programming partner for professional developers: Cursor

Cursor is an AI programming assistant designed for professional developers, capable of providing real-time assistance in writing, modifying, and understanding code within familiar development environments like VS Code (. It extends the concept of “Vibe coding,” allowing engineers to collaborate with AI using natural language commands to complete programming tasks.

The core features of Cursor are high integration, not disrupting the existing development process, and providing instant code suggestions, error corrections, refactoring recommendations, and code annotations. Jen-Hsun Huang also revealed in this interview that his engineering team has already adopted Cursor in their daily development.

The advantage of Cursor lies in its ability to delve into professional development processes, enhancing productivity and programming efficiency. However, challenges include the security of generated code, contextual understanding capabilities with large libraries, as well as integration and reliability. Cursor's parent company, Anysphere, raised 900 million dollars in June 2025, currently valued at approximately 9.9 billion dollars.

A one-stop AI platform from natural language to online applications: Lovable

Lovable is a startup from Sweden that centers around the core concept of “Vibe coding”. Users can describe the desired functionality in natural language, and the AI can automatically generate code, structure, and deploy it as an application. It integrates an editor, backend system, and deployment tools, achieving a seamless process from idea to launch.

Lovable has experienced astonishing growth since it opened to general users in November 2024, completing a Series A funding round of 200 million USD in July this year, with a valuation of approximately 1.8 billion USD. Lovable is suitable for quickly building web applications, prototypes, and internal tools, providing instant previews and AI collaborative editing features, allowing users to switch freely between manual and automated processes.

Its advantages are very high development efficiency and low barriers to entry, but when facing complex systems and high-performance requirements, stability and maintainability still pose challenges.

) The latest five mainstream AI language models ( LLM ) comprehensive analysis, understand payment, applications, and security all at once (

This article was personally named by Jen-Hsun Huang! An overview of the 6 most anticipated AI potential startups in 2025 first appeared in Chain News ABMedia.