A recent TechCrunch report highlights the rise of 15 AI startups founded by ex-OpenAI employees, rapidly gaining traction in Silicon Valley. This burgeoning network, echoing the “PayPal Mafia,” is showcasing cutting-edge technologies with the potential to achieve OpenAI-level breakthroughs.
The original “PayPal Mafia” significantly impacted the internet landscape. Members like LinkedIn’s Reid Hoffman, YouTube’s Steve Chen and Chad Hurley, and Tesla and SpaceX’s Elon Musk drove transformative changes across social networking, online video, electric vehicles, commercial space travel, and big data analytics. They created trillion-dollar empires and established a legendary “alumni network” model in Silicon Valley.
A similar phenomenon is now emerging in artificial intelligence. OpenAI, co-founded by Musk, Sam Altman, and Greg Brockman in 2015, witnessed its AI chatbot ChatGPT’s rapid rise in 2019. By late March 2025, OpenAI’s valuation reached $300 billion (approximately ¥2,186.3 billion). Currently, the combined valuation of the 15 AI startups launched by ex-OpenAI staff is roughly $250 billion (around ¥1,821.9 billion), equivalent to recreating 80% of OpenAI. These companies span various cutting-edge fields, including large language models, AI Agents, robotics, and biotechnology, with some, like Perplexity, challenging Google’s search engine dominance.
Anthropic: Creator of Claude and Major OpenAI Competitor
In 2021, Dario Amodei and his sister Daniela Amodei departed OpenAI to establish Anthropic, an AI safety company based in San Francisco. In 2024, OpenAI co-founder John Schulman joined Anthropic, committed to developing “safe general artificial intelligence (AGI).”
In February 2025, Anthropic launched Claude 3.7 Sonnet, its most intelligent model to date and the first mixed-reasoning model on the market. This model provides near-instant responses while offering step-by-step reasoning processes that users can visualize. API users can precisely control the model’s thinking time. Claude 3.7 Sonnet excels in programming and front-end development.
Last year, Anthropic partnered with Menlo on Anthology, a program where venture firms invest in young AI startups from a $100 million fund. These startups gain access to Anthropic’s models, staff, and $25,000 (approximately ¥180,000) in computing credits. This month, Anthropic made its first investment in Goodfire, a one-year-old startup focused on helping AI developers understand the inner workings of their AI models. Menlo Ventures led the round, with participation from Anthropic’s other investors, Lightspeed Venture Partners and B Capital.
According to The Information, OpenAI’s revenue ($3.7 billion, or ¥27 billion) still exceeds Anthropic’s ($1 billion, or ¥7.3 billion) by more than threefold. However, Anthropic’s valuation reached $61.5 billion (approximately ¥448.2 billion) by March 2025, positioning it as a significant OpenAI competitor. Anthropic’s focus on AI safety distinguishes it from other AI developers, and its Anthology program demonstrates a commitment to fostering the next generation of AI startups. The launch of Claude 3.7 Sonnet further solidifies Anthropic’s position as a leading innovator in the AI space, pushing the boundaries of what’s possible with large language models. The model’s enhanced reasoning abilities and programming prowess make it a valuable tool for developers across various industries. As Anthropic continues to develop and refine its AI models, it will be interesting to see how it impacts the competitive landscape and shapes the future of AI.
Safe Superintelligence: Focused on Secure Superintelligence
In May 2024, former OpenAI chief scientist Ilya Sutskever left the company after an unsuccessful attempt to oust CEO Sam Altman. He then founded Safe Superintelligence (SSI) with the singular goal of focusing on safe superintelligence. The company adopts a technical approach that prioritizes several years of research and development before launching products. Sutskever’s departure and the founding of SSI underscore the growing concerns about the potential risks associated with advanced AI. By focusing solely on safety, SSI aims to mitigate these risks and ensure that superintelligence benefits humanity.
In September 2024, just three months after its founding, SSI announced a $1 billion (approximately ¥7.3 billion) funding round, valuing the company at $5 billion (approximately ¥36.4 billion). Investors included Andreessen Horowitz, Sequoia Capital, DST Global, SV Angel, and NFDG, an investment partnership managed by Nat Friedman and SSI CEO Daniel Gross. By 2025, SSI’s valuation had reached $32 billion (approximately ¥233.2 billion) despite having no products or revenue. The rapid valuation increase reflects the strong investor confidence in Sutskever’s vision and the growing importance of AI safety. The company’s commitment to long-term research and development suggests a deep understanding of the complexities involved in creating safe superintelligence. While SSI’s lack of products may raise some eyebrows, its focus on foundational research could lead to breakthroughs that have a profound impact on the future of AI.
Thinking Machines Lab: Empowering Everyone with Knowledge and Tools
Thinking Machines Lab is an AI research and product company committed to providing everyone with the knowledge and tools they need, ensuring AI serves their personalized needs and goals. Mira Murati, former OpenAI CTO, founded the company in 2024, and it publicly launched in February 2025, declaring its intention to develop “customizable, high-performance” AI systems. Murati’s departure from OpenAI and the founding of Thinking Machines Lab highlight the diverse paths that former employees are taking to shape the future of AI. Thinking Machines Lab’s focus on empowering individuals with AI tools suggests a belief in the democratizing potential of AI.
Initially, the Thinking Machines Lab team comprised approximately 30 core members, two-thirds of whom came from OpenAI. Subsequently, more former OpenAI executives joined as advisors, including former OpenAI chief research officer Bob McGrew and Alec Radford, the lead researcher behind several flagship AI models. The high concentration of former OpenAI employees underscores the company’s technical expertise and its ability to attract top talent.
Reports indicate that Thinking Machines Lab is seeking $2 billion (approximately ¥14.6 billion) in seed funding, valuing the company at no less than $10 billion (approximately ¥73 billion). If successful, it would become one of the world’s most valuable AI startups, despite being established only this year and having no revenue or products. The company aims to compete in the fiercely competitive generative AI model development landscape. The ambition to secure such a large seed funding round reflects the company’s confidence in its ability to disrupt the AI market. If successful, Thinking Machines Lab could become a major player in the generative AI space.
Perplexity: An AI Search Engine Powered by Large Language Models
Aravind Srinivas founded Perplexity in 2022 after working at OpenAI for a year. Perplexity is an AI search engine based on large language models that generates responses by integrating web search results. Perplexity’s AI-powered search engine offers a unique approach to information retrieval, providing users with concise and relevant answers based on web search results. This contrasts with traditional search engines that simply provide a list of links.
According to The Verge, Srinivas stated that Perplexity plans to release its own browser, named Comet, next month. “We’re building a browser because it’s probably the best way to build an agent,” he said. “A browser is essentially a containerized operating system. If you’re already logged into other third-party services, it can let you access those services through hidden tabs, scrape pages on the client-side, and reason and take actions on your behalf.” The development of a dedicated browser highlights Perplexity’s ambition to create a comprehensive AI-powered platform. The browser’s capabilities suggest a focus on enabling users to automate tasks and access information seamlessly.
Earlier reports suggested that Perplexity was interested in acquiring Chrome if it were to separate from Google. The company’s AI search engine has begun entering into pre-installation agreements with Android phone manufacturers. This week, Motorola announced that it would pre-install Perplexity on its new Razr phones, potentially exposing Srinivas’s AI “answer engine” to millions of new users. Although Srinivas stated that the current pre-installation depth did not meet their expectations, it represents a significant breakthrough for a startup like Perplexity. The pre-installation agreements demonstrate Perplexity’s growing influence and its ability to compete with established tech giants.
Perplexity, which has received investment from Jeff Bezos and others, has faced controversy over alleged illegal web data scraping. In March 2025, the company launched a $1 billion (approximately ¥7.3 billion) funding round with a valuation of $18 billion. The controversy surrounding data scraping underscores the ethical and legal challenges that AI companies face.
xAI: The Company Behind the Grok AI Chatbot
Kyle Kosic left OpenAI in 2023 to join Elon Musk’s AI company, xAI, where he led the development of the Grok chatbot. He returned to OpenAI in 2024. Kosic’s journey highlights the interconnectedness of the AI ecosystem and the movement of talent between different companies.
In February, xAI launched the new generation Grok 3 model, its most advanced version to date, combining powerful reasoning abilities with extensive pre-trained knowledge. The model was trained on its Colossus supercomputing cluster, utilizing ten times the computing resources of previous top models. Grok 3 represents significant improvements in reasoning, mathematics, programming, common sense, and instruction execution. The launch of Grok 3 demonstrates xAI’s commitment to developing state-of-the-art AI models. The model’s enhanced capabilities position it as a strong competitor in the AI chatbot market.
In late March, Musk announced that xAI would acquire the X platform in an all-stock transaction, forming a new holding company, xAI Holdings. Post-merger, xAI was valued at $80 billion (approximately ¥583 billion), and X was valued at $33 billion (excluding debt) (approximately ¥240.4 billion). xAI Holdings has become one of the world’s most valuable AI startups with a valuation in the hundreds of billions. While the transaction model has raised questions, it is seen as an important move to capitalize on the Musk ecosystem. The acquisition of X and the formation of xAI Holdings represent a bold move by Musk to consolidate his AI efforts and create a powerful platform for innovation.
On April 27, Bloomberg reported that Musk’s xAI Holdings is in talks for a new funding round of $20 billion (approximately ¥145.7 billion). If successful, this would be the second-largest startup funding round in history, only surpassed by OpenAI’s $40 billion (approximately ¥291.5 billion) funding round last month. The move would value the company at over $120 billion (approximately ¥874.5 billion). The potential funding round underscores the immense investor interest in xAI and its potential to disrupt the AI market.
Stem AI: Still in Stealth Mode
TechCrunch reported in 2024 that former Twitch CEO and OpenAI interim CEO Emmett Shear is secretly developing an AI startup called Stem AI (disclosed in 2024). Shear’s involvement in Stem AI suggests a focus on addressing the ethical and societal implications of AI.
Market analysis suggests that Stem AI is likely focused on addressing the “AI alignment” challenge that Shear has repeatedly mentioned. Its team background suggests that the startup combines AI safety, bioethics, and internet product genes, potentially paving a new path for AI development. The combination of AI safety, bioethics, and internet product development suggests a holistic approach to AI that considers both its technical capabilities and its impact on society.
Project details and funding scale have not been disclosed, but it has received support from Andreessen Horowitz, a top Silicon Valley venture capital firm. The backing of Andreessen Horowitz provides Stem AI with valuable resources and expertise.
Eureka Labs: Dedicated to AI Teaching Assistants
Andrej Karpathy, a computer vision expert and founding member of OpenAI, joined Tesla in 2017 to lead its autonomous driving project. He left in 2024 to found Eureka Labs, an education technology company based in San Francisco, focused on developing AI teaching assistants. Karpathy’s expertise in computer vision and autonomous driving provides Eureka Labs with a strong foundation for developing intelligent AI teaching assistants.
Eureka Labs’ core philosophy is to create an “AI-native” education platform, deeply integrating AI technology into all aspects of education rather than simply adding AI as an auxiliary tool to existing educational models. Their vision is to revolutionize traditional learning methods by creating AI teaching assistants. They believe that AI can significantly improve learning efficiency and experience through intelligent personalized guidance and feedback. The focus on creating an “AI-native” education platform suggests a belief in the transformative potential of AI in education.
Currently, Eureka Labs has launched its first product, an undergraduate-level course called LLM101n, which aims to teach students how to train their own large language models (LLMs). The launch of LLM101n demonstrates Eureka Labs’ commitment to providing students with the skills they need to succeed in the AI era.
To date, specific information regarding Eureka Labs’ funding scale and investors has not been disclosed.
Pilot: An Accounting Services Company
Jeff Arnold founded Pilot, an accounting services company, in 2017 after serving as an operations manager at OpenAI for five months in 2016. Arnold’s experience at OpenAI likely provided him with valuable insights into the challenges faced by high-growth tech startups.
Pilot.com provides bookkeeping, tax, and CFO services to high-growth tech startups and small businesses. Founded in 2017 by Jeff Arnold, Waseem Daher, and Jessica McKellar, who met at MIT. Pilot.com is their third startup together, with their previous two companies, Ksplice and Zulip, being acquired by Oracle and Dropbox, respectively. The founders’ previous success in building and selling companies provides Pilot with a strong foundation for growth.
Pilot focuses on providing financial services to startups, securing $100 million (approximately ¥730 million) in Series C funding in 2021, valuing the company at $1.2 billion (approximately ¥8.7 billion). Investors include Sequoia Capital, Index Ventures, and Stripe. The company’s focus on startups and its strong investor base position it for continued success.
Adept AI Labs: Focusing on Human-Machine Collaborative Work
David Luan left OpenAI in 2020, briefly worked at Google, and then founded Adept AI Labs in 2021 to develop enterprise-level AI tools. Luan’s experience at OpenAI and Google provided him with valuable expertise in AI research and development.
Adept AI was founded in 2022 by David Luan (CEO), Niki Parmar (CTO), and Ashish Vaswani (Chief Scientist), all of whom have significant AI research and development experience at companies such as Google (including Google Brain) and OpenAI. Notably, Parmar and Vaswani are pioneers of the Transformer architecture. The founders’ expertise in AI and the Transformer architecture provides Adept AI with a significant competitive advantage.
Adept AI Labs is a machine learning research and product lab focused on building general intelligence systems, aiming to enable humans and computers to work together creatively. The focus on human-machine collaboration suggests a belief in the potential of AI to augment human capabilities.
The company has received investment from General Catalyst, Spark Capital, Greylock Partners, Microsoft, Nvidia, and Andrej Karpathy. In 2023, it raised $350 million (approximately ¥2.55 billion), valuing the company at over $1 billion (approximately ¥7.3 billion). Luan left in late 2024 to join Amazon’s AI agent lab (Amazon has acquired Adept’s core team). The acquisition of Adept’s core team by Amazon demonstrates the company’s talent and its potential to contribute to Amazon’s AI efforts.
Cresta: An AI Customer Service Company
Tim Shi joined OpenAI in 2017, focusing on safe AGI research, and left a year later to found Cresta, an AI customer service center company. Shi’s focus on safe AGI research at OpenAI suggests a commitment to responsible AI development.
Cresta’s primary business is to provide an end-to-end generative AI platform for contact centers. The platform enhances customer interaction, improves the efficiency and performance of human customer service representatives, and ultimately helps companies optimize operating costs, increase revenue, and gain a deeper understanding of customer needs through its core products: Cresta AI Agent (AI virtual customer service), Agent Assist (real-time customer service assistance), Conversation Intelligence (AI Analyst), and AI-driven quality management and coaching. Cresta’s platform offers a comprehensive suite of AI-powered tools for improving customer service.
Headquartered in San Francisco, Cresta has raised $276 million (approximately ¥2 billion) through seven funding rounds. The most recent round was a $125 million (approximately ¥900 million) Series D funding in November 2024, co-led by World Innovation Lab (WiL) and the Qatar Investment Authority (QIA). As of March 2022, Cresta was valued at $1.6 billion (approximately ¥11.7 billion), with the valuation of the November 2024 Series D funding undisclosed. The company’s strong funding and valuation demonstrate its success in the AI customer service market.
Covariant: Warehouse Automation
Pieter Abbeel, Peter Chen, and Rocky Duan worked as research scientists at OpenAI from 2016 to 2017 and founded Covariant, a robotics foundation model company, in 2017. The founders’ experience at OpenAI provided them with valuable expertise in AI and robotics.
Covariant develops a general robotic intelligence platform primarily used in logistics and warehousing. The platform combines advanced AI technologies such as deep learning, imitation learning, and reinforcement learning. Their goal is to enable robots to observe, learn, and adapt to new tasks and environments without extensive pre-programming. Covariant’s platform offers a flexible and adaptable solution for warehouse automation.
To date, they have raised over $222 million in funding, with investors including Index Ventures, Radical Ventures, Coatue, and the Canada Pension Plan Investment Board. The most recent round was a $75 million (approximately ¥550 million) C-2 funding in May 2023.
In 2024, Amazon acquired all of Covariant’s founders and a quarter of its employees, a deal seen as a typical example of tech giants circumventing antitrust scrutiny. The acquisition of Covariant’s founders and employees by Amazon demonstrates the company’s talent and its potential to contribute to Amazon’s robotics efforts.
Living Carbon: Using Technology to Protect the Environment
Maddie Hall left OpenAI in 2019 to found Living Carbon, a climate technology company that develops genetically engineered plants to enhance carbon capture. Hall’s decision to leave OpenAI and focus on climate technology highlights the growing interest in using AI to address environmental challenges.
Headquartered in Hayward, California, Living Carbon’s core business is using synthetic biology and genetic engineering to develop plants, particularly trees, that can more effectively capture and store carbon dioxide. Living Carbon’s technology offers a novel approach to carbon sequestration.
To date, the company has raised at least $36.1 million (approximately ¥260 million), with investors including Temasek, Lowercarbon Capital, Toyota Ventures, and Felicis Ventures. The most recent public funding round was a $21 million (approximately ¥150 million) Series A funding in January 2023.
Prosper Robotics: Bringing Robots into Homes
Shariq Hashme worked at OpenAI for nine months in 2017, leading the development of the Dota game robot. He then joined data annotation company Scale AI and founded Prosper Robotics in 2021. Hashme’s experience at OpenAI and Scale AI provided him with valuable expertise in AI and robotics.
Based in London, Prosper Robotics focuses on the development of home robot butlers, competing with companies such as Norway’s 1X and US-based Apptronik. Prosper Robotics’ primary R&D and promotional product is “Alfie,” a bipedal wheeled humanoid robot designed to be a home assistant. Their robots may adopt a subscription or leasing model, meaning users may not need to purchase expensive robots outright but rather pay a monthly or annual fee. Prosper Robotics’ focus on home robot butlers reflects the growing interest in personal robotics.
Daedalus: Revolutionizing the Manufacturing of Precision Components
Jonas Schneider left OpenAI in 2019 to found Daedalus, an intelligent factory company for precision components, based in San Francisco. Schneider’s decision to focus on manufacturing highlights the potential of AI to transform traditional industries.
Daedalus’s primary business is using AI-powered robots and its proprietary FactoryOS software to revolutionize the manufacturing of precision parts. They focus on providing automated production solutions for industries such as automotive, medical technology, and mechanical engineering. Their self-learning robots can handle repetitive and error-prone tasks, enabling humans to focus on more critical aspects, thus enabling a human-machine collaborative production model. Daedalus’s platform offers a comprehensive solution for automating the manufacturing of precision components.
As of February 2024, Daedalus AI successfully completed a $21 million (approximately ¥150 million) Series A funding round led by NGP Capital, with existing investors Addition and Khosla Ventures also participating. The funding will be used to further develop Daedalus AI’s proprietary manufacturing AI platform and expand its production facilities in Germany.
Kindo: Focused on Enterprise-Grade AI Chatbots
Margaret Jennings worked at OpenAI from 2022 to 2023 and founded Kindo, an enterprise-grade AI chatbot company, in 2023. Jennings’s experience at OpenAI provided her with valuable insights into the development of AI chatbots.
Kindo is a company focused on providing enterprises with a secure and compliant AI management platform. The platform provides a centralized interface for accessing various AI models (private, commercial, and open-source) and integrates with over 200 SaaS applications, ensuring the security and compliance of AI-related activities. Kindo’s platform addresses the growing need for secure and compliant AI solutions in the enterprise.
As of July 2024, Kindo had raised a total of $27.6 million (approximately ¥200 million) in funding, including a $7 million (approximately ¥51.01 million) seed round in September 2023 and a $20.6 million (approximately ¥150 million) Series A round in July 2024. Its investors include Riot Ventures, Eniac Ventures, and Drive Capital. Notably, Kindo acquired WhiteRabbitNeo, an open-source secure AI model project, at the same time as the Series A funding to enhance its security capabilities. Jennings left the same year to join French AI company Mistral, responsible for product and R&D.
This AI behemoth, valued at $300 billion (approximately ¥2,186.3 billion), is spurring a new generation of tech startups through a unique talent spillover effect. The combined valuation of these 15 companies is approximately $250 billion (approximately ¥1,821.9 billion), equivalent to recreating 80% of OpenAI.
Founded by technical experts, these companies cover the entire industry chain from basic research to industry applications, continuing OpenAI’s technical genes while exploring differentiated development paths. The enthusiastic pursuit of capital markets reflects the industry’s optimistic expectations for the future of AI technology.
As Pat Grady, a partner at Sequoia Capital, put it, ‘We are witnessing the ‘Bell Labs phenomenon’ in the AI field, where a top research institution spawns an entire industrial ecosystem.’ In the long run, this ‘Bell Labs phenomenon’ effect helps form a more diversified industrial ecosystem and may also foreshadow a more competitive market landscape. Whether it can continue to produce breakthrough innovations remains to be seen, depending on the alignment between corporate technology evolution and market demand. The success of these companies will depend on their ability to innovate, adapt to changing market conditions, and address the ethical and societal implications of AI.