The Summit of Waves and the Weight of Reality: AIGC/LLM Big Model Application and Unicorn Inventory and Entrepreneurship Outlook 💪
In this article, we explore the current landscape of AIGC/LLM big model applications and unicorns, their significance, and potential entrepreneurial opportunities. 📚
Introduction 📝
The super cycle that combined the development of mobile internet, Chinese economic growth, and global economic globalization has come to an end. This marks a significant turning point in the TMT industry. LLM big models have emerged, igniting the field of AIGC entrepreneurship, and big model applications have taken the lead in the wave of innovation. 🔥
The Challenges in Big Model Entrepreneurship 😔
Big model entrepreneurship is not without its limitations and challenges. Data has become the core element of future business competition, similar to land and oil in the past. Companies that cannot generate or store data lack barriers, while those that cannot directly apply data to create business value will be disrupted by the extension of existing businesses by tech giants. As the importance of data becomes consensual, data security, privacy, protection, transmission, and transactions become new hot topics. Companies that do not possess competitive advantages in data creation, protection, or application will be marginalized and eliminated. Most big model startups driven by capital will disappear as the industry chain matures. The data competition in the big model field will start in the consumer market and gradually expand into finance, education, healthcare, and more. The players involved will also continue to evolve, from entrepreneurs and giants to capital, government, and countries. 🚀
Valuation of AIGC/LLM Big Model Unicorn Companies 🎓
Let’s take a look at some of the highest-valued AIGC/LLM big model unicorn companies and analyze their businesses, people, and future entrepreneurial opportunities. 😎
- Company Overview
Company Name Establishment Time Segment Direction Business Type Valuation (as of May 2023)
Stability AI 2020 Image Content Generation Allegedly $4 billion USD
Cohere 2019 Text Saas $2 billion USD
Jasper 2021 Text Content Marketing $1.5 billion USD
Runway 2018 Video Content Generation $1.5 billion USD
Repl.it 2016 Text Coding Assistance $1.16 billion USD
Character.AI 2021 Chatbots Content Generation $1 billion USD
Glean 2019 Search Saas $1 billion USD
2. Company Details
The analyzed companies include Stability AI, Cohere, Jasper, Runway, Repl.it, Character.AI, and Glean.
As startups in the industry are rapidly evolving, we focus more on fundamental information (such as establishment time, location, segment direction, and business type) and provide a brief description of their business and team.
In summary, at the business level, high-valued companies in the current stage primarily focus on text, image, and video domains. The application of big models is more direct and prevalent, resulting in frequent company updates. However, this also implies lower barriers to entry and competition from established industry giants. For example, Jasper competes with Office/Notion, and Runway competes with Adobe + Firefly. This intensifies competition and disrupts existing financial models. Therefore, we have a long-term positive outlook for B2B-oriented companies like Glean and Repl.it.
At the team level, a high proportion of founders have strong technical backgrounds with extensive experience in the field of big models. For instance, one of the founders of Character.AI is an author of the Transformer model, and a founder of Cohere has conducted significant research on Transformers during their time at Google Brain. These founders possess a “T-shaped” talent profile, with expertise in both technology and business domains. Many current startup companies are extensions of their previous work or entrepreneurial projects. The founders are often serial entrepreneurs with experience in launching one or more companies internally or externally, embodying the concept of innovation diffusion theory’s “innovators.” The team members mainly consist of “friends and family” from school, major companies, or previous entrepreneurial ventures, with 5-10 years of collaboration and a well-established mutual trust.
2.1 Stability AI
Founded in 2020 in the UK, Stability AI is a content generation company in the field of image processing, with an estimated valuation of $4 billion.
Business description in one sentence: The company behind Stable Diffusion has designed and implemented an open-source AI tool that generates images based on given text input.
Team description in one sentence: The CEO (born in 1983) has 13 years of experience in hedge funds and has started two ventures in 2015 and 2019.
Stable Diffusion is an AI technology model that generates images based on text input. It can generate high-resolution and high-quality images within seconds while maintaining realism and artistic qualities, similar to the DALL-E2 system. Unlike established companies like DALL-E2 and OpenAI, Stable Diffusion allows anyone to use and build models in an unsupervised manner.
The open-source code of Stable Diffusion enables developers to bypass data limitations and achieve functionalities that are difficult to implement on other platforms. This means that anyone can view, run, and modify the code, and even use the software for their own commercial products. Additionally, the Stable Diffusion platform allows the use of celebrity portraits and sensitive images that are prohibited on other platforms.
CEO Emad Mostaque (born in 1983) graduated from Oxford in 2005 with a degree in Mathematics and Computer Science. He has 13 years of experience in hedge funds. In 2015, he founded Ananas Foundation, a company that aims to combat extremism and create richer community education resources, which has now become a Web3 company. In 2019, he co-founded Symmitree.
2.2 Cohere
Founded in 2019 in Canada, Cohere is a Saas company in the field of text processing, with an estimated valuation of $2 billion.
Business description in one sentence: A CRM platform targeting developers and businesses.
Team description in one sentence: The CEO (born in 1995) has worked with Jeff Dean and the Google Brain team on Transformer visual understanding.
Cohere was founded in 2019 by former researchers from Alphabet, Google’s parent company. With a strong research background and close connections to Google, Cohere quickly rose to prominence and focused on training natural language processing (NLP) models. It competes with OpenAI, Anthropic, and other companies as a provider of foundational models. Cohere plans to launch a dialogue model similar to ChatGPT but with a focus on meeting the needs of developers and enterprise users. Users can generate text and optimize output by interacting with the model.
CEO Aidan Gomez (born in 1995) graduated from the University of Toronto in 2018 and dropped out of a Ph.D. program at Oxford in the same year. In 2018, he collaborated with Geoffery Hinton on research related to knowledge distillation and initialization. In 2019, he worked with Jakob Uszkoreit, Jeff Dean, and the Google Brain team in Berlin on Transformer’s visual understanding.
CTO Ivan Zhang (born in 1996) joined Ranomics in 2016 after dropping out of the University of Toronto. He has worked as an engineer in companies such as Pressly, Cortex, and FOR.ai.
During the Transformer project, Aidan participated as an intern and made contributions in the software aspect. When he joined, there were already senior researchers within Google Brain, including Ashish, Noam, and Jakob, dedicated to advancing the application of autoregressive models in the text domain. While autoregressive models were popular in the speech domain, they had not been applied in the text domain. The main work on Transformer was completed within three months, during which team members lived in the office. In an interview, Aidan mentioned that Transformer was his first involvement in formal academic research, and he had thought that all academic research was of such high intensity. One night at 3 am, they submitted the Transformer paper to the NeurIPS conference, and afterward, Ashish said to Aidan, “This will be a big change.” At that time, Aidan did not realize the impact Transformer would have in the future, but as the first author of the paper, Ashish understood the significant influence this model would have on artificial intelligence.
2.3 Jasper
Founded in the United States in 2021, Jasper is a marketing-oriented company in the text field, with a valuation of $1.5 billion.
One-sentence business description: A leading AI marketing tool and writing assistant that helps teams create branded custom content quickly.
One-sentence team description: The CEO (born in 1989) and a three-person founding team started two ventures in 2014 and 2017, with a deep bond between them.
Jasper utilizes artificial intelligence technology to provide solutions for rapid ad, blog, and social media article writing. Starting as an ad tool, it has evolved into a comprehensive AI content platform with 60+ templates covering ads, slogans, webpages, emails, blogs, and social media, among other scenarios. Integration with the Grammarly tool helps users check for plagiarism and fix errors, enhancing content quality. The Chrome extension launched in 2022 allows users to leverage AI writing capabilities on any website, enabling them to quickly create various types of content, including social media posts, emails, and blogs, using simple commands and 60+ AI writing templates. Jasper has attracted over 100,000 users, including freelancers, small and medium-sized enterprises, and marketing departments, who utilize its AI technology to create multi-format, multi-language content.
CEO Dave Rogenmoser (born in 1989) graduated with a BBA from Kansas State University in 2011 and founded a recruitment company. He later co-founded a marketing company and a CRM company, Proof, in 2017, together with the CTO.
CTO John Philip Morgan started his entrepreneurial journey with the CEO eight years ago.
COO Chris Hull.
The friendship of Jasper’s founding team, Dave Rogenmoser, Chris Hull, and John Philip Morgan, dates back eight years. They previously founded companies such as Payfunnels and Proof, eventually coming together to start Jasper AI. Despite setbacks in applying for Y Combinator and facing transformation challenges, they remained committed to their friendship and the company’s growth. Rogenmoser continues to lead the team as CEO, while Hull and Morgan’s backgrounds laid the foundation for Jasper in B2B marketing and process automation. The efforts of this team paved the way for Jasper’s success.
2.4 Runway
Founded in the United States in 2018, Runway is a content generation company in the video field, with a valuation of $1.5 billion.
One-sentence business description: The initial version of Stable Diffusion primarily involves the company and launches AI tools for real-time video editing and collaboration in the browser.
One-sentence team description: The CEO (born in 1990) briefly stayed at the university as a teacher after
graduation and met the founding team during a project at New York University in 2016.
Runway’s image processing capabilities overlap to some extent with Jasper’s products, such as text-to-image and image-to-image generation. However, Runway has unique competitive advantages in image processing, video processing, and audio processing. In the video processing field, Runway leverages an AI tool plugin called Magic Tools, which enables video editing, green screen extraction, video restoration, motion capture, and other functionalities. Its efficiency far surpasses traditional video software like Adobe After Effects. By integrating multiple processing capabilities, Runway provides a more comprehensive set of creative tools and functionalities.
CEO Cristóbal Valenzuela Barrera (born in 1990) graduated with a Bachelor’s and Master’s degree in Design Arts from the Universidad Adolfo Ibáñez (AIU) in Chile in 2012. After graduating, he chose to stay at the university as a teacher. Starting in 2016, the progress of deep learning had a profound impact on Cristóbal Valenzuela. He quit his job and went to New York University to pursue further studies, becoming a graduate student in the Interactive Telecommunications Program (ITP) at the Tisch School of the Arts. Runway was Cristóbal Valenzuela’s thesis project. During the development of the project, he met Alejandro Matamala, a compatriot from Chile with two entrepreneurial experiences. Alejandro subsequently joined the project as a co-founder. After graduation, New York University provided them with a research internship opportunity, where they encountered experienced Chilean developer Anastasis Germanidis and successfully convinced him to join Runway as the Chief Technology Officer (CTO).
2.5 Repl.it
Founded in the United States in 2016, Repl.it is an AI-assisted programming company in the software development field, with a valuation of $1.16 billion.
One-sentence business description: A browser-based cross-platform collaborative coding integrated development environment (IDE), equivalent to VS Code + Git + Node.js (development environment) + IM + forum.
One-sentence team description: The CEO (born in 1988), his wife, and his younger brother jointly established a “family business”.
Replit positions itself as the “first fully online multiplayer coding environment” and has quickly attracted a large user base since its inception. The platform offers powerful online editing capabilities, allowing users to collaborate on editing code in real-time without the need for downloading or setting up. Users have commented that this platform encompasses almost all the essential features such as editing, version control, development environment, debugging, package management, making it an enhanced version of GitHub. Replit supports over 50 programming languages, and users can build applications and websites here using any browser and device, including mobile devices. Additionally, it provides project collaboration and sharing features, as well as access to containers for running code.
CEO Amjad Masad (born in 1988) graduated from PSUT in Jordan in 2010 with a bachelor’s degree. After graduation, he worked as a front-end developer at research institutes and Yahoo in Jordan. In 2011, he moved to New York and became a founding engineer at Codecademy. In 2013, he joined Facebook as a software engineer.
CTO Faris Masad is the CEO’s younger brother.
Design Haya Odeh (born in 1985) is the CEO’s wife. She graduated from Oman University in 2007 with a bachelor’s degree and has worked as a graphic/web designer in multiple companies.
Replit was initially founded as a family business by Amjad Masad, his wife Haya Odeh, and his younger brother Faris Masad. Haya Odeh, Amjad’s wife, is a talented designer, while his brother is an excellent programmer.
By 2019, this startup had grown to a size of six employees and one million users, supported by $5 million in seed funding over three years. Despite having a user base at that time, as Amjad’s wife answered on Quora, “Replit was not profitable,” and the platform, built on Google Cloud services, “required a certain amount of funding, and the service was not cheap.”
2.6 Character.AI
Founded in the United States in 2021, Character.AI is a content generation company in the chatbot field, with a valuation of $1 billion.
One-sentence business description: Creating advanced AI characters and engaging in conversations with them.
One-sentence team description: The CEO (born in 1976) is one of the authors of Transformer and had previously worked with the founding team on the Meena chatbot project at Google.
Character.AI is a web-based neural language model chatbot application that can generate human-like text responses and engage in contextual conversations. This beta model was developed by Noam Shazeer and Daniel De Freitas, former developers of Google LaMDA, and was opened to the public in September 2022. Users can create their own “characters,” shape their “personalities,” set specific parameters, and then publish them to the community for others to engage in conversations. These characters can be based on fictional media resources or celebrities, or they can be entirely original. Some characters are created for specific purposes, such as assisting in creative writing or as part of text-based adventure games. Users can interact with individual characters or organize group chats with multiple characters, allowing them to.
CEO Noam Shazeer (born in 1976) graduated from Duke University with an undergraduate degree in 1998. From 2000 to 2021, he worked as a software engineer at Google, serving as a Google Distinguished Engineer and one of the authors of Transformer.
President Daniel De Freitas (born in 1989) graduated from the University of São Paulo in 2011. In 2012, he joined Microsoft Bing as a software engineer and later joined Google in 2016 as a research engineer, leading research on Meena and LaMDA.
Noam Shazeer and Daniel De Freitas started collaborating on the development of the chatbot Meena while at Google. The technology they adopted was said to surpass any other technology available at the time. Meena, a conversational AI program, could confidently engage in philosophical discussions and even come up with puns on the spot. According to insiders, the researchers told their colleagues that with the advancement of artificial intelligence, chatbots would fundamentally change the way people use search engines and interact with computers. They urged Google to provide external researchers with access to Meena and attempted to integrate it into the Google Assistant smart assistant. They also requested that Google publicly demonstrate the capabilities of this chatbot, but Google executives repeatedly and clearly rejected their requests, citing concerns about the AI system’s safety and fairness. Google’s cautious approach to chatbots prompted the two researchers to leave and start their own venture.
2.7 Glean
Founded in the United States in 2019, Glean is an enterprise-focused company in the search field, with a valuation of $1 billion.
One-sentence business description: An AI work assistant that connects 100+ SaaS applications, allowing users to search enterprise data across applications.
One-sentence team description: CEO (born in 1974) previously worked as a software engineer at Microsoft and Google and co-founded Rubrik in 2014.
Glean’s cross-application search functionality acts as a layer above all SaaS products, allowing users to search enterprise data without having to individually open each SaaS application. It simplifies common tasks and has become the most frequently used and time-consuming application for many Glean users, with the potential to become a core platform within organizations. If ChatGPT is the new gateway to the internet, Glean has the potential to be an entry-level product designed for enterprise scenarios, serving as the primary interface for all SaaS applications and an AI assistant for all employees.
CEO Arvind Jain (born in 1974) graduated with an undergraduate degree from IIT in 1996 and a master’s degree from UWash in 1997. Prior to entrepreneurship, he worked as a software engineer at Microsoft and Google and co-founded Rubrik in 2014.
Product Engineering Tony Gentilcore (born in 1982) graduated with an undergraduate degree from Saint Louis University in 2004. He worked as a software engineer at Google for ten years and co-founded a social platform as a founding member in 2016. He started doing startup consulting in 2018.
Search Piyush Prahladka (born in 1983) graduated from IIT in 2005 and has held Engineering Lead positions at Google and Uber.
Infra – TR Vishwanath (born in 1975) graduated with an undergraduate degree from IIT in 1997 and a master’s degree from UT Austin in 1999. Prior to entrepreneurship, he worked at Oracle and Microsoft, and before starting his own venture, he served as a Principal SE at Facebook.
Arvind Jain has been passionate about entrepreneurship since childhood. Before founding Glean, he had experiences in technology giants, startups, and co-founding ventures, with a high level
of technical expertise. Arvind worked as a software engineer at Microsoft and joined a startup called Akamai Technologies during the booming internet and startup wave, where he served as an architect for three years.
After Akamai, as a founding engineer, Arvind joined Riverbed Technology and gained experience in building a company from scratch. Arvind, who was introverted, learned how to communicate with customers and sell products through his entrepreneurial experiences.
From 2003 to 2014, Arvind Jain worked at Google for eleven years as a Distinguished Engineer, leading the Google Search, Maps, and YouTube product teams. In 2014, Arvind left Google and co-founded Rubrik, which became one of the fastest-growing companies in the cloud data management field. In 2019, Arvind founded Glean.
Entrepreneurial Opportunities
Where are the potential opportunities for entrepreneurs?
In the 2C market, existing unicorns face fierce competition (Jasper, Runway), so there are no significant opportunities, except in vertical fields that have not yet caught the attention of the tech giants, where there may be small opportunities.
The only relatively promising opportunity lies in the 2B market, which has a sufficiently large market size and where the incremental value brought by large models is high. There is a higher probability of making new breakthroughs.
[If you are in China there is basically no opportunities for entrepreneurs without “resources”:
In the consumer sector swept by the mobile internet, there are currently no opportunities in the domestic 2C market. The large market has already attracted widespread attention, and the incumbents hold the data and application entry points, making it difficult for newcomers to disrupt the incumbents. In the small market, as the market matures, there is severe homogenization and squeezed profits, making it challenging to establish a viable financial model.
In the domestic 2B market, due to the prevailing payment concept and efficiency ratio, and the preference of large companies to directly engage in the market, there are also limited opportunities.]
In traditional sectors such as finance and healthcare, there may be opportunities domestically, but they may not necessarily belong to entrepreneurs. Entrepreneurs tend to focus on deepening specific industries and have resources.
Opportunities are scarce, and the requirements for entrepreneurs are high, requiring a fair amount of luck. In a structurally downward environment, finding upward opportunities is the demand of our time.
In fact, it is not very complicated. The principle of Occam’s Razor states that if there are no necessary entities, do not introduce them unnecessarily. Many things can be judged using common sense. What we lack is not the methodology and thinking models needed for decision-making, but the courage to face facts and the execution to make things happen.
As pioneers riding the tide of large models in the entrepreneurial wave, entrepreneurs must balance the tide brought by new technologies with the gravitational pull of the real world and move forward towards the distance.
