Two months surge by 40%! Cohere, the OpenAI rival backed by NVIDIA, is valued at $3 billion.

Artificial intelligence has become the hottest trend in the technology industry. Cohere, a Canadian startup in the field of generative AI and considered one of the rivals of OpenAI, has seen its valuation increase by nearly 40% in just two months.

According to insiders who spoke to the US technology media outlet The Information, Tiger Global Management, one of the world's largest unicorn investors, plans to sell a portion of its stake in Cohere, raising the company's valuation to approximately $3 billion.

In June of this year, Cohere participated in a Series C funding round led by NVIDIA, Oracle, and CRM software leader Salesforce. At that time, Cohere's valuation was around $2.2 billion. This represents a 36% increase in its latest valuation.

Reports indicate that Tiger Global Management intends to sell 2.1% of its stake in Cohere for approximately $63 million, which is exactly the same as its initial investment of $62.7 million in December 2021. "The savvy Tiger Fund has quickly recouped its investment."

After selling a portion of its shares, Tiger Global Management will still hold approximately 5% of Cohere. In February of last year, Tiger Global led a Series B funding round for Cohere, investing $125 million with the aim of accelerating the platform's development and expanding its international influence.

It is reported that Cohere raised a total of $270 million in its Series C funding round in June, with a valuation of $95 per share. The valuation for the shares that Tiger Global plans to sell has now increased to between $130 and $140 per share.

Analysts believe that due to the significant decline in the stock prices of many listed companies it holds, Tiger Fund is facing liquidity constraints. It is currently selling its holdings in many portfolio companies and even attempting to sell shares in small venture capital funds that it supported during the peak of the bull market.

Pitchbook, a private capital market data SaaS company under Morningstar, stated that John Curtius, a former partner at Tiger Global who has now left the company, led the investment in Cohere and also led its Series B funding round. The valuation of Cohere at that time was not disclosed, and generative AI technology was still relatively unknown. Tiger Global is also an investor in high-valued technology companies such as OpenAI and ByteDance.

It is said that Tiger Global is facing fundraising challenges. In February, the company lowered the fundraising target for its next fund from $6 billion to $5 billion. Securities filing documents show that as of June, Tiger Global had only raised $2.7 billion. In 2021, the company significantly reduced the value of its portfolio of private technology company investments.

Public information shows that Aidan Gomez, former Google AI researcher, is the CEO and co-founder of Cohere. In 2017, he co-authored a groundbreaking academic paper that first proposed the concept of the Transformer neural network, which is now widely used for text processing optimization. It has become one of the most important breakthroughs in machine learning and artificial intelligence technology in recent years. Also driving the most advanced large language models like OpenAI's GPT-4.

Cohere, founded in 2019, provides access to a range of generative AI models through a cloud-hosted API. Its models, optimized with a core Transformer neural network called Command, generate text based on user prompts, similar to ChatGPT. Its main applications include creating marketing copy, generating emails, summarizing documents, and providing customer support through chatbots.

Cohere has also developed more specialized AI models, including one for text summarization, another for website search, and a third model that converts words into vector embeddings, a data format that is easier for AI applications to process than raw text.

Like OpenAI, Cohere allows customers to fine-tune its AI models by training them on custom proprietary datasets. This enables customers to train Cohere's models to perform more specific tasks and engage in conversational interactions. It also offers a significant advantage by allowing customers to deploy these models on any public cloud platform or within their own data centers.

In an interview in March of this year, Gomez stated that Cohere's large language models can be "deployed anywhere, in any industry" because "language is everywhere." He also predicted that "next year will see some major changes" and that currently people are only beginning to understand the ultimate capabilities of generative AI models:

"We've already seen the first hints of this capability through tools like Bing Chat, where external knowledge bases can be used to enhance these dialogue models. Now, these models can stay up-to-date in real-time at the millisecond level because they can search the web."

In the weeks leading up to Cohere's Series C funding round in June, another startup called Anthropic, which develops advanced generative AI models, raised $450 million in funding led by Spark Capital, bringing its valuation to $4.1 billion. In February of this year, Google invested over $2 billion in Anthropic, a "ChatGPT competitor," to compete against Microsoft-backed OpenAI.

Last week, Wall Street News mentioned that several popular large language models tend to "talk nonsense" (i.e., "AI illusions"). According to tests conducted by Arthur AI, OpenAI's GPT-4 performed the best overall among all models, with Anthropic's Claude-2, supported by Google, ranking second, Meta's Llama-2 performing in the middle, and Cohere being the most "confidently creative" in making things up. At the same time, Meta plans to launch an open-source AI software in the near future, aiming to help developers automatically generate programming code. It is another bold move that may disrupt the field of artificial intelligence, following the Llama 2 language model. It is referred to as an open-source model using "free version programming tools" to challenge the charging ecosystem and target vendors such as OpenAI, Cohere, and Anthropic, who sell similar technologies to enterprises.