In 7 months, the valuation increased by 15 times.
The wealth-creation myth in the AI industry continues to unfold.
According to foreign media reports, AI startup Reflection AI recently completed a new round of financing worth $2 billion, with a post-investment valuation of up to $8 billion (approximately RMB 57.1 billion).
In March this year, Reflection AI received a $130 million Series A financing, with a valuation of about $545 million at that time. That is to say, in just seven months, the company's valuation has increased by an astonishing 15 times.
The list of investors in this round is quite impressive, including well - known institutions and investors such as Lightspeed Venture Partners, Sequoia Capital, DST Global, former Google CEO Eric Schmidt, and 1789 Capital, where Donald Trump Jr. is a partner.
Notably, NVIDIA also made an appearance in this deal - it directly invested $800 million as the lead investor. This is already NVIDIA's eighth move in the AI field since September.
Silicon Valley Geniuses' Entrepreneurship
Like many AI startups that have received large - scale financing recently, the founding team of Reflection AI also has a dazzling "star aura" - both founders are from Google DeepMind and have a profound R & D background in the AI field.
Among them, Ioannis Antonoglou worked at DeepMind for 12 years and participated in the development of AlphaGo. This AI system defeated the world champion in Go in 2016, becoming one of the milestone events in the history of AI development.
Misha Laskin was the former core leader of Google Gemini and also led Google's directions in reinforcement learning and large - language models.
In March 2024, the two co - founded Reflection AI. When talking about the reason for leaving DeepMind, Misha Laskin publicly stated that they believe the realization of Artificial General Intelligence (AGI) is not far away, and an independent new company will be more conducive to the rapid implementation of technologies.
This confidence stems from their judgment on the development stage of AI. Laskin predicted that "intelligent agents for small tasks" will be implemented first, and "general super - human intelligent agents" will appear in about three years. They hope to prove that "even without relying on technology giants, top - notch talents can independently build the most cutting - edge models."
At the product level, Reflection AI initially focused on autonomous programming agents and launched Asimov, a code - understanding agent for engineering teams, in July this year. This product can analyze enterprise data such as emails and Slack messages and generate corresponding codes. It is reported that Asimov has initially achieved enterprise - side revenue.
However, Laskin pointed out that Asimov is just the first step. In the future, Reflection AI plans to expand "enterprise - level super - intelligence" to fields beyond coding, covering "team memory" and knowledge management in product, marketing, human resources and other aspects.
The 'American Version of DeepSeek'
In addition to the halo of the star - studded team, another core selling point of Reflection AI is its focus on "open - source", so it is also called the "American version of DeepSeek".
During the Spring Festival in 2025, DeepSeek emerged. Its open - source strategy allows developers around the world to join the project for optimization, breaking the monopoly of a few enterprises on cutting - edge models.
This model is not only regarded as a "software of national - fortune level" that may change the global industrial pattern but also has a strong impact in the US technology circle. US technology media bluntly stated that "DeepSeek has completely subverted the research consensus of the past decade."
Laskin also publicly stated: "Models like DeepSeek are a wake - up call for us. If we do nothing, the global intelligence standards will be set by others, not by the United States."
He further pointed out that if the most cutting - edge technologies continue to be concentrated in a few closed laboratories, it will lead to a "snowball - style" monopoly of capital, computing power and talents, excluding the vast majority of participants.
"The United States currently lacks an existence like DeepSeek," he emphasized. "This is exactly the reason why Reflection AI must exist."
In terms of specific strategies, Reflection AI is not completely open - source. Instead, it chooses to open the model weights (i.e., the core parameters of AI) for researchers and developers to use, but the training data and the complete process remain private. This balance also supports its business model: researchers can use the model for free, and the company's main revenue will come from enterprise customers building products based on its model and participating in the construction of "Sovereign AI" systems in various countries.
The concept of Reflection AI has been recognized by the US technology circle and policy - makers.
David Sacks, the White House AI and Crypto Affairs Commissioner, publicly stated on social media: "I'm glad to see more American open - source AI models emerging. A considerable part of users in the global market value the cost, customizability and control brought by open - source. We hope the United States can also lead in this field."
The response from the capital market is more direct. In March this year, Reflection AI completed a $130 million Series A financing, with investors including top - tier institutions such as Sequoia Capital, CRV and NVentures under NVIDIA. And just seven months later, the company completed a new round of financing worth $2 billion.
According to media reports, Reflection AI completed this round of financing within a few weeks, which is enough to reflect the strong interest and recognition of investors.
As of now, the Reflection AI team has about 60 people, mainly consisting of AI researchers and engineers in infrastructure, data training and algorithm development. The company has deployed a large - scale computing power cluster and plans to release a cutting - edge language model with a training scale of "trillions of tokens" next year.
NVIDIA's Eight Moves in Two Months: Going All - Out
With the completion of the Reflection AI deal, NVIDIA, the lead investor behind it, also deserves attention.
My colleague, Teacher Zhang Xue, described NVIDIA's madness in the article "Investing 65.5 Billion in 7 Days: Jensen Huang Goes All - Out". In just one week in mid - September, NVIDIA officially announced five external investments in Intel, Enfabrica, Wayve, Nscale and Dyna Robotics, with a total amount of more than $9.2 billion (approximately RMB 65.5 billion).
Before that, Jensen Huang also announced that NVIDIA would invest up to $100 billion in OpenAI to build a data center with a computing power scale of at least 10GW and deploy about 4 - 5 million NVIDIA chips.
On October 8th, Bloomberg reported that Jensen Huang officially confirmed in an interview with CNBC that the company plans to invest $2 billion (approximately RMB 14.269 billion) in xAI, an artificial intelligence startup founded by Elon Musk. This investment will be in the form of a combination of debt and equity, mainly to support xAI's "Colossus 2" supercomputer project.
Adding this lead investment in Reflection AI, NVIDIA has made eight moves in the AI field since September, with a cumulative investment of over $100 billion.
According to a report released by investment tracking institution PitchBook, as of October 9th, NVIDIA has participated in 50 AI - related financings this year, exceeding the total of 48 in the whole year of 2024.
Behind these frequent moves is the support of abundant funds.
FactSet data shows that NVIDIA has generated a total of up to $72 billion in free cash flow in the past four fiscal quarters and is expected to exceed the $100 billion mark by the end of this fiscal year. In this regard, Cailian Press gave an intuitive comparison in its report: this figure has exceeded the expected total free cash flow of all technology giants this year except Apple.
It is not difficult to speculate that NVIDIA, with a large amount of cash on hand, will continue to make investments in the future. As for the selection of investment targets, NVIDIA also has a clear direction, that is, to expand the AI ecosystem by supporting startups it considers to be "game - changers and market - creators".
This article is from the WeChat official account "China Venture Capital", author: Wang Manhua. It is published by 36Kr with authorization.