Author｜Hualin Dance King

In 1956, a group of scientists gathered at Dartmouth to formally discuss "Can machines think?" They optimistically believed they could solve this problem in a summer.

Seventy years later, this question still has no answer. But one company, which was just founded four months ago, has secured $500 million in funding, with a valuation reaching $4 billion - all because it claims to have found a way for AI to learn to conduct research and evolve on its own.

This company is called Recursive Superintelligence.

GV, Google's venture arm, led the investment, with NVIDIA participating. The positions of both companies in the AI ecosystem are well-known and need no further elaboration. The fact that both have chosen to invest in a startup that has not even made its products public is worth serious analysis.

01 "Removing Humans from the Loop"

Let's first discuss what Recursive Superintelligence is actually doing.

The company was founded by former Salesforce chief scientist Richard Socher, with its core team coming from Google DeepMind and OpenAI. This is not an unfamiliar combination - over the past two years, there has been a noticeable wave of engineers and researchers leaving top labs to start their own ventures.

Richard Socher's X personal homepage, Altman has clearly taken note of this talent｜Image Source: X

Socher is not the kind of founder who merely comes from a big tech company to gain some prestige. Born in Germany in 1983, he studied under AI pioneers Andrew Ng and NLP authority Christopher Manning at Stanford University, completing his doctoral thesis in 2014, winning the Best Doctoral Thesis Award from the Stanford Computer Science Department.

Richard Socher is one of the key figures who truly brought neural network methods into the field of natural language processing - his early research on word vectors, context vectors, and prompt engineering directly laid the technical foundation for today's BERT and GPT series models, with over 180,000 citations on Google Scholar.

In the year he graduated with his doctorate, he founded the AI startup MetaMind, which was acquired by Salesforce two years later through a strategic merger. Since then, he has led Salesforce's AI strategy as Chief Scientist and Executive Vice President for several years, overseeing the launch of enterprise AI product lines such as Einstein GPT.

After leaving Salesforce, he founded the AI search engine You.com in 2020, which completed its Series C financing in 2025, reaching a valuation of $1.5 billion. This time, he has shifted his focus from search to more fundamental propositions.

Thinking Machines Lab, Safe Superintelligence, Ineffable Intelligence, Advanced Machine Intelligence Labs...each one emerges with the label “core team of the previous XX large models,” each tells a story of “next-generation AI.”

But Recursive's approach is more radical than most of its peers.

Its core proposition is "self-learning AI" - not to make AI smarter at answering questions, but to allow AI to autonomously complete the entire process of scientific research: formulating hypotheses, designing experiments, evaluating results, and iterating directions. In other words, it aims to completely remove human researchers from this loop.

This is not a novel direction, but Recursive has placed it within an extremely realistic business logic. Currently, top AI researchers earn annual salaries ranging from $15 million to $20 million; if a system can accomplish the same work at a lower cost and faster pace, the economic model of frontier research would be fundamentally rewritten.

Investors evidently see this logic. Reports indicate that the funding round was over-subscribed, with the final amount potentially reaching $1 billion.

02 Google and NVIDIA Bet at the Same Time

GV led the investment, with NVIDIA participating. This combination of investors is itself a signal.

Google's logic is easy to understand. DeepMind has long been the most important explorer in the "AI for Science" direction, with AlphaFold solving the protein folding problem and AlphaGeometry defeating top human competitors in mathematics competitions.

But DeepMind's path is to use AI to solve specific scientific problems, while Recursive aims to do something more fundamental - enabling AI systems to autonomously drive the process of scientific discovery itself. This represents both competition for Google and an investment worth betting on.

More importantly, just earlier this month, Google announced a multi-generational AI infrastructure collaboration agreement with Intel. This indicates that Google's layout at the AI infrastructure level is accelerating comprehensively. Investing in Recursive is a piece on this larger chessboard - whoever gets ahead in the race for the leading model is something Google wants to have a stake in.

NVIDIA's logic is more straightforward. The core bottleneck of self-learning AI is not the algorithm but the computing power. If AI is to autonomously run experiments and iterate models, the required scale of GPU clusters grows exponentially. NVIDIA's investment in Recursive, to some extent, is an investment in its future orders.

The simultaneous moves by the two companies also release a more subtle signal - this track may have reached the stage where "not investing may be too late."

03 Is a $4 Billion Valuation in Four Months Reasonable?

It is expected that when everyone saw the $4 billion figure for the first time, the immediate reaction was "Here we go again."

The topic of the AI startup valuation bubble has not been new over the past two years. A PDF, a demo, a few slides, plus a couple of names from top labs can leverage hundreds of millions of dollars - this is no longer a legend in Silicon Valley and London, but daily reality.

But looking closely at Recursive's situation, there are a few points that are different from the ordinary "PPT unicorns."

First, the weight of the founding team. Richard Socher has real academic experience in the NLP field, not purely relying on the halo of being from a "big company" for packaging. The core team's experience in DeepMind and OpenAI also means they have actually encountered the pain points of frontier research.

Second, the fact that financing was oversubscribed. This indicates that market demand far exceeds supply, investors are vying to get in rather than being persuaded to come on board.

However, a $4 billion valuation for a four-month-old company with no public product relies on expected value rather than reality for its pricing. Essentially, this is paying for a direction rather than for a product or revenue.

Such pricing logic is becoming increasingly common in the AI era, stemming from investors' deep-seated fear of "missing the next OpenAI." Safe Superintelligence also achieved a sky-high valuation in a state of having almost no product, Ilya Sutskever's name being its strongest asset.

Recursive is following the same path. This is not criticism, but an objective observation.

04 What Lies Behind the Door of "Self-Learning"

The name Recursive Superintelligence already makes the company's ambitions quite clear.

"Recursive" means recursion. In computer science, recursion is a structure where a function calls itself, which is the core mechanism of many complex algorithms. Applied to AI research, "Recursive Superintelligence" suggests a system capable of continuously optimizing itself and spiraling upwards.

This concept is not new; its extreme version is "intelligence explosion" - when a system surpasses a certain critical point, it can autonomously accelerate its evolution, ultimately achieving levels of intelligence beyond human comprehension. This has long been one of the core concerns in AI safety.

However, what Recursive is currently doing is likely far from that level. A more realistic interpretation is that they are trying to build a system that can autonomously drive the cycle of scientific exploration, aiming to significantly reduce the labor and time costs of AI research.

If it can truly achieve this, the impact will not be confined to the AI realm. It means that fields such as drug development, materials science, and physics could enter a phase of "rapid advancement without human scientists' involvement."

Of course, this is still "if."

The distance from claiming to realizing is never linear in the AI industry.

05 The Logic of the Wave

Since the second half of 2025, there has been wave after wave of talent leaving top labs to start their own businesses. Thinking Machines Lab, Safe Superintelligence, Ineffable Intelligence... this list keeps growing.

Recursive is the latest and currently the highest valued company in this wave.

The structural reasons behind this are simple - competition from OpenAI, Anthropic, and Google DeepMind has made these leading labs increasingly resemble large companies, with KPIs, compliance, and politics.

Researchers who really want to bet on the most radical directions feel that starting their own ventures allows for more freedom.

At the same time, the logic in the capital markets is reinforcing this trend. For top researchers backed by big companies, the window for starting a business right now might be the best in history - investors are more willing than ever to pay for "direction."

The core question of this wave is not "Who will succeed?" but "What does success mean?"

If Recursive ultimately proves the feasibility of self-learning AI, it will rewrite the underlying paradigm of AI research. If it fails to do so, after burning through $500 million in funding, what will remain is another overhyped concept.

Both possibilities genuinely exist.

Four months, $4 billion valuation, this figure excites yet also alerts us. As the AI arms race has evolved to today, even the act of "how to conduct research" itself has become a battleground of competition.

Scientists debated a question at Dartmouth for an entire summer, and now someone plans to answer it with AI - using AI to research AI, racing towards superintelligence in a recursive manner.

Where this path leads is something no one truly knows. But evidently, Google and NVIDIA have decided that wherever it leads, they cannot afford to be absent.