GV and Greycroft lead $650M round for Recursive Superintelligence, a stealth AI lab building self-improving systems. NVIDIA and AMD Ventures also participate. Public launch set for mid-2026.
Recursive Superintelligence has raised $650 million at a $4.65 billion valuation in a funding round led by GV and Greycroft, with participation from AMD Ventures and NVIDIA. The stealth AI lab, founded in 2025, is building self-improving AI systems designed to automate parts of the research process, including model architecture, training methods, evaluation, and research direction.
The round signals a shift in venture capital flows toward capital-intensive research labs pursuing autonomous AI, even as other tech sectors see tighter purse strings. Recursive Superintelligence argues that the next stage of AI development will require systems that can improve how they learn, rather than relying only on larger models and greater computing power.
The $650 million raise at a $4.65 billion valuation places Recursive Superintelligence among a growing group of venture-backed companies pursuing artificial general intelligence. The round’s lead investors, GV and Greycroft, are known for backing deep-tech startups. AMD Ventures and NVIDIA joined as participants, signaling hardware makers’ interest in early-stage AI research labs that will need significant compute infrastructure.
For investors tracking AI exposure through public equities, NVIDIA’s participation is notable. NVDA carries an Alpha Score of 69/100 (Moderate) on AlphaScala, reflecting the market’s balanced view of its AI dominance and valuation risk. The company’s involvement in this round suggests it sees Recursive Superintelligence as a potential future customer for its GPUs.
Recursive Superintelligence was started by former leaders and researchers from OpenAI, Google DeepMind, Meta AI, Salesforce AI, and Uber AI. The founding team includes:
This concentration of talent from top AI labs gives the startup credibility in a field where recruiting is the primary bottleneck.
Recursive Superintelligence is building software that can generate, test, and refine new capabilities in a continuous cycle with less direct human supervision. The company compares the process to biological evolution, where improvements accumulate over time and lead to more advanced forms of intelligence.
This contrasts with the dominant approach of scaling existing model architectures with more data and compute. The company’s thesis is that the next leap in AI will come from automating the research process itself, not just throwing more GPUs at larger models.
The company frames its approach as analogous to biological evolution: incremental improvements compound over time, leading to emergent capabilities. This is a deliberate departure from the “bigger model, more data” paradigm that has driven recent advances. The lab is developing self-improving systems that can modify their own architecture and training methods.
Recursive Superintelligence currently operates from San Francisco and London and has grown to more than 25 researchers and engineers. The company plans to use the funding to expand its compute infrastructure and research operations. This capital-intensive model is typical of AI labs pursuing AGI, where compute costs can run into hundreds of millions of dollars annually.
The round highlights how AI funding is moving beyond enterprise software and into research labs that require large amounts of compute and technical talent. For public market investors, this means that companies like NVIDIA and AMD benefit from the proliferation of these labs, the labs themselves carry high execution risk.
Venture investors have become more selective in other technology sectors, AI labs continue to attract large rounds. Recursive Superintelligence’s raise is part of a broader trend where capital is flowing to research-heavy startups rather than traditional SaaS companies. This shift has implications for the public markets: it increases demand for compute hardware while creating a new class of private companies that may eventually IPO or be acquired.
The deal adds to a competitive field that includes startups working on world models, reinforcement learning, and safety-focused superintelligence research. Recursive Superintelligence is seeking to differentiate itself by attempting to automate more of the AI development pipeline, rather than focusing only on scaling existing model architectures.
The company is preparing to run its first “Level 1” autonomous training system, according to the company. A public launch is planned for mid-2026. This timeline gives the lab roughly two years to demonstrate that its self-improving approach can produce tangible results.
The mid-2026 public launch is the next concrete milestone. If Recursive Superintelligence can demonstrate a working autonomous training system, it would validate the thesis that AI research can be partially automated. Failure to deliver on that timeline would raise questions about the approach’s feasibility.
The primary risk is execution. Automating AI research is a moonshot, many similar efforts have failed. The lab also faces competition from well-funded incumbents like OpenAI, Google DeepMind, and Anthropic, all of which are working on autonomous systems. Compute costs could escalate faster than expected, forcing the company to raise additional capital at a lower valuation.
For public market investors, the most direct read-through is to NVIDIA and AMD, which supply the hardware these labs need. Recursive Superintelligence’s success or failure will not move those stocks, the broader trend of capital flowing to AI research labs supports demand for GPUs. The Alpha Score of 69/100 for NVDA suggests the market is already pricing in this demand, leaving limited upside from this specific round.
Recursive Superintelligence’s $650 million raise is a bet that the next phase of AI will be defined by self-improving systems. The company has the talent and the capital to attempt it. The question is whether the approach works faster than the competition can scale.
Prepared with AlphaScala research tooling and grounded in primary market data: live prices, fundamentals, SEC filings, hedge-fund holdings, and insider activity. Each story is checked against AlphaScala publishing rules before release. Educational coverage, not personalized advice.