Deep Learning
CAPITAL FIGURES ARE MEDIA-EXTRACTED ESTIMATES, NOT VERIFIED FILINGS.
EXTRACTED FROM 25+ PODCASTS & VC NEWSLETTERS · MEDIA-REPORTED FIGURES, NOT VERIFIED FILINGS
Market Context
Deep learning is experiencing a surge of capital deployment across scientific discovery, physical AI, and vision systems, with $26.97B deployed across 29 deals in the last 28 days — driven by the convergence of foundation model architectures, agentic workflows, and real-world robotics applications. Chinese academic institutions including Shanghai AI Laboratory, Fudan University, Shanghai Jiao Tong University, and Zhejiang University are emerging as serious counterweights to US/UK labs in VLA (Vision-Language-Action) model research, while investors like Lux Capital and NVIDIA each backed 7 deals in the period. The regulatory environment is also tightening, with Illinois passing legislation requiring frontier model developers including Google DeepMind to undergo independent third-party safety audits.
Investment Activity
- Skild AI raised a $14B Series A led by Felicis Ventures, signaling massive conviction in robotics foundation models.
- PhysicsX raised $32M in a Series A led by General Catalyst, backing deep learning simulation software for advanced industry.
- Ultralytics closed a $30M Series A with participation from Felicis Ventures, accelerating open-source vision AI via its YOLO models.
- Sakana AI raised $30M in a seed round led by Lux Capital, with participation from Khosla Ventures, Jeff Dean, and Clement Delangue.
- Autoscience raised $14M in seed funding led by General Catalyst and Toyota Ventures for a fully autonomous AI research lab.
- Phylo raised $13.5M in a seed round from Andreessen Horowitz for its AI Scientist agentic orchestration platform.
- Unlearn.AI raised $12M in a Series A from 8VC for clinical trial digital twins powered by deep learning.
- Sim raised $7M in a Series A backed by YCombinator, Felicis Ventures, and Perplexity Fund for agentic workflow infrastructure.
- Axiom Math raised $200M at a $1.6B valuation in March 2026 for AI-powered formal mathematical reasoning.
Key Players
- Skild AI is a robotics foundation models company that attracted an extraordinary $14B Series A from Felicis Ventures, underscoring the scale of bets being placed on general-purpose physical AI.
- PhysicsX is the leading deep learning simulation software company, backed by General Catalyst and Standard Investments in a $32M Series A to bring AI-driven simulation to advanced industry.
- Sakana AI is building nature-inspired foundation AI models out of Japan, having closed a $30M seed round with backers including Lux Capital, Jeff Dean, and Clement Delangue.
- Ultralytics is the developer of the widely-used YOLO vision AI models and raised a $30M Series A from Felicis Ventures and SquareOne to scale its open-source platform.
- Axiom Math is an AI math reasoning startup valued at $1.6B after raising $200M, using formal proof language Lean via its AxiomProver tool, competing directly with Harmonic which is backed by NVIDIA.
Market Signals
- Lux Capital and NVIDIA each led 7 deals in the past 28 days, indicating that deep learning is a top-priority theme for the most active crossover investors.
- Chinese academic institutions — Shanghai AI Laboratory, Fudan University, Shanghai Jiao Tong University, and Zhejiang University — are producing frontier VLA research rivaling US labs, as demonstrated by BORA's 86% dexterous task success rate and VisualThink-VLA's 22.8× latency speedup over ECoT.
- A new 100M-image permissive corpus (GPIC) was released on Hugging Face, co-authored by Salesforce Research and Radical Numerics, providing commercially usable training data that could accelerate vision model development industry-wide.
- VLA-Pro from Fudan University achieved a 207% relative improvement in simulation benchmarks and an ~11× improvement on unseen real-world manipulation tasks, indicating rapid capability gains in physical AI.
- Illinois enacted what may be the US's strongest state-level AI safety law, requiring Google DeepMind, OpenAI, and Anthropic to submit to independent audits — a signal that regulatory pressure on deep learning labs is accelerating.
- Deal velocity stands at 29 deals in 28 days with a rising velocity score of 1.13, confirming sustained acceleration rather than a one-time spike.