How World Labs, Nvidia Cosmos, DeepMind Genie 3, Odyssey, Decart, Skild AI, Wayve GAIA, Runway, Luma Dream Machine, Marble.
AI-native world models, spatial intelligence, embodied simulation, neural physics, 3D foundation models is the frontier layer underneath every robotics, autonomy, AR/VR, game, film, XR, digital-twin roadmap in 2026 — validated by Fei-Fei Li's World Labs $230M seed-plus-A Sep-2024 (a16z-NEA-Radical-AIX Ventures-Adobe-AMD Ventures-Databricks-Nvidia-Intel Capital, Marble text-to-3D-world release Nov-2025, $1B valuation), Nvidia Cosmos world-foundation-models open release Jan-2025 CES, Cosmos-Predict, Cosmos-Transfer, Cosmos-Reason for physical-AI, DeepMind Genie 3 Aug-2025 real-time-interactive-world-generation at 720p 24fps, minutes-long consistency, Odyssey $27M seed+A EQT Ventures-GV-Pioneer Fund-Air Street-Kaya-Ted Persson, Explorer interactive-video Dec-2024, Decart $32M seed Sequoia, Oasis Minecraft-world-model Oct-2024, real-time-video-model, Skild AI $300M A Coatue-Lightspeed-Softbank-Bezos Expeditions, robot-foundation-model $1.5B valuation Jul-2024, Wayve $1.05B C Softbank-Nvidia-Microsoft May-2024, GAIA-2 driving-world-model, AV2.0 end-to-end-driving, Runway $308M D General Atlantic-Nvidia-Fidelity Apr-2024, Gen-3 Alpha, Gen-4, Act-One, Frames, Aleph video-to-video Sep-2025, Luma AI $43M B Andreessen Horowitz Jan-2024, Dream Machine, Ray 2, Reframe Video, Modify Video, Genesis physics-engine open-source Dec-2024 CMU-Stanford-MIT-Nvidia-CogniCore, 43x-faster-than-real-time robot-simulation, Physical Intelligence π0, π0.5, $400M Series A Amazon-OpenAI-Thrive-Lux Nov-2024 $2.4B valuation, generalist-robot-policy, Google Veo 3, Veo 3.1, Veo 3.1 Fast May-2025, audio, longer-clips, higher-consistency, OpenAI Sora 2 Sep-2025, Sora Turbo, steerability, physics-consistency-benchmark, Meta V-JEPA 2 Jun-2025 self-supervised-video-world-model, zero-shot-robot-planning — finally proved diffusion, autoregressive, JEPA, Gaussian-splatting, NeRF, neural-radiance-fields, 3D-foundation-models, video-diffusion for interactive, controllable, physically-consistent, minutes-long, real-time world generation at commercial scale.
Four unlocks: (1) World Labs $230M seed-plus-A Sep-2024 a16z-NEA-Radical-AIX-Adobe-AMD-Databricks-Nvidia-Intel, Marble text-to-3D-world Nov-2025, $1B valuation reset the spatial-AI, 3D-foundation-model bar and validated Fei-Fei Li, Justin Johnson, Christoph Lassner, Ben Mildenhall as the reference-team. (2) Nvidia Cosmos Jan-2025 CES open-release, Cosmos-Predict, Cosmos-Transfer, Cosmos-Reason, Cosmos-Drive, Cosmos-Physics, GR00T, Isaac, Omniverse ecosystem, DeepMind Genie 3 Aug-2025 720p 24fps minutes-long real-time, Meta V-JEPA 2 Jun-2025 self-supervised-video-world-model, zero-shot-robot-planning finally proved diffusion, autoregressive, JEPA, Gaussian-splatting, NeRF, neural-radiance-fields for interactive, controllable, physically-consistent, minutes-long, real-time world generation at commercial scale. (3) Robotics, autonomy foundation-model wave — Skild AI $300M Coatue-Lightspeed-Softbank-Bezos $1.5B, Physical Intelligence π0-π0.5 $400M Amazon-OpenAI-Thrive-Lux $2.4B, Wayve $1.05B Softbank-Nvidia-Microsoft, GAIA-2, Figure Helix, 1X Neo, World Model, Nvidia GR00T-Cosmos — pulled world-model, neural-simulator, sim-to-real into the critical-path for every generalist-robot, AV2.0 roadmap. (4) Video-generation supercycle — Runway Gen-4, Act-One, Aleph, Luma Dream Machine, Ray 2, Modify, Google Veo 3.1, audio, OpenAI Sora 2, physics, Kuaishou Kling 2.1, Bytedance Seedance, MiniMax Hailuo, Pika 2.2, Ingredients, Higgsfield camera-control — expanded the addressable market from film, advertising, game to real-time interactive world-model, robot-simulation, AV-simulation, XR, digital-twin.
Seed $10-300M+ for platform, first-model, first-benchmark, first-downstream-partnership (Odyssey $27M EQT-GV-Pioneer, Decart $32M Sequoia, Luma $43M B a16z earlier-round, Skild $300M A outlier, World Labs $230M seed-plus-A outlier, Physical Intelligence $70M seed Thrive-Khosla-Sequoia-Lux earlier-round). Series A/B $100M-$1B for first-real-time, first-minutes-long, first-multi-modal, first-robot-fleet-deployment, first-AV-fleet-deployment, first-XR-deployment, first-film-production-deployment (Wayve $1.05B Softbank-Nvidia-Microsoft, Skild $300M A Coatue-Lightspeed-Softbank-Bezos $1.5B, Physical Intelligence $400M A Amazon-OpenAI-Thrive-Lux $2.4B, Runway $308M D General Atlantic-Nvidia-Fidelity, Luma $43M B a16z, rumored-larger-round-2025). Series C/D, growth, IPO $500M-$10B+ for hyperscaler-partnership, multi-vertical-deployment, international-expansion, IPO (Wayve $1.05B benchmark, Skild $1.5B, Physical Intelligence $2.4B, World Labs $1B, Runway ~$3B rumored-2025-round, Luma ~$3B rumored-2025-round). Non-dilutive material — Nvidia Inception, Nvidia NVentures, AMD Ventures, Intel Capital, Google for Startups, Google DeepMind partnerships, Microsoft for Startups, Microsoft Founders Hub, Amazon AWS Activate, Anthropic Claude for Startups, Meta Llama Grants, Databricks Ventures, Snowflake Ventures, Cloudflare Startup Program, Vercel for Startups, Modal Labs, Together AI, Fireworks AI, Baseten, Replicate, Groq for Startups, Cerebras, SambaNova, Tenstorrent, Etched credits, partnership, strategic-alliance. Government, defense — DARPA, IARPA, ARPA-H, AFWERX, SBIR, STTR, NSF, NIH, DOE, DOD, Space Force, Army Applications Lab, Army Futures Command, Naval Postgraduate School, Air Force Research Lab AFRL, DIU Defense Innovation Unit, In-Q-Tel non-dilutive.
Overclaiming world-model, spatial-AI capability without downstream-partnership, robotics, autonomy, AR/VR, game, film, XR, digital-twin, defense, industrial-deployment, third-party-benchmark validation — post-World Labs Adobe-AMD-Databricks-Nvidia-Intel, Skild Softbank-Bezos, Physical Intelligence Amazon-OpenAI, Wayve Nissan-Uber-Softbank, Runway Lionsgate-Getty, Luma AMD-Google credible-downstream-partnership bar, operators without a signed, published, third-party-validated downstream-deployment or hyperscaler-partnership get flagged by a16z, NEA, Radical, AIX, Sequoia, Coatue, Lightspeed, Softbank, Bezos, Nvidia, Amazon, Microsoft, Google, Meta, Fidelity, T. Rowe, General Atlantic, Thrive, Founders Fund, Khosla, Kleiner, Accel, Bond, IVP, Insight, TCV, Bessemer, Iconiq, Altimeter, D1, Dragoneer, Tiger, Coatue crossover diligence within 12-18 months. Underestimating compute, Nvidia H100-H200-B100-B200-GB200-Rubin, AMD MI300X-MI325X-MI355X-MI400, Google TPU v5-v6-v7, AWS Trainium 2-3, Cerebras CS-3-4, Groq LPU, SambaNova SN40L, Tenstorrent Wormhole, Etched Sohu accelerator-supply, inference-cost, latency, FLOPs, tokens/sec, frames/sec, resolution, duration, consistency, steerability, controllability constraint. Ignoring physical-consistency, object-permanence, occlusion, collision, gravity, rigid-body, soft-body, fluid, cloth, articulated, contact-rich, sim-to-real, zero-shot-transfer benchmark (Cosmos Physical AI Benchmark, PhysGen, Physics-IQ, VideoPhy, WorldSim-Bench, RoboCasa, ManiSkill3, RoboVerse, LIBERO, MetaWorld, RLBench, CALVIN, Franka-Kitchen).
Investor directory · Fundraising library · Articles A–Z · Company funding database
