Cohorts of 2026

Enabling Model-Hardware Co-Design at Scale

AI that simulates physics 1000x faster

Calejo is an intelligence platforms that leverages the AI-potential helping corporates to optimize their business.

Zant simplifies and accelerates the deployment of AI models on edge devices, where cloud-based execution is limited by privacy, latency, or connectivity constraints.

Floating nuclear powered data centers at sea.

AI chips for ultra-fast model training and inference

The easiest way to make private AI models.

Buy, sell, and redeem GPU hours for any week with instant liquidity

Least cost routing system to reduce 80%+ token spend

AI-controlled optical switching for AI clusters

A Quant Neolab building self-improving agents from quant trading

We make Silicon on the Moon.

Open source OpenRouter that turns your traffic into a better model

Building intellectually curious AI

Lightning fast chips with hardcoded AI models

Distributed, hardware-agnostic AI inference

Training human brain cells for AI compute

Earned intelligence for every company

Teaching AI to Design Chips

Market for future inference capacity

Combining open-source AI models for better answers at lower cost

Self-optimizing neocloud that cuts LLM bills by 80%

Continual Learning Models.

AIデータセンター向けに、多波長レーザーを統合した光源PICを設計するファブレス企業

QodeX Quantum builds a platform to train and deploy quantum-native AI models for research and engineering companies - faster, cheaper, and beyond classical limits.

We are the Anyscale for quantum computing.

Fastino provides task-specific language models for efficient and accurate AI development.

Unlock wasted GPU capacity.

AI model built for high accuracy deterministic tasks

Kugel is an on-prem Voice AI model supporting 30+ languages

Near-real-time video infrastructure for LLMs

Emotive foundation voice models

The AI-native observability platform for fast-moving engineering teams

Distributed runtime for AI workloads at scale

Home - An Electronics Photonics Integration Company

Foundation models for 24/7 Earth Observation

AI Cost Management: Track and attribute AI spend across every provider

Distilling Proprietary Intelligence

AI models that learn from experience

The Fastest Multimodal Inference OS

Building AGI that can innovate.

AI Infra for real-time vision applications

The Networking Layer of AI

Building secure polymorphic apps

Infrastructure for Autonomous Science

A faster, lower cost path to custom silicon

The GreenThread inference engine allows you to multiplex many LLM's and SLM's on the same GPU, swapping them subsecond.

Unlocking the next generation semiconductors

Developing an Edge Infrastructure platform that enables the use of Cloud Native SW stacks

Neuromorphic hardware for training foundation models

Adaptive data transfer optimization for high-performance infrastructure, Berkeley CS founders.

Unlocking the next one billion specialized models