Qingcheng.AI

RadixArk, founded in 2025 in Palo Alto, California, USA, develops AI systems around the SGLang inference engine and Miles reinforcement learning framework, improving model inference and training efficiency by reducing compute and memory requirements.

Seed · $100M · 2026-05 · Valuation $400M

Autoloops provides inference infrastructure for voice-agent builders: serverless APIs and on-demand GPU clusters serving open-source LLMs such as Gemma at realtime speed. Its proprietary Hanoi engine targets low-latency voice workloads, claiming 2–2.5x lower p50/p95 latency than DeepInfra Turbo and tuned vLLM at high concurrency. Public case study: voice-AI testing platform Cekura. Pitched at GMI Cloud's SCALE accelerator Cohort 2 Demo Day (2026-10-23).

Inference.net is a San Francisco-based AI inference infrastructure company. It began in early 2024 as Kuzco, a Solana-based decentralized network aggregating idle GPU compute for LLM inference, and rebranded in mid-2024. It now offers an LLM lifecycle product suite (Deploy, Observe, Trace, Train, Evaluate) that lets AI-native teams run open-source and custom models at lower cost than proprietary APIs.

Fangqing Technology, founded in 2022 in Shanghai, China, develops multi-core NPU and decoupled distributed AI computing chips.

Series A · $71M · 2026-08 · Valuation $1.5B

Xingyun Integration, founded in 2023 in Beijing, China, is a GPU chip developer.

Undisclosed · 2026-09

APMIC (Accelerate Private Machine Intelligence Company; formerly Asia Pacific Machine Intelligence Company), founded in Taipei in 2017, provides enterprise on-premises AI: private LLM fine-tuning/distillation (PrivModel), an end-to-end on-prem AI platform (PrivStation) and a no-code enterprise assistant builder (CaiGunn), serving regulated industries such as finance, government, manufacturing and healthcare. An NVIDIA Inception member, it expanded into the U.S. in 2025.

SLNG, founded in 2025 in Barcelona, builds a global execution layer for real-time voice agents, sitting between orchestrators such as LiveKit and Pipecat and the underlying models, with a unified API across 30+ STT/TTS/voice-cloning models plus in-region routing, failover and compliance.

$4M

Callosum, founded in 2024 in London, UK, develops systems-level software that enables different AI models to run collaboratively across diverse chip architectures, focusing on AI infrastructure for heterogeneous computing and multi-chip coordination.

Seed · $110M · 2026-08

Olix, founded in 2024 in London, develops photonic AI chips targeting inference workloads to reduce high-bandwidth memory constraints.

Series B · $532M · 2026-08 · Valuation $3.3B

Decart, founded in 2023 in Tel Aviv, Israel, improves the efficiency of AI models to enable faster and more reliable training and real-time inference. Its diffusion model can generate video in real time as users watch; its model powers the open-world video game Oasis, rendering new portions of the game world in real time based on player movement inputs.

Undisclosed · $453M · 2026-05 · Valuation $100M

Paritok is an open-source context compression gateway for AI coding agents, sitting between agents such as Claude Code, Cursor and Codex and LLM APIs to cut token costs via tool-schema filtering, content compression and history summarization. It ships Paritok-4B, a compression model trained on 45K real agent trajectories, under Apache-2.0, with an optional paid hosted tier.

Speculum AI calls itself 'The AI Runtime Company'. It began stealth R&D on an independent AI runtime architecture in Taiwan in January 2024 and launched commercially in January 2026, operating across Taiwan, Japan and Singapore with Japan as its first enterprise market. Mirror OS (deployed as Mirror Shield, Mirror Runtime, Mirror Foundry) governs runtime behavior stability of production AI, and Mirror Token manages consumption efficiency, for enterprise, government and mission-critical deployments.