Today for AI
HOT RADAR
industryHEAT 0.0°

AI CURATED HOT STORY · 10/12/2026

Infinigence AI Raises 2.3 Billion Yuan in Series C, Setting AI Infra Record with PDD Architecture Cutting First-Token Latency by 51.5%

3 reports archived0 independent sourcesupdated 10/13/2026, 12:01:15
Synthesis & Latest Updates

Infinigence AI has completed a 2.3 billion yuan Series C funding round, bringing its total raised to 4.3 billion yuan and setting new records for single-round and cumulative financing in China's AI-native infrastructure sector. Specializing in the 'middle layer' between models and chips, the company aggregates heterogeneous computing resources to address inference cost bottlenecks amid the Agent boom, with plans to list on the HKEX by mid-2027.

Technically, Infinigence employs an M×N paradigm for mixed training across 16 major chip types, including Huawei Ascend and NVIDIA, achieving up to 97.6% utilization. Its proprietary PDD architecture reduces first-token latency (TTFT) by 51.5% and per-token costs by 37.5%. The investor lineup includes state-backed entities, industrial giants like Tencent and Baidu, and top-tier VCs such as Sequoia China, signaling strong market confidence in high-efficiency infrastructure solutions.

This milestone underscores the maturation of domestic AI Infra. By optimizing token production from every watt and second of compute, Infinigence has deployed Agentic Infra solutions in healthcare and entertainment. This approach not only benchmarks against global peers like Fireworks AI but also redefines heterogeneous scheduling standards, offering developers a more cost-effective foundation for large model services.

LATEST/Infinigence AI has secretly filed for an IPO on the HKEX, targeting a listing by mid-2027. The company is accelerating the deployment of its Agentic Infra solutions in industries like healthcare and entertainment, validating its commercial viability in heterogeneous computing aggregation and token cost optimization.

HEAT TRENDHourly heat curve

74 fully observed hours
Now
0.0
Peak
25.213:00
24h change
-100%

The line compares only sources observed throughout; gap hours are interpolated to keep the trend continuous.

TIMELINECoverage timeline

Total 3 reports · Latest first
  1. 量子位官网T2·68 pts
    • Infinigence AI's 2.3B yuan Series C and 4.3B yuan total raise set new records for domestic AI infrastructure funding
    • The company operates as middleware connecting models and chips, addressing fragmented heterogeneous compute resources and costs
    • Benchmarking against high-valuation overseas peers like Fireworks AI highlights the growing inference compute gap driven by Agent adoption
  2. IT之家 · 智能时代T2·78 pts
    • Infinigence AI closed a 2.3 billion RMB Series C, totaling 4.3 billion RMB in funding, setting a new record for China's AI infrastructure sector.
    • Core technology focuses on heterogeneous compute scheduling, supporting mixed training across 16 major chips (including Huawei Ascend and NVIDIA) with up to 97.6% utilization.
    • Proprietary Cross-Cluster Heterogeneous PD Separation Architecture (PDD) optimizes inference: reducing TTFT by 51.5% and per-token costs by 37.5%.
    • The company has confidentially filed for an HKEX IPO, aiming for a listing as early as the first half of 2027.
  3. 量子位 (微信公众号)T2·68 pts
    • Infinigence AI raised 2.3B yuan in Series C, totaling 4.3B yuan, setting new fundraising records for domestic AI Infra.
    • Positioned as the middleware between models and chips, it aggregates heterogeneous compute to lower token production costs.
    • The explosion of Agents drives massive inference requests, sustaining compute shortages and prompting capital bets on efficient infrastructure.