Core Events: Funding, IPO, and Product Launch Timeline
- OpenAI: Negotiating preliminary discussions for a new funding round that could value the company at over $1.2 trillion before IPO;
- Anthropic: Has selected NASDAQ as its potential IPO venue with a target timing of October this year; no IPO filing yet, and market expectations reach up to $2 trillion;
- Huawei: Ascend 960 chip accelerated on schedule—Ascend 960DT to launch in Q1 2027, three quarters earlier than planned; Ascend 960PR to follow in Q3 2027;
- Alibaba: Qwen3.8-Omni-Flash multimodal model launched with enhanced audio-video AI capabilities;
- AutoNavi: NavigationLive to launch in 2026, focusing on complex AI dialogue for trip planning.
Funding and IPO: Capital Race in AI’s Hottest Funding Cycle
OpenAI and Anthropic’s financing/IPO moves mark a key phase of capital monetization in AI. OpenAI’s projected $1.2 trillion valuation would surpass Microsoft and Apple; Anthropic’s $2 trillion target—if achieved—would make it the largest tech IPO in history.
A subtle divergence emerges in their regulatory stance. OpenAI CEO Sam Altman publicly welcomes a federal safety framework, stating the company “welcome[s]制定一套针对前沿人工智能的联邦框架”. Anthropic’s Dario Amodei previously advocated for slowing development pace. While both prioritize safety, their growth-rhythm philosophy differs.
Compute infrastructure has become the hard constraint. Broadcom expects AI semiconductor revenue at $115 billion in FY2027, doubling to $230 billion in FY2028. Zhipu’s scaling model estimates 3 billion RMB (~$420M)建10万P compute, with theoretical annual revenue of 40 billion RMB—though real-world performance depends on utilization, discounts, and pricing volatility.
Huawei and Tech Giants’ Product Cadence: Ascend 960 Accelerated, Model Capabilities Expanded
Huawei is accelerating both hardware and software ecosystems. The accelerated Ascend 960 launch reflects urgency in domestic substitution and large-scale model support. Wang Tao noted at AIDC 2026 that 100,000+ chip ultra-node clusters will become baseline by 2027, providing downstream demand.
Huawei Cloud unveiled multiple enterprise AI products at HUAWEI CONNECT 2026:
- Lingqu Ascend 950 intelligent compute cluster service globally live;
- Agentic MaaS platform aggregating models across a wide variety of types;
- Enterprise agent platform ZhiGuo AgentArts serves 100+ enterprises;
- OceanStor M900 AI memory storage supports “one-hop direct connection” L3.5 PB-level KV cache, improving SSD lifespan 16x.
Domestic model usage shows initial strength. OpenRouter data indicates Chinese models maintained 20 consecutive weeks lead in global API call volume, with 61.17 trillion tokens this period versus America’s 21.76 trillion. Yet growth decelerated: China +7.85% vs. America +31.56% week-over-week, warranting attention.
Product Comparison and Adoption Advice
Table: Product parameters from source material only.
| Product | Key Features | Status |
|---|---|---|
| Qwen3.8-Omni-Flash | Native multimodal model, enhanced audio-video Agent capabilities | Launched |
| Dobby Cabin Assistant | Complex request understanding, whole-vehicle control, user preference memory | Partnered with SAIC; Roewe Jiayue 07 pre-sales coming |
| Feishu 8.0 | Agent-centric redesign, CLI points increased from 247 to 767 | Launched; agents can join groups |
| NavigationLive | Multi-modal perception + spatial understanding in driving/walking/touring | Launching 2026 |
Reader Recommendations
- Enterprise users: Consider Feishu 8.0 + Dobby Cabin Assistant integration for Agent pipeline builds; its “independent identity, permission control, and memory” suits organizational collaboration; 70% YoY growth among small/medium customers indicates strong mid-market fit;
- Developers: Zhipu AI’s recent apology regarding data practices underscores risk: sensitive code projects should disable default data features; prefer open-source tools with transparent data policies;
- Wait-and-see: Anthropic’s financial details await mandatory pre-roadshow filing window.
Final Note
AI evolution has shifted from pure technology competition to dual-track race: commercialization validation and infrastructure build-out. Regulatory focus is shifting from “whether needed” to “how designed”. While China leads in model call volume, slowing growth implies a transition from scale expansion to deep operations.