Xpeedic Releases EDA 2027 Version, STCO and AI for EDA Become the Focus of the 2026 User Ecosystem Conference

On October 30, the 2026 Xpeedic User Ecosystem Conference was held in Shanghai under the theme "Chip Leaps Across Systems, Intelligence Masters All Changes." More than a thousand representatives from industry chain companies and universities exchanged views on STCO (chip-to-system co-optimization), system-level EDA, and AI for EDA. Xpeedic released the EDA2027 version at the conference.

In her address, Chen Wen, Deputy Secretary-General of the China Semiconductor Industry Association, said that artificial intelligence is driving semiconductor industry innovation to shift from process scaling to system integration. She affirmed Xpeedic's exploration in recent years of continuously expanding system-level EDA capabilities along the "chip-to-system" direction, and expressed the hope that Xpeedic will continue to solidify multi-physics simulation as its "ballast stone" and promote the integration of AI multi-agent systems with EDA.

Zhou Feng, Director of the Integrated Circuit Institute at CCID Research Institute, noted from the perspective of industrial planning that advanced packaging, multi-chip integration, and EDA have been listed as priorities for industrial foundational capabilities in the "15th Five-Year Plan," and that EDA needs to move further from point tools toward systematization and intelligence.

Computing Power Depends on the System, Design Requires Collaboration

In his opening keynote, Dr. Ling Feng, Chairman of Xpeedic, pointed out that AI is changing the way computing power is measured. In the past, chip performance was judged by a single chip; in the AI era, the computing unit has expanded to racks and even clusters, and computing power must be measured by the token output efficiency and overall cost of the entire system.

He used the barrel theory to explain: system computing power depends on the shortest of computing power, storage capacity, transport capacity, and power, and if any link falls behind, even the strongest chip cannot perform to its full potential. Moreover, these "staves" are interconnected—heat affects electricity, electricity affects signals—and optimizing each in isolation can no longer solve the problem.

Therefore, Ling Feng believes STCO is imperative: design must start at the system level from the outset, target overall energy efficiency and cost, consider chips, packaging, boards, and racks together, and enable multi-physics issues such as electrical, thermal, mechanical, and optical to be solved collaboratively in the early design stage.

According to his introduction, Xpeedic, based on system-level design and multi-physics simulation capabilities, is introducing AI agents upward and extending simulation tools outward from chips all the way to complete machines, and has already promoted the implementation of STCO through multiple collaborations with industry chain partners.

Roundtable: STCO System Design in the AI Era

The morning roundtable "STCO System Design in the AI Era" was moderated by Dr. Dai Wenliang, President of Xpeedic, and featured Peng Li, CTO of MetaX; Chen Qiongnan, CPO of Luxshare Technology; Feng Ming'an, SVP of Murata; Zhou Jun, CTO of Wuhan Xinxin; and Tian Fujun, Senior Technical Expert of Longsys. The discussion covered key links in the AI hardware industry chain, including AI chips, storage, advanced packaging, high-speed interconnect, core components, and interconnect complete machines.

Guests from different positions in the industry reached a consistent judgment: in the AI era, local optimization is increasingly unable to guarantee system-level optimization. The interactions among chips, packaging, storage, interconnect, power supply, and heat dissipation are growing stronger. Chen Qiongnan, CPO of Luxshare Technology, mentioned that without system-level joint simulation, problems often only surface in later stages, and a single design change can cost tens of millions of yuan.

This is precisely where the value of STCO lies: enabling different links to collaborate earlier around the same system goal, and allowing models, data, and multi-physics simulation to truly run through the design process. As the design space continues to expand, AI will also take on more solution exploration and optimization work, but professional EDA and trustworthy physics simulation will still be needed for engineering verification. "System-level EDA lets us see the system from the chip, and AI for EDA lets AI truly enter the main design flow," Dr. Dai Wenliang concluded.

Mo Zhijian, Chief Engineer & Director of Lenovo Group, shared the practice of Lenovo and Xpeedic jointly developing EDA Agent 2.0 over more than a year. Mo Zhijian, Chief Engineer & Director of Lenovo Group, shared the practice of Lenovo and Xpeedic jointly developing EDA Agent 2.0 over more than a year. In the live demonstration, EDA Agent has moved from point assistance to multi-agent collaboration, from answering design specifications and undertaking library creation and placement and routing to checking while designing, covering multiple R&D links, and from reading device manuals to outputting DDR analysis reports, all completed automatically. This process increased design iteration efficiency by about 5 times and reduced manual operations by about 90%.

Jiang Liguo, Vice President of R&D at Xpeedic, released the EDA2027 version on site. In addition to the all-new 3DIC PDK and optoelectronic joint simulation platform, the more important change in this release is that AI for EDA is beginning to move from point functions toward platformization and productization. The Agentic AI super-agent platform AgentEvo is responsible for understanding design tasks, orchestrating engineering workflows, and invoking professional EDA tools, driving AI from assisting operations further into the real design flow. The Physics AI trustworthy physics intelligence foundation XAI uses AI for modeling, prediction, and optimization, expanding the design space engineers can explore. Both core AI capabilities are ultimately built on Xpeedic's system-level EDA and multi-physics simulation capabilities, forming a closed loop of "AI exploration, professional EDA verification."

Sub-forums: Real Engineering Practice of STCO and AI for EDA

English text is being prepared.

From this conference, it can be seen that along the two main lines of STCO and AI for EDA, Xpeedic's system-level EDA is continuously expanding the value boundary of domestic EDA, and also provides a new practical sample for Chinese EDA to enter AI computing infrastructure and next-generation intelligent system design.