On October 27, NVIDIA announced that it will invest $3 billion (about RMB 200 billion) in Meta’s share issuance, and that the two companies will expand their cooperation on Muse, Meta’s personal AI agent, and the Muse family of models, to improve inference and training performance. More than 10 million people have already set up the Muse personal agent.
The investment is part of a much larger Meta financing plan. On October 26, Meta announced plans to raise $95 billion (about RMB 633.5 billion): first $35 billion through an underwritten public offering, followed by up to $60 billion via an at-the-market issuance program. Beyond NVIDIA, AMD and Broadcom have also signaled their intention to invest.
Viewed from the industry landscape, a chip giant taking a direct equity stake in a foundation-model company marks a new depth in the binding of AI compute and model ecosystems — model vendors lock in advanced compute supply, while chip vendors trade capital for ecosystem influence. Even more noteworthy is the “personal agent” form factor: as tens of millions of user-level agents come online, inference workloads will shift from the cloud toward devices and the edge, further elevating the strategic value of on-device compute and edge infrastructure.
For Junyi Microelectronics, the scaled landing of personal agents is exactly the on-device AI trajectory we focus on. The AI-BOX-3588 and 5G-EDGE-3399Pro edge computing hosts target localized AI inference, the K-series edge AI boards and the J-3568/J-3576/J-3588 core boards provide a compute gradient for intelligent terminals from entry level to flagship, while the IPC-i5-1235U industrial mini PC and touch panel all-in-one machines cover interaction and deployment form factors. As agent applications spread across industries, the demand curve for edge compute hardware is being redrawn.
Source: https://k.sina.com.cn/article_5952915720_162d24908067052kq4.html (2026-10-27)