Multiple Layouts in Edge AI and AI Agents; Mininglamp Technology (2718.HK) Accelerates Expansion into the Trillion-Dollar AI Market

On June 10, the "Implementation Opinions on the Innovative Development of 'Artificial Intelligence + Information Communications' (2026–2028)" released a highly significant bellwether signal: by 2028, more than 30 high-value typical scenarios will be formed, and a batch of typical applications and characteristic agents will be built.

Following the release of these opinions, Mininglamp Technology's (2718.HK) share price surged by nearly 50% over the two days of June 11 and June 12. The capital market may be repricing the strategic value of its Edge AI and AI Agents.

Edge AI: From "Concept" to "Trillion-Dollar Industrial Blueprint"

The market narrative of Edge AI no longer remains at the speculative stage; it has already manifested as concrete products, propelling the entire industry into a trillion-dollar scale of development.

Data from LeadLeo Research Institute shows that the scale of China's Edge AI industry was less than RMB 200 billion in 2023 and is expected to break through RMB 1.9 trillion by 2028, exhibiting a compound annual growth rate (CAGR) as high as 58% from 2023 to 2028. Concurrently, Frost & Sullivan forecasts that from 2025 to 2029, the global Edge AI market scale will leap from RMB 321.9 billion to RMB 1.22 trillion, with a CAGR of 40%.

The force driving this growth comes from the dual resonance of policy and market. On June 10, the "Implementation Opinions on the Innovative Development of 'Artificial Intelligence + Information Communications' (2026–2028)" issued by the Ministry of Industry and Information Technology (MIIT) explicitly proposed to "vigorously develop products such as AI smartphones, AI PCs, smart home devices, and intelligent wearables."

Subsequently, the national authorities' "Artificial Intelligence +" Action Opinions further set a quantitative target—by 2027, the penetration rate of the new generation of smart terminals will exceed 70%. Extrapolating from this penetration rate, the scale of Edge AI devices will reach 1.05 billion units in 2027 alone, corresponding to a hardware market of over RMB 500 billion.

TF Securities explicitly pointed out that 2026 is expected to usher in a grand year for Edge AI, with the deep integration of computing power and applications driving the rapid expansion of the ecosystem. Soochow Securities similarly judged that 2026 is welcoming a critical inaugural year for the large-scale volume release of edge-side AI, and Edge AI, as a brand-new C-end traffic entry point, possesses ample explosive power.

Since the beginning of 2026, Mininglamp Technology has executed precise and rapid layouts in large models and other related fields of Edge AI.

In May 2026, Mininglamp Technology intensively open-sourced its edge-side GUI-VLA agent model Mano-P and inference acceleration framework Cider. The 72B version of Mano-P ranked first globally on the OSWorld specialized model leaderboard with a success rate of 58.2%, while its 4B quantized version achieved a prefill speed of 476 tokens/s on Apple M4 Pro. Under the W8A8 mode, Cider's operator speed improved by 1.4 to 1.9 times compared to native MLX.

With OpenClaw natively entering Windows, the Mano-CUA skill has been integrated as an OpenClaw ecosystem skill, achieving edge-side Agent implementation capabilities across both macOS and Windows dual platforms.

Wu Minghui, founder of Mininglamp Technology, summarized this path as "Scaling Out"—instead of feeding data into a centralized brain, it allows intelligence to grow in a distributed manner right where the data is generated.

This judgment also echoes Jensen Huang's keynote at the GTC conference, which defined that AI has officially transitioned from the large language model stage into the era of Agentic AI, characterized by the capacity to autonomously observe, reason, plan, and invoke tools.

AI Agents: A Tens-of-Billions Track Racing Toward Three Hundred Billion

In an AI-driven economy, what exactly is a company's core asset?

In mid-June 2026, Microsoft CEO Satya Nadella gave his own answer. He believes that a company's core competitiveness in the AI era is not determined by which large model it chooses, but rather by whether it can build a "learning closed loop" atop the models where human capital and token capital compound and grow together.

From the current perspective, Mininglamp Technology's layout in intelligent agents since 2026 has been equally forward-looking.

Public disclosures indicate that Mininglamp Technology has constructed a three-tier product matrix spanning from models to platforms to hardware. Among them, the open-source Agent collaboration platform Octo is the world's first open-source, trustworthy Agent collaboration network. It has already been deployed internally among approximately 1,400 employees, with over 2,900 AI Agents running efficiently in actual workflows—making the number of Agents more than double the number of employees. Octic, the AI Native hardware released in May 2026, completed the software-hardware integrated closed loop from perception to processing to collaboration.

According to the White Paper on AI Agents Empowering Industry Decision-Making: Trends and Practices (2026) released by Kezhi Consulting, the market scale of China's enterprise-grade AI agents was RMB 8.6 billion in 2024, jumped to RMB 21.2 billion in 2025, is expected to reach RMB 44.9 billion in 2026, and is poised to break through RMB 332 billion by 2029, showcasing a CAGR as high as 107%.

IDC predicts that 2026 and 2027 will be the two years with the fastest growth rate for the number of active agents in Chinese enterprise scenarios, with a year-on-year growth exceeding 200% in a single year, and will reach 350 million active agents by 2031.

The global perspective is equally optimistic. Deloitte predicts that the global agentic AI market scale is expected to reach USD 45 billion by 2030. Gartner expects that by 2035, agentic AI will generate up to USD 450 billion in market revenue opportunities. Goldman Sachs further calculates that by 2030, consumer-end and enterprise-end AI agents combined will drive global token consumption to grow by 24 times compared to 2026 levels.

Forward-Looking Layout in the Second Half of AI; Accelerating the Commercial Implementation of AI

The transformation of this technical form, AI Agent, is directly transmitting to the pricing logic of business models. Nadella explicitly pointed out that a company can outsource a certain task or even a certain position, but it can never outsource its own "learning capacity."

In response to this trend, explorations at the Chinese application layer appear more concrete. Mininglamp Technology founder Wu Minghui believes that future software services might evolve into a billing model of "token cost plus management fee." This model resembles the logic of traditional advertising agencies helping clients manage their ad placement budgets and charging a fixed percentage as a service fee. When AI applications cross the pure tool stage, the valuation logic for service providers might also transition from traditional outsourcing headcount fees to the long-term value of helping enterprises build and operate infrastructure.

On the commercialization level, Mininglamp Technology's Agentic Services generated over RMB 100 million in its first year in 2025, covering four major scenarios: performance advertising, brand advertising, content e-commerce, and micro-drama production.

Since 2026, in addition to delivering products in edge-side AI large models and hardware devices, Mininglamp Technology has also been accelerating the broadening of its commercial application scenarios.

On June 12, Mininglamp Technology and PATEO signed a strategic cooperation framework agreement. The two parties will integrate Mininglamp Technology's capabilities in AI large models and AI Agent suites with PATEO's resources in edge-side hardware entry points, AI computing power infrastructure, and in-vehicle application scenarios. Utilizing token economics as the underlying commercial logic, the platform aims to open up a complete closed loop of "brand marketing budget—vehicle-side AI computing power consumption—user scenario-based consumption."

Meanwhile, the scale of China's intelligent connected vehicle track reached RMB 455 billion in 2025 and is expected to grow to RMB 1.56 trillion between 2026 and 2030. This cooperation signifies an extension of Mininglamp Technology's edge-side AI capability from enterprise office scenarios into this trillion-dollar market of intelligent connected vehicles.

Furthermore, Mininglamp Technology also signed a memorandum of strategic cooperation with The Education University of Hong Kong in early June to jointly build an "AI + Education" joint research platform, opening up AI tools such as DeepMiner, Octo, and CoCraft to extend edge-side AI capabilities into the fields of education and scientific research.

China Galaxy Securities pointed out that as large model inference capabilities, multimodal technologies, tool invocation, and multi-agent collaboration continue to mature, AI is evolving from a content generation tool into a "digital employee" endowed with autonomous planning and task execution capabilities. The key to future competition will no longer merely be model performance, but rather who can take the lead in constructing an Agent ecosystem encompassing models, tools, data, and scenarios, while driving the deep embedding of AI into real-world production and daily life scenarios.

Over the past half-year, Mininglamp Technology's path from the open-sourcing of Mano-P and Cider, to the implementation of the Octo collaboration platform, and further to the launch of Octic hardware and the strategic cooperation with PATEO Connect+, constitutes a clear progression chain of technology-product-commercialization.

Under the trend where Edge AI transitions from a "concept" to a "trillion-dollar industrial blueprint" and AI Agents race from a "tens-of-billions track" toward a "three-hundred-billion market," the company is validating a core proposition through its differentiated path of "Scaling Out." In the second half of AI, it is not about who possesses the largest model, but rather who can transform AI into a "digital workforce" that can truly do the work, collaborate, and deliver results.

NewTimeSpace Disclaimer: All content herein is the original work of NewTimeSpace. Any reproduction, reprinting, or use of this content in any other manner must clearly indicate the source as "NewTimeSpace". NewTimeSpace and its authorized third-party information providers strive to ensure the accuracy and reliability of the data, but do not guarantee the absolute correctness thereof. This content is for reference only and does not constitute any investment advice. All transaction risks shall be borne by the user.