The Global AI Hardware Battle: 117 Companies Compete, with 70% Based in China

Can AI Hardware Give Rise to New Industry Giants?

Should companies focus on technological advancement or market expansion? The emergence of new product categories such as AI glasses, AI Q&A devices, AI toys, AI wearables, and standalone AI terminals may offer a fresh direction for AI model developers who are struggling with the increasing competition in the field.

The Rise of Consumer AI Hardware

At the beginning of 2025, following the boom in AI models and AI applications, consumer AI hardware has started gaining traction among both the market and investors. Beyond smartphones and PCs, AI is now transforming the form and function of consumer electronics. As of now, at least 117 companies have entered the AI hardware sector, focusing on six key areas: glasses, headphones, education, companionship, assistants, and health. Notably, over 70% of these companies are based in China, with additional players from the U.S., Japan, South Korea, India, and Israel.

While most global tech giants remain cautious about entering the consumer AI hardware space, some exceptions exist. Outside of China, Ray-Ban and Meta successfully ignited an AI glasses trend. In China, ByteDance and Alibaba made brief attempts in the AI hardware market but did not go all in. Companies like Xiaomi, Samsung, and Google prefer to integrate AI into existing smart hardware rather than develop entirely new AI-centric devices. Meanwhile, Baidu is expected to launch its Xiaodu AI glasses this year, and Amazon and Tencent have yet to make significant moves in the consumer AI hardware sector. Amazon remains focused on AI chips and cloud services, while Tencent continues to build out its Hunyuan AI ecosystem.

Chinese Startups Driving the AI Hardware Wave

Among the 117 consumer AI hardware companies, around 30% were founded within the past five years, and 80% are based in China. This trend is closely tied to China’s well-established manufacturing and supply chain ecosystem. Several AI hardware startups have already achieved impressive sales figures. For example:

• Limitless’ AI Pendant

• Plaud.AI’s AI recording device, Plaud Note

• Leap Innovation’s AI toy, BubblePal

• Super Daddy’s AI Q&A device, Xiaochuang

Each of these products surpassed 10,000 units in sales shortly after launch, demonstrating the potential of AI hardware startups.

Currently, the 130 AI hardware products analyzed can be categorized into 13 subcategories, including AI Q&A devices, AI learning devices, language practice machines, AI robots, AI toys, AI wearables, standalone AI terminals, AI recording pens, AI/AR glasses, AI headphones, AI rings, and AI smartwatches. Among these, five categories—AI glasses, AI Q&A devices, AI toys, AI wearables, and standalone AI terminals—are entirely new product types, attracting the most startups. Meanwhile, other categories are dominated by major tech firms that simply enhance existing smart products with AI capabilities.

Industry insiders suggest that new entrants, especially in tech hubs like Guangzhou and Shenzhen, face challenges due to the token-based cost structure of AI services. Unlike traditional consumer electronics, AI hardware must account for ongoing expenses related to large language model usage, which disrupts the typical fast-paced development cycle of hardware supply chains in China. Despite these challenges, consumer AI hardware is still in its early stages and is expected to grow rapidly.

Three Approaches to Consumer AI Hardware

Currently, AI hardware development follows three distinct technical approaches:

1. Directly accessing cloud-based AI models via API – The simplest and most cost-effective approach, commonly used in voice and conversation-driven AI products.

2. Embedding powerful on-device AI models – Requires high computing power and results in higher costs. This approach is mainly used in AR glasses, which need strong visual rendering capabilities.

3. Relying on smartphones for AI processing – AI hardware connects to mobile devices via Bluetooth or other methods to leverage the phone’s AI capabilities, reducing costs. Many AI glasses, inspired by the success of Ray-Ban Meta, follow this model.

Recently, most popular AI hardware products rely on cloud-based AI models via API calls, especially those centered around speech and conversation functionalities. This provides an optimal balance between user experience and cost. However, as AI services improve and multi-modal AI models become more affordable, AI hardware relying solely on cloud models may become less competitive.

At present, chip manufacturers focusing on edge AI processing (on-device AI) have not benefited much from the consumer AI hardware trend. However, in the long run, as AI agent capabilities advance and costs decline, standalone AI devices with on-device processing power may gain greater market value.

For lightweight AI devices like headphones and smart rings, relying on smartphone AI capabilities is likely to remain the preferred approach due to weight and battery constraints. More AI products may emerge in the future as extensions of the smartphone AI ecosystem.

Can AI Hardware Create New Industry Giants?

Of the 117 AI hardware companies, 36 (approximately 30%) are startups founded within the past five years. These startups are primarily focused on AI glasses (10 companies), AI headphones (4 companies), AI education devices (6 companies), AI companionship devices (10 companies), and AI assistants (6 companies). Notably, there are no AI health hardware startups.

The presence of tech giants in a particular AI hardware segment often discourages new entrants. Many AI hardware startups strategically avoid product categories where smartphone giants already dominate, such as smartwatches and fitness bands.

The most successful AI hardware startups so far—Limitless, Plaud.AI, Leap Innovation, and Super Daddy—have all introduced entirely new product categories rather than competing head-on with existing smart hardware. For instance, Plaud Note found success by fulfilling a niche market need for iPhone-compatible AI recording solutions.

While these companies have the potential to become leading AI hardware brands, achieving dominance remains uncertain. The supply chain for AI hardware is still controlled by major smartphone manufacturers, which gives them a competitive edge in cost efficiency and distribution. Standardized AI hardware products may also face price wars from smartphone giants if they gain significant market traction.

However, AI hardware startups that carve out unique vertical niches still have the opportunity to become category leaders.

AI Hardware: A New Factor in the AI Model Wars

Among the 117 AI hardware companies, only a handful—including Meta, Google, ByteDance, Alibaba, Baidu, and iFlytek—have their own large language models. The vast majority of AI hardware companies rely on external AI model providers.

As these AI hardware startups grow in market influence, they could play a significant role in shaping the competitive landscape of the AI model industry. For instance, ByteDance’s Doubao and Amazon have been actively engaging with AI hardware companies to explore potential collaborations. While AI hardware products do not yet have large user bases, their user engagement metrics (such as time spent and monthly active users) are promising.

The AI model market remains highly competitive, with new entrants emerging every few months. For example, China’s DeepSeek O1 recently achieved top rankings in LiveCodeBench, matching OpenAI O1’s mid-tier reasoning performance. Just a few months earlier, DeepSeek was a relatively unknown player in the AI industry.

This intense competition forces AI model developers and their investors to make a strategic choice: focus on improving technology or expanding the market. The emergence of AI hardware—such as AI glasses, AI Q&A devices, AI toys, and AI wearables—could provide an alternative growth avenue for AI model companies that struggle with pure technological advancements.

Some AI model companies have already chosen to specialize in industry-specific AI applications. For example, Baichuan AI, led by Wang Xiaochuan, has targeted the medical field. However, most industry AI models have been integrated into traditional businesses rather than creating new AI hardware products.

Shenzhen and Guangzhou-based AI hardware startups recognize the potential of AI hardware but struggle with the token-based cost structure of AI services. Some industry professionals suggest a business model similar to how Tmall collaborated with white-label TV box manufacturers: startups profit from hardware sales, while AI model providers offer free AI services in exchange for user acquisition.

The key challenge, however, remains: how will AI model providers monetize users once AI services become free?

The Future of Consumer AI Hardware

As the consumer AI hardware market is still in its infancy, the future remains uncertain. Will major tech companies take a wait-and-see approach? Can startups establish themselves as industry leaders? Will new AI hardware form factors emerge?

Only time will tell.