Cover Image | Created by Pencil News

OpenAI has acquired AI imaging company Glass Imaging for over $300 million (approximately 2.1 billion RMB).

Glass Imaging leverages AI to enhance the performance of small cameras, enabling them to "shoot more clearly." Its core technology, GlassAI, has already been integrated into Honor's 600 series smartphones.

"For AI companies, venturing into hardware is almost a prerequisite. Whether it's for data acquisition or application deployment, hardware is ultimately essential—it provides a more direct connection to users," Li Yachong, founder of Putaoteng Technology, told Pencil News.

As competition intensifies for hardware entry points, the demand for superior tools grows.

- 01 - The Technology is Already Generating Revenue

Founded in 2019, Glass Imaging was co-established by Ziv Attar and Tom Bishop, both former Apple employees who contributed to the development of the iPhone's Portrait mode technology.

Smartphone manufacturers have recently faced a conundrum: devices must become thinner and lighter, yet camera quality must improve annually.

Traditionally, the solution has been to add more hardware. Glass Imaging, however, took the opposite approach: since hardware cannot be infinitely stacked, they focused on making the software more intelligent.

GlassAI processes raw data from the camera's sensor and trains neural networks tailored to specific lenses and sensors to correct noise, lens aberrations, and detail loss.

This approach differs somewhat from conventional "AI photo retouching."

Glass Imaging aims to restore the information truly captured by the camera, rather than relying on generative AI to "imagine" textures that do not exist.

The Honor 600 series serves as a prime example.

This smartphone lacks a dedicated telephoto lens, instead relying on cropping from a 200-megapixel main camera for zoom functionality. GlassAI restores details that are typically lost during cropping, while also reducing noise and enhancing texture.

For smartphone manufacturers, if software can replace some expensive and space-consuming optical hardware, the same body space can accommodate fewer lenses or achieve effects that previously required more complex hardware, using smaller cameras.

Glass Imaging's integration into Honor's smartphone lineup at least demonstrates that it has overcome one of the toughest challenges: mass production.

Thus, OpenAI's valuation of Glass Imaging at over $300 million reflects the technology's proven ability to enter the consumer electronics supply chain.

Despite being founded more than seven years ago, Glass Imaging has maintained a small, agile team, currently totaling just 37 employees.

- 02 - OpenAI is Vying for Entry Points

OpenAI is not lacking in AI expertise; it aims to bring that expertise into the physical world.

In 2025, OpenAI spent $6.5 billion to acquire io, a hardware company co-founded by Apple design guru Jony Ive. This year, reports surfaced that OpenAI's first consumer hardware product might be a portable, screenless smart speaker equipped with cameras and multiple sensors to perceive its surroundings.

This is where the significance of Glass Imaging becomes apparent.

In the ChatGPT era, humans actively share their world with AI by uploading images or documents for the model to interpret.

If AI is to become a hardware device that remains with users long-term, it must be able to see and hear independently.

For instance, a user might point to medicine on a table and ask, "How do I take this?" or wear glasses and inquire, "What building is in front of me?" or ask AI to check if a machine's dashboard displays any abnormalities.

No matter how intelligent the model is, a prerequisite exists: it must first see clearly.

Li believes the true value of hardware lies in serving as an "entry point." Currently, ChatGPT primarily interacts with users through phones and computers. Once OpenAI has its own devices, it can directly engage with users in high-frequency life scenarios.

This strategy has already proven successful. Meta's AI glasses, developed in collaboration with EssilorLuxottica, exemplify this point.

In the first half of 2026, EssilorLuxottica revealed that sales of its AI smart glasses, developed with Meta, nearly doubled year-over-year. This year, Meta introduced a new generation of AI glasses starting at $299.

Once OpenAI secures a new entry point for continuous user interaction and environmental information acquisition, Glass Imaging's role is to ensure AI sees more clearly. If OpenAI only sought to enhance a single product's camera, it could simply license GlassAI.

By acquiring the company outright, OpenAI gains the entire team, core intellectual property, and the ability to jointly optimize future products around its own cameras, chips, and form factors.

This $300 million investment is about preemptively enhancing the visual capabilities of next-generation AI hardware.

- 03 - AI Hardware is Prone to Failure

However, OpenAI's acquisition of Glass Imaging does not imply that startups should rush into AI hardware now.

In fact, many have already learned this lesson the hard way in recent years.

Humane was once one of Silicon Valley's most talked-about AI hardware startups.

Its founding team also hailed from Apple, and its AI Pin aimed to replace some smartphone functions with a screenless device that could be pinned to clothing. Initially priced at $699, the company raised approximately $240 million in funding.

But the AI Pin received overwhelmingly negative reviews after launch, leading Humane to halt sales in 2025 and sell core assets—including CosmOS, patents, and part of the team—to HP for just $116 million, less than half of its total funding.

Many sold AI Pins subsequently lost major functions like cloud AI, calls, and messaging.

On one hand, Meta's glasses nearly doubled in sales; on the other, Humane sold off its company. The difference lies in the problems the products actually solve.

Li believes the truly promising aspects of AI hardware may not lie in "perception layers" like cameras and sensors but rather in "application-level hardware" designed for specific needs.

AI glasses are one example; companionship hardware is another. In the future, dedicated devices for office work, education, healthcare, or industry could emerge, each designed around a specific high-frequency need.

But identifying use cases is also the toughest challenge.

"Hardware requires supply chains, upstream and downstream partnerships, distribution channels, and capital. Inexperienced entrepreneurs simply can't compete," Li said.

Meta's glasses succeeded partly because they did not start from scratch: Ray-Ban already had brand recognition, an eyewear supply chain, and global sales channels.

Humane, by contrast, demonstrated how difficult it is to create a new hardware category from scratch.

Algorithms may not offer easy entrepreneurial opportunities either.

Li believes that as foundational AI capabilities grow stronger, gaps between pure algorithm offerings could quickly narrow. Meanwhile, components, supply chains, and distribution channels are strengths cultivated by traditional hardware manufacturers over time.

Thus, for ordinary startups, a more realistic path may not be to "see AI hardware is hot and immediately build a complete device" but rather to first enter the industry as a participant and then find their niche in algorithms, supply chains, components, distribution, or specific use cases.

OpenAI has the financial resources to spend $6.5 billion on a hardware team and another $300 million on a pair of "eyes." Other startups could still become acquisition targets for industry giants.

This article does not constitute any investment advice.