Zhang Yiming und Altman müssen beide von Zuckerberg lernen.

"The current form of computers does not match the sci - fi 'AI companion' we imagine. The truly suitable form is yet to be explored, but ultimately, what matters will be the hardware forms that can continuously perceive the environment and maintain natural interaction with humans." Last month, OpenAI CEO Sam Altman said this in a podcast.

Altman's view has been reflected in OpenAI's business layout. This leading model company is putting 'hardware development' on the agenda.

First, in May this year, the io Products team founded by Apple's legendary designer Jony Ive joined OpenAI. And last week, a report from The Information revealed that OpenAI has signed a contract with Luxshare Precision and contacted Goertek, another assembler of Apple products, hoping that the latter will provide components such as speaker modules for its future AI product series. It is understood that this mysterious AI device is "similar to a smart speaker without a display screen." OpenAI is also considering developing glasses, digital recorders, and wearable brooches. The first batch of products is planned to be launched at the end of 2026 or early 2027.

On the other side of the ocean, Zhang Yiming and ByteDance are also reorganizing their hardware strategies. Since this year, multiple reports have shown that ByteDance is planning to launch self - developed AI smart glasses, which are expected to integrate the AI functions of ByteDance's self - developed model 'Doubao'. This will be ByteDance's second AI hardware product after the AI earphone OlaFriend. According to a disclosure by Digitimes last week, ByteDance is close to launching a prototype of the AI glasses.

In addition, recent reports show that ByteDance is also making continuous efforts in the robotics field. After mass - producing more than 1,000 wheeled logistics robots (AMR), ByteDance is expanding the scale of its Seed team. The recruitment positions specifically mention 'next - generation general robots'.

Altman and Zhang Yiming, these two'software experts' from Silicon Valley and Zhongguancun, have both set their sights on the consumer - grade AI hardware market. However, before them, another'software expert', Mark Zuckerberg, has already presented a phased answer: the Meta Ray - Ban series of AI glasses. This series of products, which 'looks like ordinary sunglasses', has become the top - selling AI glasses in the world. Recently, Meta also launched the new - generation Ray - Ban Display with colored in - lens displays and an electromyography wristband, priced at $799, which is equivalent to the price of an iPhone 17.

According to data previously disclosed by EssilorLuxottica, the parent company of Ray - Ban and Meta's partner, more than 2 million pairs of Meta Ray - Bans have been sold, and the goal is to produce 10 million pairs of Meta glasses annually by the end of 2026. While other model companies are still planning how to enter the market, Meta has turned this track from a prototype into a'real business'.

A

In May this year, io, a startup founded by famous designer Jony Ive, was acquired by OpenAI. Ive is a legendary designer at Apple and led the design of classic products such as the iPhone and iMac. Ive himself will not join OpenAI's staff. His design company, LoveFrom, will continue to operate independently, but LoveFrom will 'take over all of OpenAI's design work'.

After the acquisition, Altman said that he envisioned collaborating with Ive to create a 'family of devices'. The first device is said to be small in size, context - aware, and screenless. This also means that OpenAI's first AI hardware will not be glasses.

On the other hand, OpenAI is also seeking to jointly create AI hardware products with some ecological partners. In June this year, Mattel, the manufacturer of Barbie dolls, announced a strategic cooperation with OpenAI to jointly develop toy and game products based on AI technology. The first AI product is expected to be launched later this year.

While engaging in 'co - branding' cooperation, OpenAI has also started to form an internal hardware team and has poached some key members from Apple's hardware team. It is understood that OpenAI's hardware team is led by Tang Tan, a former Apple hardware executive, and has recruited a group of backbone members from Apple's hardware team.

In September, OpenAI's hardware plan is getting closer to implementation. Last week, a report from The Information revealed that OpenAI has reached a manufacturing cooperation agreement with Luxshare Precision and has contacted Goertek regarding acoustic modules. Both of these companies are well - known members of the 'Apple supply chain'. OpenAI's first hardware product may carry Apple's DNA from the start. However, hardware manufacturing requires a large amount of upfront investment in tools, supply chain management, quality control, and inventory management. Although OpenAI has been laying out the 'Apple supply chain' and recruiting Apple team members, integrating these capabilities and links still takes time.

In addition, although the team members are highly capable, the specific form of OpenAI's first hardware product is still unclear. The description of this product remains at what Altman previously disclosed,'similar to a screenless smart speaker'. Altman and Ive seem to have more confidence in this product form than in AI glasses.

In July this year, Altman was publicly interviewed by the media while wearing a pair of white sunglasses. When asked, 'Are you wearing smart glasses? Is this a smart glass?' he replied, 'No, definitely not. I don't like smart glasses.' Although it was an informal statement, it seems that this CEO is not a big fan of AI glasses. By clearly excluding glasses as the form of its first product, OpenAI has seemingly forced itself into the niche market of'speakers'.

Previously, well - known analyst Ming - Chi Kuo further revealed information about this screenless device. He said that the prototype is equipped with a camera and a microphone and is expected to enter mass production in 2027. He carefully used the word 'expected', seemingly implying that there are still many unknowns about this hardware product.

This somewhat stubborn product concept has made some industry insiders highly skeptical about OpenAI's hardware plan, especially after the'screenless' form was determined. In fact, OpenAI's greatest advantage lies in ChatGPT's ability to process and generate rich multimodal content, and screenless devices largely limit the presentation of this content.

'What problems can a screenless AI companion solve that a smartphone cannot?' After covering the consumer technology field for five years, I still can't answer this question. OpenAI obviously can't either. Industry analyst Derick David wrote in an analysis.

Obviously, this AI product, which may not be available until a year later, does not plan to provide any display function. Such a product logic seems a bit 'outdated'. In the analysis, Derick David said that in 2025, creating a consumer technology product without a screen is like 'building a car without windows'. From the very beginning, the device's functions and user experience are deliberately restricted. 'I believe that if Steve Jobs were still alive, he would definitely strongly promote an AI/VR glasses product.'

However, OpenAI is still very confident about its future hardware product plan. Altman previously claimed that this could be 'the coolest technology in the world'. With the support of design guru Ive, OpenAI hopes that this product can become a revolutionary product like the iPhone, even though it may only be the size of an iPod and has no screen.

On the other side of the ocean, Zhang Yiming and his hardware team have already tried AI hardware such as earphones and speakers. ByteDance launched its first AI earphone, OlaFriend, last year, but the product received a lukewarm response and did not become the game - changing product that ByteDance had hoped for. The feedback from the AI glasses market also shows that the screen - based/display solutions are becoming the mainstream in the third quarter of 2025. Companies such as Meta, Alibaba, and Rokid have all introduced mature 'waveguide display + AI' solutions.

So, will ByteDance, which owns the VR manufacturer PICO, copy Meta's approach in AI glasses?

B

For ByteDance and Zhang Yiming, hardware is not an unfamiliar field. Since the early days of the large - scale learning lamp and PICO, ByteDance has been trying to integrate more hardware capabilities. In the AI hardware market, although Meta has captured a large share of the global AI glasses market, the domestic AI glasses market remains a blue ocean due to regional policies.

In fact, before the disclosure of ByteDance's AI glasses plan, this internet giant had already created its first AI wearable hardware, the AI smart earphone OlaFriend, by acquiring the earphone manufacturer Oladance. This product became ByteDance's first wearable AI smart device, but it did not receive a strong market response and failed to become the game - changing product that ByteDance had expected. On the other hand, ByteDance has a foundation in screen - based hardware. In fact, with the accumulation from its PICO business, ByteDance has good technical capabilities in the VR/MR field. However, in the AI era, it needs to figure out how to avoid the problems that PICO faced.

In August 2021, ByteDance spent billions of yuan to acquire PICO, the leading VR manufacturer in China. Although PICO has long held the top position in the domestic VR market, the VR game ecosystem in China has faced some challenges. The high cost of VR game development and introduction, combined with the low willingness of domestic users to purchase single - player games, has made it difficult to form a large - scale user base. Eventually, in 2023, PICO scaled back its business due to lower - than - expected sales. Its team of nearly 1,600 people was reduced to a technical team of only 200.

In terms of market share, PICO is not a failed business. However, with the slow growth of the entire VR market, PICO cannot meet ByteDance's ambition to expand its ecosystem in the hardware market. In the smart wearable market, head - mounted MR/VR devices are too heavy and large to be 'everyday' items. In the traditional entertainment market, they also face competition from other game consoles. As a result, although PICO ranks first in domestic sales, it is still difficult to shake off the label of being a 'niche product'.

A horizontal comparison shows that the hardware development paths of Meta and ByteDance are somewhat similar. Both companies are internet companies that started as content communities. Meta acquired Oculus VR in 2014 and then entered the VR hardware market with its Quest series, just as ByteDance acquired PICO. Comparing the current weakness of PICO with Meta's success in AI glasses, learning how to develop hardware without relying solely on'software thinking' may be one of the most important lessons for ByteDance.

As early as February this year, there were rumors in the industry about ByteDance's AI glasses solution, which uses the Airoha 2800 + Yanjiwei ISP chip to meet the imaging and image - processing requirements of AI glasses. This product solution does not have a display function and focuses on voice interaction and shooting, mainly competing with the Meta Ray - Ban and the Xiaomi AI glasses launched a few months ago. However, a report from The Information in July showed that ByteDance is following Meta's technical solution and developing a lightweight mixed - reality (XR) glasses, similar in appearance to goggles, competing with Meta's similar products.

It is understood that PICO is collaborating with suppliers to develop this device. PICO had previously considered developing a high - end headset similar to the Vision Pro, but as Meta shifted to lightweight devices to increase the accessibility of the mixed - reality experience, PICO also made corresponding adjustments.

Obviously, ByteDance is much more proactive than OpenAI in learning from Meta.

For both OpenAI and ByteDance, there is a current shortcoming that is difficult to address - offline channels. Whether it is AI speakers or AI glasses, the 'last mile' of ToC hardware products requires offline try - ons and services. Especially for wearable hardware, users are more likely to make a purchase after experiencing the product offline.

In this regard, Meta did not choose to build its own stores from scratch. Instead, it has a deep partnership with Ray - Ban, integrating try - ons, eyewear fitting, and after - sales services into Ray - Ban's existing retail and fitting network. Consumers can try on the products, get their glasses fitted, and receive after - sales services at these stores.

However, the key to Meta's success lies in the precise positioning of the product form of the Meta Ray - Ban series. This series prioritizes core functions such as video shooting, calls, and external audio playback in a lightweight product form, and also integrates AI in the second - generation products. This strategic plan, which only provides 'adequate' functions, may seem 'lacking in ambition', but it has contributed to Meta's current hardware market position.

C

Since the wave of large - scale models swept the world in 2023, some new forms of AI hardware have been launched. The once - impressive AI Pin and Rabbit R1 both claimed to revolutionize the way we interact with AI. However, in actual use, they are not significantly different from AI apps on smartphones.

To summarize Meta's hardware strategy for its successful AI glasses in one sentence, Meta does not aim to replace existing smart hardware but rather integrates AI into 'ordinary products'.

Looking back at the success of the Meta Ray - Ban series, a key factor is its appearance, which is widely acceptable to the public. Instead of developing the entire product design independently, Meta partnered with EssilorLuxottica, the world's largest eyewear manufacturer. The Meta Ray - Ban series retains more 'fashion sense' than 'technological sense', maintaining the basic attribute of glasses as a daily fashion item. The glasses are intelligentized through unobtrusive cameras, microphones, and the Meta AI voice assistant.

After establishing a foothold in the smart glasses market, Meta carried out a new - generation technology upgrade at Connect 2025. The new - generation Ray - Ban Display integrates an in - lens display and a wristband. The color HUD in the right lens is used for information display and subtitle translation, and the new Neural Band changes the interaction logic of the glasses. Gesture and micro - motion input further enhance the experience of 'using AI quietly in public'. The previous generation of AI glasses relied mainly on voice interaction.

However, Zuckerberg's demonstration at Connect 2025 encountered some problems. Two functions demonstrated on - site, using the camera to check food ingredients and making a WhatsApp call, did not work smoothly. Meta CTO Andrew Bosworth later explained that these issues were caused by voice commands triggering all devices at the scene to simultaneously request the server and a bug that prevented notifications from being displayed when the screen was in sleep mode. Fortunately, these problems do not seem to be related to the hardware design, and Meta later announced that it had fixed the functions.

On the other hand, the $799 price of the Ray - Ban Display has also been questioned by some industry insiders. As a daily - use head - mounted glasses product, a price comparable to that of an iPhone seems a bit too high.

However, as the leading company in the industry, Meta is still confident in its products. IDC predicts that the combined shipments of AR/VR and non - display smart glasses will increase by 39.2% in 2025, and Meta is considered the most important contributor to this growth. It is understood that Meta invested $3.5 billion this year to acquire a 3% stake in EssilorLuxottica, the parent company of Ray - Ban. The company's previous financial report showed that its revenue reached a new high of $16.25 billion, and the sales of its product, the Ray - Ban Meta, exceeded 2 million pairs.

Looking back at Zuckerberg's hardware strategy, we can see that Meta did not insist on creating a 'geeky' product. Instead, it partnered with EssilorLuxottica to integrate AI functions into a well - designed product form.

On the other hand, Zuckerberg's keen sense of the market has given Meta an early advantage in this 'hardware' transformation. ByteDance and OpenAI, which only started to take action this year, now face not only competitors like Meta but also traditional hardware giants such as Apple