Latest observation: AI screens are driving up display output value

Since 2016, the display industries of both LCD and LED technology lines have been looking for the same key - the key to unlocking the ceiling of output value.

LED display companies are using MLED to target the 200 million TV market, trying to leverage this massive stock; the LCD camp is also using MLED to set its sights on the LED display market with higher gross margins. Both sides are trying to use the same technology to penetrate the other side's deeply cultivated territory.

As a result, the two camps that were originally dominant on one side began a long two-way penetration.

Ten years have passed and no one has lost, but no one has won yet.

The reason is not complicated: both parties are competing in the stock market, competing for parameters, spacing, and brightness, but they have not created essential service differences. Existing products in the large-screen market have diminishing marginal effects as more parameters are added; the small- and medium-sized screen market is occupied by LCD and OLED, and the TV market seems large, but the path is full of fog.

The plot is still unfolding, but the narrative is no longer exciting.

Until AI evolves.

At the beginning, most people believed that the most pragmatic aspect of AI is nothing more than empowering image quality and energy saving.

But recently, experts say that Display and Mofa Technology CEO Professor Chai Jinxiang, TPV Commercial Display and Hotel TV China General Manager Wu Guangfa, Unilumin Technology Group Vice President Mei Zhimin, and Hengshi Technology (State Grid Software Supplier) CTO Qian Zhengyang are four industry veterans. After having an in-depth exchange of opinions on AI screens, we found that AI screens have spawned application scenarios that did not exist in the past. AI has not changed the hardware of the screen, but has redefined the role of the screen: "white-collar" positions that used to rely on real people will have the opportunity to start being filled by a screen.

The stock market is no longer a burden of production value, but a mineral deposit waiting to be activated - because this is not a parameter competition on the old battlefield, but a revaluation of values on the new battlefield.

Inflection point: AI screen redefines the value of "display"

In this in-depth exchange, experts said that Display discovered that in the past, the screen was just a piece of hardware - its value depended on size, image quality and brightness, and it was mainly "viewed". It played advertisements, displayed information and presented pictures - you asked it questions, it was silent; you talked to it, it was indifferent. No perception, no understanding, no response.

Now, when AI capabilities are injected, large models give it the ability to understand and generate, digital human technology gives it interactive images and sounds, and breakthroughs in edge computing power give it low-latency real-time response. A once passive display terminal is evolving into a service carrier that can complete the closed loop of "perception-understanding-expression".

This is not an evolution of parameters, but a reconstruction of roles: the AI screen essentially puts service personnel into the screen - turning a screen into a staff member. Work that used to be done by "white-collar workers" - information provision, communication and coordination, service guidance, promotion explanations, etc. - can now be undertaken by AI screens.

Therefore, wherever "face-to-face service between people" is required, there may be an explosion of AI screens.

The company's budget anchor may also shift from "how much money per square meter" to "replacement or replacement labor costs." This change is the real starting point that shows that the logic of industrial valuation has been rewritten.

Once these new scenarios are verified to be effective, large-scale replication will begin, which means that the value of display screens will be completely transferred - display companies can truly shift from one-time income from "selling hardware" to continuous income from "selling services", which in turn will drive new demand for screen hardware.

The two are independent of each other, but also reinforce each other: the better the stock is activated, the greater the imagination space for new scenarios; the more new scenarios are implemented, the stronger the motivation to upgrade the stock - which may eventually generate service revenue of the same scale or even greater, thus completely breaking the display industry output value ceiling.

So the imagination of the AI screen is that it will create scenes that did not exist in the past, instead of competing for "who displays better" in the existing scenes.

From the perspective of the AI world, AI screens may be the best carrier of embodied intelligence in recent years.

But this judgment will naturally lead to an unavoidable question: Why is the best carrier of embodied intelligence not a humanoid robot, but a screen?

Question: Why a screen and not a humanoid robot?

When talking about embodied intelligence, humanoid robots are often the first carriers that come to mind. This intuition makes sense, but it may not be complete.

The answer lies in the division of labor.

Humanoid robot: It solves the relationship between "people and things". Handling, operation, manufacturing, cleaning, and sorting—whether it is production line operations in industrial scenarios or housework in future families, the core challenge lies in physical interaction technologies such as joint control, sensor fusion, and power systems. Tesla Optimus and other products have left the laboratory and entered factory trials, and will clearly extend to home scenarios. However, from technical verification to large-scale commercial deployment, the dual thresholds of cost and stability still need to be overcome.

AI screen: It solves the "people-to-people" relationship. Information provision, communication and coordination, service guidance, promotion explanation - it does not need to grab objects, it only needs to complete the closed loop of "perception, understanding, and response". This is the most mature capability of current AI technology and the closest to commercial realization.

One is dedicated to the operation of things, and the other is focused on serving people. Both are equally important, but have different evolutionary paths.

Humanoid robots are still overcoming the problem of "how to physically interact", while AI screens are already capable of the task of "how to provide services". In the next 3-5 years, the large-scale implementation of the former still needs to break through multiple bottlenecks, while the latter can directly access AI on the existing ubiquitous screens to achieve batch employment.

There are two types of embodied intelligence, one serving blue-collar jobs and the other serving white-collar jobs. One still takes time, and the other is taking the lead.

After pursuing this point, another question surfaced: Since the stock screen can open new scenes by connecting to AI, why didn't it work in the past, but now?

Turnaround: Reconstruction of the underlying logic of human-computer interaction

The answer does not lie in the screen hardware or AI itself.

In the past decade, the brightness, contrast, and resolution of screens have all improved, but these evolutions have not changed the essence of the screen—it is still a "passive information output device."

In addition to screen hardware, interactive technologies such as large models, digital people, and voice are also developing rapidly. However, even so, AI screens have still not been implemented on a large scale.

Because the problem is not that "big models aren't smart enough" or that "digital people aren't realistic enough". But the point is: if an AI screen can truly enter real scenes such as hospitals, government halls, shopping malls, and exhibition halls, it will face much more than just "displaying a digital person."

It needs to complete the recognition of the user's voice and behavior in milliseconds, understand the intention, generate an answer, and synchronously drive the digital human's voice, mouth shape, expression, movement and picture presentation. At the same time, it is also necessary to ensure long-term stable operation, high-frequency interaction between multiple people, adaptation to different terminals, and controllable costs after large-scale deployment.

In other words, the real bottleneck of AI screens is not whether there is a breakthrough in a single point of capability, but whether the entire real-time driver system can simultaneously meet low latency, low cost, high concurrency and lightweight deployment.

In the past, the reason why many digital human projects stayed in the "project-based" and "customized" stages for a long time was not because they could not produce interactive demonstrations, but because once they entered large-scale real service scenarios, real-time drive, computing power consumption, system stability, and terminal adaptation would all become unavoidable thresholds for commercialization.

The large model solves the problem of "understanding and generation" - it allows AI to understand human speech and figure out how to answer. But what the industry is really missing is another capability: how to allow AI to "perform" answers in a real-time, natural and stable manner, and to deploy them on thousands of screens at a low enough cost.

The breakthrough of Mofa Technology revolves around this issue.

What it provides is an end-to-end embodied interaction solution for screen manufacturers and scene customers. But what really supports its large-scale implementation is not just a certain capability - but the integration and reconstruction of the entire system.

The technical indicators brought by this set of capabilities are: driver delay within 500 milliseconds, end-to-end interaction within 1.5 seconds, and support for thousands of concurrent connections. High image quality, low latency, and scalability are all achieved simultaneously for the first time.

In the past, in order for the screen to be connected to AI and run smoothly, it was often necessary to reserve computing power before leaving the factory. Existing screens were almost incapable of AI. Mofa's breakthrough allows existing screens to obtain real-time interaction capabilities without replacing hardware; it allows AI screens to move from "project-based" to "standardized products"; it allows the display industry to move from "selling hardware" to "selling services."

At this point, the large-scale implementation of AI screens has a real basis for reality.

This also means that billions of screen terminals around the world can obtain AI capabilities at extremely low marginal costs without replacing hardware.

Specifically, Mofa’s AI screen solution can be flexibly embedded in three links:

First, the control system side. Accessing computing power and memory from the lowest control card has the best effect. At present, Mofa has been adapted to the mainstream playback boxes (LED control boards) and commercial display motherboards in the display industry, and can drive AI interaction relying on the original computing power of the board/motherboard.

The second is the screen side. Integrate AI capabilities directly on the screen locally. Mofa Nebula has completed multi-terminal adaptation and supports mainstream operating systems such as Android, iOS, Windows, Linux, and Hongmeng. It can be used right after booting.

The third is the system integration end. Provide multi-modal perception capabilities through peripherals such as cameras and microphones. This path requires no additional computing power or additional chips to complete the access.

For existing screens, AI capability injection can be completed through the screen end or system integration end without the need to replace the hardware. For new screens, if higher-performance chips and computing power can be embedded in the control system before leaving the factory, and a complete AI screen solution can be pre-installed, the effect will be better and the implementation can be completed in one step.

At present, Mofa Technology’s AI screen solution has completed application verification in leading companies such as Unilumin, Leyard, TPV, Novavax, etc., and has entered the stage of large-scale embedding.

At this point, the screen is still the same screen, but the capabilities of the screen have been redefined.

Experts say Research summary: a watershed worthy of industry expectations

Experts say that together with Mofa Technology, when working with Unilumin Technology, TPV Technology and Hengshi Technology, they found that customer needs have quietly changed - they are no longer satisfied with a screen that is only for "viewing", but hope that the screen can "interact" and "serve". Specifically speaking, it means turning the screen into the front desk, sales, and customer service.

This change is not conceptual hype. AI screens have been put into practical applications in many scenarios such as airports, stores, exhibition halls, and conference rooms. More importantly, projects equipped with AI screen solutions have been transformed from "technical highlights" into "bid winning chips" - the market is expressing its position with orders.

The signal behind this is worth pondering: the market is paying for "screens that can serve".

When a screen is no longer just a cost item, but a "digital employee" that can create value, companies will be significantly less sensitive to its price and focus instead on its service capabilities and operational stability. This is the true direction that shows the return of industrial value - not an infinite competition of parameters, but the support of 7x24 uninterrupted "white-collar services."

From this we also extend an imagination:

For more than 20 years, Micro LED, which has attracted much attention in the industry, has always been known as the "ultimate display technology", but it has never found a battlefield where it cannot be used.

Today, the emergence of a new species called AI screen may point out a direction worth looking forward to for Micro LED.

The subsequent evolution path of AI screens may be: first complete the activation of existing screens, let "screens that can serve" become popular in all walks of life, and form a large-scale application ecosystem. When AI screens really take off, scenes with extreme requirements on brightness, lifespan, and resolution may emerge on a large scale. By then, the long-established technical advantages of Micro LED may truly have its moment of glory.

As a result, when the role of the screen changes from "a screen to be viewed" to "a service terminal housing staff", the valuation logic of the entire display industry will be completely rewritten. For LED large-screen displays, this means the possibility of breaking through the 50 billion output value ceiling; for LCD panels, this means opening up new incremental space beyond the 500 billion stock.

The AI screen itself is the real protagonist of this revolution. Micro LED is the most anticipated technology partner in the future after existing screens are activated into AI screens. As for who will stand next to it in the end, time will tell. But the start button has been pressed for this revolution.