ZTE unveils agentic AI smartphone

ZTE Unveils Agentic AI Smartphone, Redefining Mobile Experience

In a significant development, ZTE Corp. Has unveiled a range of smartphones with integrated artificial intelligence services, joining other Chinese manufacturers in making agentic AI a central part of mobile operating systems. The state-backed telecommunications company demonstrated the NaviX Ultra at the World Artificial Intelligence Conference in Shanghai, describing it as the world's first agentic AI smartphone. This move marks a shift in the smartphone industry, as manufacturers turn to AI to boost sales and create a new user experience.

The NaviX Ultra is available in various colors and launches ByteDance's Doubao assistant through a voice command or dedicated button. Unlike conventional phones that add isolated AI features, agentic devices are designed to execute instructions autonomously across multiple applications. On-device models handle frequent tasks requiring low latency, with cloud-based systems processing more complex requests. This approach enables the AI to navigate apps, perform multi-step tasks, make payments, edit media, and more, with minimal user input.

Chinese manufacturers, including StepFun, Honor, and Xiaomi, are turning to AI as higher memory costs, inflation, and weak consumer demand weigh on the global smartphone industry. The market is projected to record its steepest-ever annual decline, with budget-focused Chinese brands facing particular pressure from thinner margins. In response, these companies are redesigning the phone around on-device AI agents, which demands a rethink of the phone itself rather than another spec bump.

The shift to agentic AI matters far beyond China, as the country ships roughly a fifth of the world's smartphones through domestic brands, and its supply chain touches almost every device sold globally. When Chinese brands move together on a hardware idea, component makers across the region follow. This is why the current debate about what an AI-first phone should look like is, in practice, a regional question for Asia's electronics economy.

The trigger for this shift is agentic AI, software that does not just answer questions but completes tasks across apps. An assistant that books a train, edits a photo, and drafts a reply needs constant access to the screen, the microphone, and the user's data. Running those models on the device, rather than in the cloud, keeps latency low and personal data local, which matters in markets where privacy scrutiny is rising.

On-device inference is expensive in silicon terms, requiring more memory, a dedicated neural processing unit, and a battery that can sustain sustained compute without overheating the handset. Chinese makers have responded by widening the physical envelope of the device, testing thicker vapor chambers, larger silicon-carbon batteries, and, in some concept units, secondary displays and always-listening sensors. The result is a category of phone designed first for an assistant and second for the human holding it.

Honor has said it will spend heavily on AI over several years, positioning its devices around an on-board agent. Xiaomi has folded its in-house models into its wider hardware line, while Huawei continues to build around its own operating system and chipset after years of United States export limits. Each is betting that the assistant, not the camera, becomes the reason to upgrade.

The design language is changing to match, with prototypes making room for cooling and battery volume, and some concepts adding a dedicated button or sensor bar for the assistant. It is an admission that the human interface is no longer the only thing being served. The device is also being built for a piece of software that never stops running.

Every one of these design choices lands on suppliers across Asia, with more memory meaning more orders for Korean makers, dedicated neural silicon leaning on Taiwan's foundries, and batteries, cooling, and sensors running through a dense network of component firms in southern China, Vietnam, and Malaysia.

This concentration is a strength and a risk, as it lets Chinese brands iterate quickly, but it also ties the region's factories to a single unproven bet. If buyers do not pay a premium for an AI-first handset, the inventory correction ripples straight back through the same suppliers.

Analysts at Counterpoint Research and Canalys have repeatedly flagged that AI features have yet to lift replacement rates in a measurable way. Surveys of buyers in China, India, and Southeast Asia show interest in specific assistant features, such as translation, photo cleanup, and call summaries, but little willingness to pay more for them.

The danger for Chinese makers is that they add cost and complexity that the mid-range buyer, who drives volume across Asia, does not reward. There is also a software gap, as an agent is only as useful as the apps it can reach, and the most-used services in China sit inside a handful of platforms that guard their data closely.

Until those platforms open up, an on-device agent may be limited to first-party functions, which is a thinner promise than the marketing suggests. What is an AI-first smartphone? It is a handset designed around an on-device assistant that can act across apps, rather than a conventional phone with AI features bolted on.

The hardware priorities, memory, neural silicon, and battery, are set by the model rather than the camera or the screen. Why are Chinese brands moving now? On-device models have become small and efficient enough to run useful tasks locally, and a saturated market has left makers searching for a reason for buyers to upgrade.

Moving together also lets them shape the regional supply chain to their needs. The risk for Asia's suppliers is that the bet concentrates orders in memory, foundry, and component firms across Korea, Taiwan, China, and Southeast Asia. If AI-first phones do not command a premium, that same concentration turns a boom in orders into a sharp correction.

In the global AI phone race, China's phone makers, forced into independence by circumstance, currently hold the more coherent hand. A market once defined by who could source the best camera sensor is now defined by who owns the model running behind the lock screen.

Huawei's comeback is a significant development, with the company climbing to roughly twenty-two percent of Chinese shipments in the second quarter, its strongest position in years. What gets less attention is why the comeback happened. Cut off from global chip supply chains by sanctions, Huawei was forced to build everything itself, its own Kirin processors, its own HarmonyOS, and its own Pangu large language model running Agentic AI features natively on the device.

That isolation, originally seen as a weakness, became a structural advantage once global memory prices spiked and rivals dependent on imported components had to raise retail prices by hundreds of yuan just to protect margins. Huawei did not choose vertical integration as a marketing strategy; it was forced into building the whole stack, and the whole stack is exactly what the AI phone era now rewards.

The same domestic hardware logic extends into China's foundation model sector, with Zhipu's GLM 5 training and serving entirely on Huawei's own Ascend silicon rather than foreign chips. This gives Chinese state buyers and export-sensitive enterprises a model that is immune to the kind of supply restrictions that reshaped Huawei's phone business in the first place.

The global AI smartphone market remains tiny in absolute terms, valued near forty million dollars in 2026, but is forecast to grow at a compound rate above fifty percent through 2035 as agentic features move from novelty to expectation. Whoever controls the default assistant layer on the world's billions of phones stands to capture a disproportionate share of that growth, and right now, that layer is being contested on two very different models: China's vertically owned stack against a Western model of licensed intelligence stitched onto existing hardware.

Geopolitics sits underneath all of this, with export controls on advanced chips meant to slow Chinese AI progress, yet they appear to have accelerated domestic chip and model co-design instead. The lesson China's smartphone makers have absorbed is that sanctions create urgency, and urgency, applied consistently across chips, operating systems, and models, produces exactly the kind of integrated stack Western firms are now scrambling to license piece by piece.

None of this guarantees Chinese brands will win the global AI phone race outright. Apple's installed base, brand loyalty, and privacy-focused architecture remain formidable, and its Private Cloud Compute approach still gives it a genuine differentiator once the Gemini partnership matures. But the balance of technical self-sufficiency has shifted in ways that were unthinkable five years ago.

A market once defined by who could source the best camera sensor is now defined by who owns the model running behind the lock screen, and on that measure, China's phone makers currently hold the more coherent hand.