Why did Yin Qi develop a smartphone that was never intended for sale?

July 13, 2026, Shanghai. At the product launch event of StepStar, Chairman Yin Qi stood on stage and pulled out a smartphone.

No specifications, no pricing, no release date — none of that was announced. Reporters in the audience exchanged bewildered glances.

This did not feel like a typical consumer electronics launch. There was no hands-on experience with a real unit, no mass production commitment, not even a clear statement of "when this phone will be available for purchase." A supplier later revealed that the AI agent phone showcased at the event "has no plans for mass production and will undergo a full redesign."

It looked more like a publicity stunt. Or rather, more like a polished presentation deck prepared for the Hong Kong Stock Exchange.

Yin Qi told the audience that he had consulted many industry friends in the terminal sector, and their feedback was unanimous: they all advised him to stay away from hardware. "We really wanted to listen to that advice, but we ultimately decided to go ahead anyway."

The statement sounded like a display of unwavering resolve, but upon reflection: would a company genuinely committed to building phones open its presentation by saying hardware is something to avoid?

Huawei never said that when it launched the Mate series, Xiaomi did not say that for its flagship digital series, and even ByteDance did not make such remarks when it unveiled its Doubao phone at the end of 2025. Despite facing the same setback of being blocked by WeChat, Doubao at least delivered functional engineering units ready for shipment.

What StepStar presented instead was a "concept."

This was not a real phone — it was a narrative about a "phone." To understand why this narrative had to be told, we must first look at the position StepStar finds itself in right now.

A "Phone" That Was Never Meant for Mass Production

The narrative is titled "The Trinity of Model × Software × Hardware".

StepStar unveiled its terminal brand STEPX, the agent-native operating system Step AOS, the personal agent StepAmoo, and the so-called "world's first large model-native agent phone" STEPX Neo. Of the four announcements, three were software and brand-related — only the last one was hardware.

And this hardware was precisely the physical carrier that tied the first three together.

The logic seemed seamless: the large model is the brain, the operating system is the runtime environment, and the hardware is the delivery vehicle. None of the three can be omitted, so building a phone was necessary.

But if you ask when this phone will actually be available for purchase, the answer remains shrouded in uncertainty.

Ni Jiale, President of StepStar Terminal, responded: "We focus more on the symbiosis between agents and users, hoping that more consumers will welcome AI agents into the physical world." Translated, that means: we haven't figured out the sales strategy yet.

Even more thought-provoking was Yin Qi's admission during the post-launch media group interview that in the first phase, the company would "not simply pursue shipment volume." The business model would not rely on hardware sales for revenue, and the company would ultimately return to the core logic of token consumption.

This means STEPX Neo was designed from the very beginning as a non-commercial product. Its purpose is not to end up in consumers' pockets, but to capture the attention of investors.

If you treat it as a consumer electronics product, the launch event was undoubtedly a failure: no price, no on-sale date, no review units available.

But if you regard it as a strategic presentation for the capital market, the event was a resounding success. Using a tangible device that could be held in hand, it firmly linked the two concepts of "AI large model company" and "next-generation interactive terminal."

And this year, "coincidentally," is the year StepStar plans to go public.

In April, the company completed its shareholding restructuring, converting from a limited liability company to a joint-stock company — the standard prerequisite for a Hong Kong IPO. In May, rumors emerged that it had secured nearly $2.5 billion in Pre-IPO financing and dismantled its red-chip structure.

Its pre-money valuation was around $5 billion to $6 billion, and the post-money valuation is heading straight for $10 billion. The goal is to complete a Hong Kong listing by the end of 2026, with a target cornerstone valuation of approximately $10 billion.

$10 billion. What does that figure mean?

According to *Caijing* reporting, StepStar's 2025 revenue was close to 500 million yuan, and its 2026 projected revenue is around 1.2 billion yuan. Based on a $10 billion valuation, its price-to-sales ratio exceeds 50x. By comparison, Zhipu AI and MiniMax — two other members of the "AI Six Dragons" that have already listed in Hong Kong — operate on a similar revenue scale, yet their valuations are nowhere near that multiple.

The capital market is willing to pay for this story, but the story needs to be compelling enough.

Over the past year, StepStar's narrative revolved around its foundational large models. However, the commercialization challenges facing large models are well-documented: low per-unit API revenue, long enterprise-market deployment cycles, and ToB operations that struggle to cover exorbitant R&D costs.

Yin Qi himself admitted that the commercialization path for China's large model industry is not just about coding. So what is it? His answer lies in the integration of software and hardware.

Since Yin Qi officially took over as Chairman of StepStar in January 2026, "AI + Terminal" has been established as the company's long-term, unwavering strategic direction.

Yin Qi brings not only his background from Megvii but also hands-on experience in terminal commercialization. He is all too familiar with the predicament the "AI Four Dragons" faced in the AI 1.0 era: leading technology but no viable commercial closure, with cash flow under sustained pressure.

Hardware is the answer Yin Qi found for StepStar to close that loop — or more precisely, the answer the capital market is willing to believe.

The narrative of "AI + Terminal Defining the Future of Interaction" is far more capable of supporting a $10 billion valuation than simply "selling APIs."

A "Collusion" Written Into the Prospectus

If this were just a one-man show by a single company, the story would lack credibility. StepStar's brilliance lies in its ability to pull key players across the industrial chain into the narrative.

Look at the list of Pre-IPO investors: Huaqin Technology, Longcheer Technology, OmniVision Group, and ZTE.

Huaqin and Longcheer are top-tier global mobile phone ODMs, responsible for end-to-end R&D, design, and manufacturing. OmniVision is a leading image sensor supplier, positioned upstream in critical mobile camera components. These are no ordinary financial investors — they are industrial chain endorsements for StepStar's "phone story."

Huaqin Technology is not just an investor; it is also the contract manufacturer for STEPX Neo. Their partnership is described as a "deeply integrated collaborative relationship, far beyond a simple OEM model."

Huaqin took a strategic stake during the B+ round financing and continued to increase its investment in the subsequent C round. This means that when StepStar writes in its prospectus that "we have built a full-stack layout spanning models to operating systems, with strategic investment and manufacturing support from the world's leading ODMs," every word is backed by tangible evidence.

ZTE also deserves attention.

Last October, ZTE launched the nubia Z80 Ultra, which was powered by StepStar's GUI Agent model. On the very same day StepStar unveiled the STEPX Neo, nubia announced that it would showcase its own AI agent phone at the World Artificial Intelligence Conference (WAIC).

Partners on one hand, potential competitors on the other. This delicate dynamic can be packaged as "industrial synergy" in the IPO narrative.

Hong Kong Investment Management Corporation (HKIC) — known as Hong Kong's "Temasek" — also appears on the shareholder list. Its investment represents not just capital backing, but also strong political endorsement for a Hong Kong IPO.

A complete chain has taken shape: a model company tells a terminal story → ODMs invest and manufacture → supply chain enterprises take stakes to endorse → Hong Kong institutional investors show support → Hong Kong IPO. Every link fits perfectly, and every participant knows exactly what role they are playing.

Of course, no story can withstand close scrutiny of its fine details.

For StepStar's "agent phone" to truly execute cross-app tasks, it must resolve the interface compatibility issues with third-party apps.

The Doubao phone incident serves as a stark warning.

At the end of 2025, ByteDance partnered with ZTE to launch the Doubao phone. Just two to three days after shipments began, mainstream apps including WeChat and Taobao collectively blocked it, as its GUI-based simulated clicks were classified as unauthorized plug-in behavior.

Having learned from that lesson, StepStar announced that all partner apps will connect through dedicated APIs rather than GUI automation, with initial ecosystem partners including Ctrip, Alipay, Didi, Meituan, Baidu, JD.com, WPS, and CapCut. But one name is conspicuously absent: WeChat.

Tencent is already a shareholder of StepStar. Yin Qi's response was that "we have had very in-depth discussions with Tencent, and hope to have the opportunity to cooperate in the future." Translated, that means: no deal has been reached yet.

How much real-world value can an "agent phone" offer if it cannot operate WeChat? The answer to that question does not need to appear in the prospectus. The prospectus only needs to state that "we have reached ecological cooperation with multiple leading application providers."

Meanwhile, a research report from TF International Securities analyst Ming-Chi Kuo shows that OpenAI is also accelerating the development of its first AI agent phone, moving the mass production timeline forward from the originally planned 2028 to the first half of 2027 — one of the key reasons being to "align with a potential IPO by the end of the year."

When the world's most valuable AI company is using the same logic to tell a hardware story, StepStar's narrative is not just reasonable — it becomes almost necessary.

Epilogue

So, will STEPX Neo ever enter mass production?

Maybe yes, maybe no. StepStar's current plan is: launch a "wish pool" on July 20 to let users vote on desired phone features; launch a "Fantasy Project" in partnership with Bilibili on August 20, making devices available to thousands of makers first.

This sounds more like a crowdsourced innovation experiment than a formal product launch. Yin Qi's subtext that the first phase will return to the core logic of token consumption implies that the hardware is not necessarily meant for sale — it is a tool to verify the business model, collect data, and improve the ecosystem.

Even if STEPX Neo never enters large-scale mass production, it will have fulfilled its purpose. In the prospectus, in roadshow presentations, and in media coverage, it already exists as proof of the company's strategic layout.

The capital raised needs a clear destination. For investors, backing a company that is "redefining human-agent symbiosis in the agent era" is far more appealing than investing in a company that "sells APIs."

But we must also clearly recognize that using hardware purely as a narrative tool is not without risks.

Once the IPO is completed, the market will expect tangible shipment figures, user activity metrics, and verifiable progress on ecosystem partnerships. If STEPX Neo remains stuck at the concept stage by then, the capital market's patience will quickly evaporate.

The lesson of the "AI Four Dragons" — whose share prices remained under long-term pressure after listing — is that technological leadership cannot compensate for underwhelming commercialization results. Yin Qi knows this better than anyone, having come directly out of Megvii.

Furthermore, StepStar's "terminal strategy" carries a hidden cost.

By transforming itself from a "large model supplier" into a "terminal brand," it has created a potential competitive conflict with existing partners such as OPPO and Honor. Those terminal manufacturers that might otherwise have licensed StepStar's model capabilities will now reassess the risks of collaboration.

Yin Qi emphasized during the group interview that "we will not build a closed all-in-one ecosystem, and hope to be friends with all terminal manufacturers." But in the business world, once trust is broken, the cost of repairing it becomes extremely high.

Returning to the original question: is hardware an end in itself, or just a means?

In the traditional consumer electronics logic, hardware is the end goal: build it, sell it, make a profit.

But in StepStar's logic, hardware is clearly just a means. A means to demonstrate technical strength, to build an ecological moat, and to tell a compelling story.

And the real audience for this story is not the tech enthusiasts leaving comments in the launch live stream — it is the Hong Kong Stock Exchange listing reviewers, cornerstone investors, and public market fund managers.

STEPX Neo may never become the phone in your pocket, but it has already become the most impressive slide in StepStar's prospectus.

This article is from the WeChat public account "Decoding NewSight" by Decoding NewSight, republished with authorization from 36Kr.