AI News AI资讯 2d ago Updated 1d ago 更新于 1天前 52

Exclusive | This company raises nearly 200 million yuan to accelerate the full-color process of AI glasses with lossless Micro-LED 硬氪首发 | 获近2亿元融资,这家公司用无损Micro-LED加速AI眼镜全彩化进程

The mass production story of Micro-LED has been told for many years, with several changes in the leading players, but the script always stalls at the same point: the dismal yield and light efficiency at the chip level. Even now, when a company less than three years old claims to simultaneously achieve yields above 6N and operating temperatures exceeding 140°C using a "damage-free architecture," your first reaction should be skepticism. However, Qiushui Semiconductor's nearly 200 million yuan in Micro-LED的量产故事讲了好多年,主角换了好几拨,剧本却总卡在同一个地方:芯片端一塌糊涂的良率和光效。直到现在,当一家刚成立不到三年的公司,宣称能用一套“无损架构”同时把良率拉到6N以上、工作温度干到140℃以上,你第一反应应该是怀疑。但秋水半导体拿到近2亿融资,以及投资方近乎亢奋的表述,却暗示这次可能不是又一个PPT故事。

75
Hot 热度
75
Quality 质量
70
Impact 影响力

Analysis 深度分析

The mass production story of Micro-LED has been told for many years, with several changes in the leading players, but the script always stalls at the same point: the dismal yield and light efficiency at the chip level. Even now, when a company less than three years old claims to simultaneously achieve yields above 6N and operating temperatures exceeding 140°C using a "damage-free architecture," your first reaction should be skepticism. However, Qiushui Semiconductor's nearly 200 million yuan in financing and the near-euphoric statements from investors suggest that this time it may not be just another PPT story.

The root of the problem lies in the old "clumsy methods." The entire industry has been akin to micro-sculpting: starting with a 4-inch sapphire substrate, metal-bonding LED wafers with driver circuits, and finally using brute-force etching to partition pixels. The result? The light-emitting material ends up scarred, chip efficiency plummets, yields are catastrophically low, and the light emission angle scatters like water from a bucket. This has directly caused the two most anticipated applications—AR glasses and smart vehicle lighting—to remain stuck below the one-million-unit shipment threshold. Without capable chips, all terminal imaginations are empty talk.

Qiushui Semiconductor's approach sounds somewhat "rebellious"—it bypasses etching altogether. Its claimed damage-free chip architecture is based on the core logic of creating electrical isolation between pixels through innovative physical design, rather than physically removing material. As investor Chaohui Capital puts it, this is like building an "invisible wall of electrical insulation" without destroying the physical structure. This indeed represents a conceptual shift. Combined with an 8-inch silicon substrate and hybrid bonding, it claims to narrow the light emission angle from an absurd ±60° to a useful ±10°, while significantly improving temperature tolerance. If the data holds true, this is like turning a floodlight scattering in all directions into a flashlight beam that can be precisely controlled. Its claimed 3.75-micron pixel interconnection process even asserts alignment with Huawei's "Tao's Law" at the same node, showcasing considerable technical ambition.

However, the chasm between "can be done" and "can be mass-produced" is immeasurably deep. The Micro-LED field has never lacked astonishing laboratory data; what it lacks is stable, repeatable, and economically viable mass production capability. Qiushui’s choice of a fab-lite model (building its own hybrid bonding line while outsourcing other steps) is a pragmatic compromise. Compared to building an IDM line from scratch, it is indeed faster. The company claims to be "the first domestic startup to complete the line qualification for 8-inch hybrid bonding," with plans to begin production in October and achieve annual output of tens of millions of chips. Speed is an advantage, but it also places the core process lifeline entirely in its own hands—this is both a technical moat and a concentration of risk. If issues arise in the commissioning of this self-built line, the entire mass production schedule could stall.

Founder Jiang Zhenyu's resume and the team's background (Huawei HiSilicon, Yangtze Memory Technologies, Intel) lend technical credibility to the company. However, his forecast for the future market—"shipments of AI glasses surging from several hundred thousand units to tens of millions within the next year"—sounds more like a motivational slogan than a solid industry projection. More crucially, he correctly identifies the ultimate test for Micro-LED: colorization. He admits that traditional red Micro-LED suffers over 97% light efficiency loss due to etching, and "damage-free technology is the key to solving this." The company plans to achieve a breakthrough in colorization by the end of the year. This is a monumental promise. If successful, it will be a true game-changer; if delayed, the foundation of its entire business narrative could crumble.

Investor Tongshang Fund positions Qiushui within the high-growth sectors of "automotive intelligence and AI-AR hardware adoption," praising its "significant technological moat and business positioning advantages." This reflects capital's thirst for a "hard tech" solution. But a hot track does not guarantee a ticket to success. On the Micro-LED battlefield, giants like Samsung and Sony have never exited, and numerous seasoned players in China are also eyeing the opportunity. Qiushui’s damage-free architecture is a differentiating card, but the game has only just begun.

In essence, Qiushui Semiconductor's story is a precise bet: wagering that a disruptive process route can outrun time, outpace yield challenges, and outperform all competitors, ultimately crossing the finish line first at the ultimate hurdle of colorization. The money it has raised serves as both ammunition and a countdown. Over the next year, the wafers on that Ningbo production line and the year-end technical report on colorization will be more persuasive than any financing press release. The market doesn’t need the "next Apple"; it needs pragmatists who can truly turn technology into shipment volume.

Micro-LED的量产故事讲了好多年,主角换了好几拨,剧本却总卡在同一个地方:芯片端一塌糊涂的良率和光效。直到现在,当一家刚成立不到三年的公司,宣称能用一套“无损架构”同时把良率拉到6N以上、工作温度干到140℃以上,你第一反应应该是怀疑。但秋水半导体拿到近2亿融资,以及投资方近乎亢奋的表述,却暗示这次可能不是又一个PPT故事。

问题的根源在于过去那套“笨办法”。整个行业像在搞微雕,先用4英寸蓝宝石衬底,再把LED晶圆和驱动电路金属键合,最后通过粗暴的刻蚀来分割像素。结果呢?发光材料被刻得伤痕累累,芯片效率暴跌,良率惨不忍睹,出光角度散得像泼水。这直接导致了AR眼镜和智能车灯这两个最被期待的场景,出货量始终被卡在百万台门槛下徘徊。芯片不行,终端的一切想象都是空谈。

秋水半导体的打法,听起来确实有点“叛逆”——它直接绕开了刻蚀。其宣称的无损芯片架构,核心逻辑是通过创新的物理设计在像素间建立电性隔离,而非物理去除材料。用投资方朝晖资本的话说,这就像在不破坏物理结构的情况下,建起一道“电性绝缘的隐形墙”。这确实是一个思路上的转折。配合8英寸硅衬底和混合键合,它声称能将出光角度从离谱的±60°收窄到有用的±10°,温度耐受性也大幅提升。如果数据属实,这相当于把一盏四处乱照的泛光灯,变成了一束可以精准控制的手电筒光。其宣称的3.75微米像素互联工艺,甚至放言与华为的“韬定律”处于同一节点,技术野心不小。

但“能做”与“能量产”之间,隔着一条深不见底的鸿沟。Micro-LED领域从不缺惊人的实验室数据,缺的是稳定、可重复、有经济性的量产能力。秋水选择的Fab-lite模式(自建混合键合产线,其他环节外包),是一种务实的折中。相比从零开始建IDM整线,这确实更快。公司称自己是“国内第一家完成8英寸混合键合通线的初创企业”,预计10月通线,年产千万颗芯片。速度是优势,但同时也把核心工艺命脉完全抓在自己手里——这既是技术壁垒,也是风险集中点。一旦这条自建产线调试出现问题,整个量产节奏就会停摆。

创始人蒋振宇的履历和团队背景(华为海思、长江存储、英特尔)为公司增添了技术信用。但他对未来市场的判断——“未来一年内AI眼镜出货量从几十万台跃升至千万台”,更像是一种激励性的口号,而非扎实的行业预测。更关键的是,他正确地指出了Micro-LED的终极考验:彩色化。他坦言,传统红光Micro-LED因刻蚀导致光效损失超97%,而“无损技术是解决这一问题的关键”。公司计划年底实现彩色化突破。这是一个巨大的承诺。如果成功,那将是真正的破局;如果延迟,其整个商业叙事的基础都会动摇。

投资方通商基金将秋水置于“汽车智能化、AI-AR硬件普及”的高景气赛道中,赞誉其“技术壁垒与业务卡位优势显著”。这反映了资本对一个“硬科技”解决方案的渴求。但赛道热,不代表入场券就能兑现。Micro-LED的战场上,巨头如三星、索尼从未退场,国内亦有众多老手在虎视眈眈。秋水的无损架构是一张差异化的牌,但牌局才刚刚开始。

所以,秋水半导体的故事,本质上是一场精密的赌注:赌一种颠覆性的工艺路径,能跑赢时间、跑赢良率、跑赢所有竞争对手,最终在彩色化这个终极关卡前,率先撞线。它拿到的钱,既是弹药,也是倒计时。接下来一年,宁波那条产线上的晶圆,以及年底那份关于彩色化的技术报告,将比所有融资新闻稿都更有说服力。市场不需要下一个“下一个苹果”,它需要的是能真正把技术变成出货量的实干家。

Disclaimer: The above content is generated by AI and is for reference only. 免责声明:以上内容由 AI 生成,仅供参考。

芯片 芯片 融资 融资 产品发布 产品发布
Share: 分享到: