· 个股论点

解析$AEHR作为AI芯片热测试瓶颈解决者,正从研发转向量产,受益于巨头资本开支周期。

涉及标的:

中文翻译

更简明的 $AEHR 概览 - 11 亿美元(无行话)。这是我的一个多头持仓: 在 AI 领域,散热是一个已知的瓶颈。 随着新一代 $NVDA 或 $AMD 类型的芯片发热量增加,如果一颗芯片熔毁,将是灾难性的。 因此科技巨头 -> 对每一颗芯片进行热应力测试,以查看其是否会在现场过早失效。 $AEHR 出售用于测试这些芯片的机器。 去年,他们类似于 $POET(处于预认证/测试阶段) -> 但现在他们已经通过了认证。 我们正看到从研发到大规模生产的拐点: 今天: -> $AEHR 获得 1400 万美元的芯片老化系统(wafer level)订单,以帮助扩大芯片生产。 但这台 1400 万美元的机器每周只能处理有限数量的晶圆。 -> 随着该 AI 客户扩大规模,他们将物理上耗尽热测试吞吐量。 因此,如果他们喜欢 $AEHR,就会购买更多,从而推动收入增长。 他们还有一个独立的产品线(Sonoma),用于测试封装后的芯片。 超大规模云服务商一直在订购此产品,因此有两个不同的测试阶段,两条收入来源。 管理层指引 2026 年下半年新增订单为 6000 万至 8000 万美元(这是巨大的增长)。 你可以参考 $TER(510 亿美元)或爱德万测试(约 1250 亿美元)来看看这类公司能有多大。 他们还在其他数据中心层进行测试: - 他们正在与 NAND Flash 内存供应商合作 - 硅光子学 我尽量去除了大部分技术行话,以便大多数人更容易理解,但这确实是一个微妙/技术性的领域。 $AEHR 具有潜力,因为它在多年的研发和认证后,现在开始实现真实规模的量产。 我们正处于一个巨大的资本支出周期中(想想 $ASML 的周期,人们为建设订购大量机器)。

英文原文

A simpler overview of $AEHR - $1.1B (without the jargon). One of my long positions: With AI, thermal is a known bottleneck. With new gen $NVDA or $AMD type chips that gets hot, if one chip melts down, it's a disaster. So tech giants -> every single chip stress tested with heat to see if it fails prematurely in the field. $AEHR sells the machines that does the testing of these chips. Last year, they were like $POET (pre-qualification/test phase) -> but now they've passed. We're seeing that inflection from R&D to mass production: Today: -> $14M order for $AEHR burn-in systems to help scale chip production (wafer level) But that $14M machine order can only handle a finite number of wafers per week. -> As this AI customer scales up, they will physically run out of thermal testing throughput. so they buy more if they like $AEHR, and this ramps up revenue. They also have a separate product line (Sonoma) for testing chips after they're packaged. Hyperscalers have been ordering this, so two different testing stages, two revenue streams. Management guided for a massive second half 2026 $60M to $80M in new bookings (which is big growth). You can look at $TER ($51B) or Advantest (~$125B) to see the ceiling of how big a lot of these companies can get. They also do testing across other data center layers too: - So they're working with NAND Flash memory suppliers - Silicon photonics Tried to strip out as much technical jargon as possible so it's more understandable to majority, but it's a very nuanced/technical field. $AEHR has potential as it's now scaling up real-volume after years of R&D and qualification. And we're sitting in an enormous capex cycle (think $ASML cycles, where people order a lot of machines for the buildout).

在 X 上查看原推 ↗