Neosem：更高容量与QLC转型推动SSD测试设备需求激增
Neosem公司表示，随着AI基础设施投资扩大，SSD测试设备市场进入快速增长期。SSD容量提升及向QLC技术转型，推动了对测试设备的需求激增。该公司预计，更高容量和QLC转变将带来测试设备需求的显著增长。
原文深度解读
本文报道了Neosem公司董事Kim Shin-young在2026年AI半导体测试与插座技术会议上的发言，指出随着AI基础设施投资扩大，SSD测试设备市场进入快速增长期。关键信息包括：1TB SSD测试时间从30分钟增至16TB的约2.5小时；SSD容量从8TB增至64TB以上，QLC技术普及；PCIe Gen 6接口采用PAM4信号导致测试需求增加；Neosem在PCIe Gen 5市场近乎垄断，并在Gen 6市场领先；四大超大规模企业资本支出达7300亿美元，全球AI投资接近1万亿美元。文章与电子元器件供应链相关，因为测试设备是存储芯片制造和验证的关键环节。
- Neosem董事Kim Shin-young在2026年8月26日于水原会议中心举行的AI半导体测试与插座技术会议上表示，1TB SSD测试约需30分钟，而16TB产品约需2.5小时。
- Kim指出，保持相同产量需要比过去多2到3倍的测试系统。
- SSD容量从8TB增至16TB、32TB和64TB，内存制造商路线图包含超过120TB的高容量产品。
- NAND架构从TLC（每单元3位）向QLC（每单元4位）转变，QLC物理数据稳定性较低，增加了老化测试和系统级测试（SLT）的需求。
- PCIe Gen 6将传输速率从Gen 5的32 GT/s翻倍至64 GT/s，采用PAM4信号而非NRZ，但增加了信号失真和误码率，需要额外验证信号完整性、数据完整性和错误恢复。
- Neosem在全球SSD测试市场占据强势地位，主要竞争对手Advantest去年退出该业务，使Neosem在PCIe Gen 5设备市场近乎垄断，并在Gen 6市场领先。
供应链影响
- 随着SSD容量提升和QLC转型，测试时间延长，内存制造商可能增加测试设备采购，从而扩大Neosem等测试设备供应商的订单机会。
- PCIe Gen 6的采用可能推动新一代测试设备需求，Neosem的早期领先可能巩固其市场地位，但竞争格局可能取决于其他厂商的进入。
- AI基础设施投资增长可能带动存储设备需求，进而影响测试设备供应链，但具体影响取决于投资落地速度和规模。
- QLC SSD的稳定性问题可能增加老化测试和SLT需求，可能影响测试设备的技术规格和产能配置。
- Advantest退出SSD测试业务可能改变竞争格局，Neosem的垄断地位可能使其在定价和供应上具有优势，但需关注潜在反垄断或新进入者风险。
系统解读，仅作为判断线索，不构成备货、出货、涨价或投资建议。
随着人工智能（AI）基础设施投资的扩大，固态硬盘（SSD）测试设备市场正进入快速增长期。SSD容量的提升和技术进步延长了测试时间，并推动设备需求激增。
在8月26日于水原会议中心举行的AI半导体测试与插座技术会议上，Neosem公司董事Kim Shin-young表示，测试1TB SSD此前约需30分钟，而16TB产品约需两个半小时。
“保持相同的产量需要比过去多两到三倍的测试系统，”Kim说。
SSD测试时间的增加源于更高容量、NAND架构的变化以及高速接口标准的采用。
SSD市场正同时向更高容量和四层单元（QLC）技术发展，QLC每个单元存储4位。SSD容量已从8TB增至16TB、32TB和64TB。内存制造商的产品路线图现在包括超过120TB的高容量产品。随着AI计算变得更加先进，需要存储和备份的数据量呈指数级增长，增加了必须验证的物理数据量。
NAND架构也正从三层单元（TLC，每单元存储3位）迅速转向QLC。QLC每单元提供4位，而TLC为3位。QLC在实现更高容量方面优于TLC，但物理数据稳定性较低。这增加了老化测试和系统级测试（SLT）的需求，这些测试在极端条件下验证产品运行。此前，SSD测试主要侧重于读写等基本功能。
基于第六代PCI Express（PCIe）互连接口的SSD也已开始量产。PCIe Gen 6将传输速率从Gen 5的32 GT/s翻倍至64 GT/s。这是通过采用四电平脉冲幅度调制（PAM4）信号而非非归零（NRZ）实现的，每个信号传输的信息量从两个电平（1位）增加到四个电平（2位）。
然而，PAM4也显著增加了信号失真和误码率（BER）。这需要额外验证信号完整性、数据完整性和错误恢复。信号完整性确定电信号是否无噪声或失真地传输，数据完整性验证接收的数据是否无失真地解释和传递，错误恢复测试验证系统在发生错误时能否自行识别和解决问题。
因此，高容量SSD需要在多个阶段进行更精确的验证，显著增加了测试时间。内存制造商可能会部署更多测试设备以抵消更长的测试时间，从而扩大Neosem SSD测试系统的订单机会。
Neosem在全球SSD测试市场占据强势地位。其主要竞争对手之一Advantest去年退出了SSD测试设备业务，使Neosem在全球PCIe Gen 5设备市场近乎垄断。该公司在PCIe Gen 6 SSD测试设备市场也处于领先地位。
“AI时代的瓶颈是数据，而不是计算，”Kim说。“数据的增长推动存储设备行业的进步，为测试设备行业创造了重大机遇。”
四大超大规模企业——亚马逊、微软、Alphabet和Meta——今年的资本支出合计高达7300亿美元，同比增长约90%。全球AI总投资预计将接近1万亿美元。
The solid-state drive (SSD) testing equipment market is entering a period of rapid growth as investment in artificial intelligence (AI) infrastructure expands. Increasing SSD capacities and technological advances are extending testing times and driving a surge in equipment demand.
At the AI Semiconductor Test & Socket Tech Conference 2026 held at the Suwon Convention Center on Aug. 26, Kim Shin-young, a director at Neosem, said that testing a 1TB SSD previously took about 30 minutes, while a 16TB product requires around two and a half hours.
"Maintaining the same production volume requires more than two to three times as many testing systems as in the past," Kim said.
The increase in SSD testing time stems from higher capacities, changes in NAND architecture and the adoption of high-speed interface standards.
The SSD market is simultaneously moving toward higher capacities and quad-level cell (QLC) technology, which stores four bits per cell. SSD capacities have increased from 8TB to 16TB to 32TB and 64TB. Memory manufacturers' product roadmaps now include high-capacity products exceeding 120TB. As AI computing becomes more advanced, the volume of data that needs to be stored and backed up is growing exponentially, increasing the physical amount of data that must be verified.
NAND architecture is also rapidly shifting from triple-level cell (TLC), which stores three bits per cell, to QLC. QLC offers four bits per cell compared to TLC's three bits. QLC offers an advantage over TLC in achieving higher capacities but has lower physical data stability. This is increasing demand for burn-in and system-level testing (SLT), which verify product operation under extreme conditions. Previously, SSD testing focused primarily on basic functions such as reading and writing.
Mass production of SSDs based on the sixth-generation PCI Express (PCIe) interconnect interface has also begun. PCIe Gen 6 doubles the transfer rate from 32 gigatransfers per second (GT/s) in Gen 5 to 64GT/s. This is enabled by the adoption of four-level pulse amplitude modulation (PAM4) signaling instead of non-return-to-zero (NRZ), increasing the amount of information transmitted per signal from two levels, or one bit, to four levels, or two bits.
PAM4, however, also significantly increases signal distortion and bit error rates (BER). This requires additional verification of signal integrity, data integrity and error recovery. Signal integrity determines whether electrical signals are transmitted cleanly without noise or distortion, while data integrity verifies whether received data is interpreted and delivered without distortion. Error recovery testing verifies whether a system can identify and resolve problems on its own when errors occur.
As a result, high-capacity SSDs require more precise verification across multiple stages, significantly increasing testing times. Memory manufacturers are likely to deploy more testing equipment to offset the longer test times, expanding order opportunities for Neosem's SSD testing systems.
Neosem has a strong position in the global SSD testing market. Advantest, previously one of its major competitors, exited the SSD testing equipment business last year, leaving Neosem with a virtual monopoly in the global PCIe Gen 5 equipment market. The company is also taking an early lead in the PCIe Gen 6 SSD testing equipment market.
"The bottleneck in the AI era is data, not computing," Kim said. "The growth in data is driving advances in the storage device industry, creating a major opportunity for the testing equipment industry."
The combined capital expenditure guidance of the four major hyperscalers — Amazon, Microsoft, Alphabet and Meta — is as high as $730 billion this year, up about 90% from the previous year. Total global AI investment is expected to approach $1 trillion.