Tom's Hardware spotted Kioxia's huge LC9 Series 122.88 TB NVMe SSD at NVIDIA's GTC 2025. This monster drive comes in a 2.5-inch package with BiCS FLASH 8th generation 3D QLC tech featuring 2 TB die density. The drive connects through PCIe 5.0 and offers unmatched storage space for business data centers. When hooked up, this SSD pushes data at 128 GT/s using PCIe 5.0 in either 1x4 or 2x2 setups.
The design meets NVMe 2.0 standards and fits perfectly with AI tasks thanks to its CMOS directly Bonded to Array flash structure. Its specs show 0.3 drive writes daily with 67,000 TBW stamina, making it great for reading data but less ideal for heavy writing jobs. Tests at the event showed it reads data at about 15 GB/s, pretty fast even for high-end drives. Kioxia built special AiSAQ tech into the drive that moves vector database work from RAM to the SSD itself.
This clever approach cuts system costs yet keeps the speed needed for AI retrieval jobs. Visitors saw the SSD running live demos at booth #1811 in San Jose McEnery Convention Center through March 21. These demos focused on how quickly it fetches AI datasets. Data speed and storage size have become major roadblocks for big AI language models, making this drive a big step forward.
The massive 122.88 TB capacity means companies can run AI language models using all their business data with fast response times and long-lasting hardware. This helps AI systems answer questions based on company information without waiting forever or wearing out drives quickly. Business users need these improvements as they add more AI tools that handle growing amounts of information every day.
The design meets NVMe 2.0 standards and fits perfectly with AI tasks thanks to its CMOS directly Bonded to Array flash structure. Its specs show 0.3 drive writes daily with 67,000 TBW stamina, making it great for reading data but less ideal for heavy writing jobs. Tests at the event showed it reads data at about 15 GB/s, pretty fast even for high-end drives. Kioxia built special AiSAQ tech into the drive that moves vector database work from RAM to the SSD itself.
This clever approach cuts system costs yet keeps the speed needed for AI retrieval jobs. Visitors saw the SSD running live demos at booth #1811 in San Jose McEnery Convention Center through March 21. These demos focused on how quickly it fetches AI datasets. Data speed and storage size have become major roadblocks for big AI language models, making this drive a big step forward.
The massive 122.88 TB capacity means companies can run AI language models using all their business data with fast response times and long-lasting hardware. This helps AI systems answer questions based on company information without waiting forever or wearing out drives quickly. Business users need these improvements as they add more AI tools that handle growing amounts of information every day.
