By Sebastien Jean, CTO, Phison US

This year enticed an influx of GenAI and enterprise use cases, introducing fresh, unique and data-intensive workloads for organizations' infrastructure to grapple with, particularly for data storage. In 2025, there will be a spike in technological developments tailored to meet this shift, outlined below:

1. Increased Focus on AI and Machine Learning:

• AI Training Will Go Private: As the AI buzz brought forth AI PCs, the security loopholes increased, and consumers have started to become weary of allowing data to leave their control when utilizing cloud services. This is coupled with the fact that cloud costs remain difficult to predict and often bring surprises on the monthly bill when compared against on-premises solutions. We will see a shift to close-looped AI systems where models are trained locally and deployed directly to employee laptops to enable edge inference.
• AI-Specific Storage Features: Storage controllers might include specialized hardware or software to accelerate AI workloads, particularly in edge devices where on-device processing is crucial. Imagine a USB drive that can perform basic image recognition or natural language processing tasks on the data it already contains thanks to its integrated AI capabilities.
  

2. Enhanced Performance and Efficiency:

• PCIe 5.0 and Beyond: Expect wider adoption of PCIe 5.0 for SSDs, enabling even faster data transfer rates. On the enterprise level, expect PCIe 6.0 to double the speeds to meet the demand of AI workloads. These storage servers will likely integrate cold plate conduction cooling or even immersive cooling which places the entire server in a non-conductive liquid.
• NVMe Optimization: NVMe (Non-Volatile Memory Express) will continue to be the dominant protocol for SSDs, and they will be further optimized to reduce latency and improve overall performance.
• Lower Power Consumption: As always, there will be a push for greater energy efficiency, especially in mobile devices and battery-powered applications. Controllers will play a key role in minimizing power usage without sacrificing performance.
  

3. Tailored Use of Densities:

• Increased Integration: We may see more highly integrated solutions, with controllers and NAND flash memory packaged together in smaller and more efficient form factors. This could be particularly important for compact devices like wearables (i.e. smart glasses) where electronics must conform to the shape of clothing and accessories.
  

4. Security Enhancements:

• Hardware-Level Encryption: With the rapid progress of quantum computing and the new post-quantum cryptography (PQC) guidelines from NIST, expect to see more SSDs and other flash-based devices with built-in hardware encryption capabilities, such as PQC digital signing, to protect sensitive information. Given their parallelized nature, adoption should be swift in the coming years.
• Secure Boot and Firmware Updates: SSDs will likely incorporate more robust security features such as DICE/SPDM to allow an SSD to cryptographically prove it is authorized and thus better prevent unauthorized access and malicious firmware updates, ensuring the integrity and trustworthiness of the device. This may look like self-encrypting drives (SEDs) and include key features like isolated key management, side channel attack detection and mitigation. SSDs will integrate detectors for tunneling electron beams and manage internal power usage to make differential power analysis much harder.
  

5. Customization and Differentiation:

• Application Specific Integrated Circuit (ASIC): We may see more specialized application accelerators added into SSDs for use-case-specific tasks like wear leveling, error correction, and performance optimization. This could lead to longer lifespan, improved reliability, and faster speeds for SSDs and other flash-based devices.
• Tailored Acceleration: Computational Storage will evolve to add accelerators on to SSDs for highly specific tasks that could be handled by the CPU but are offloaded to one or more SSD(s) to handle as a background process without adding substantial cost or power consumption. The accelerated SSDs can handle large volumes of data much faster, all while consuming less power than traditional processors, something that will be in high demand as AI use cases expand in 2025.
  

##

ABOUT THE AUTHOR

Sebastien Jean is the Chief Technology Officer at Phison US, where he focuses on developing technology strategy and building alliances with other innovative companies. He also works closely with engineering teams to help integrate new concepts into products. With 24 years of experience and over 30 filed patents, he has established himself as a thought leader in the storage industry. Before joining Phison, he held senior technology positions at Micron, SanDisk, and Western Digital. At Phison, he helped devise an iterative technology roadmap that advances Security, AI, PCIe Gen 5, USB4, BGA NVMe, Client Gaming Drives and Enterprise High-density E1.S. He earned a BS in Computer Science at the University of Ottawa (Canada).