- Phison’s SSD strategy cuts AI training costs from $ 3 million to $ 100,000
- Aidaptiv+ software changes AI work load from GPUs to SSDs effectively
- SSDs could replace expensive GPUs in Solid AI Model Education
The development of AI models has become increasingly expensive as their size and complexity grow, requiring massive calculation resources with GPUs playing a key role in handling the workload.
Phison, a key player in Portable SSDs have revealed a new solution aimed at reducing the cost of training a 1 trillion trillion parameter model by switching some of the treatment load from GPUs to SSDs, bringing the estimated operating cost of $ 3 million to just $ 100,000.
Phison’s strategy involves integrating its Aidaptiv+ software with high -performance SSDs to handle some AI tool treatment tasks traditionally managed by GPUs, while also containing NVIDIA’s GH200 Superchip to improve performance and keep costs manageable.
AI model growth and trillion parameter milestone
Phison expects the AI industry to reach the 1 trillion parameter milestone before 2026.
According to the company, model sizes have quickly expanded and moved from 69 billion parameters in Llama 2 (2023) to 405 billion with Llama 3.1 (2024), followed by Deepseek R3’s 671 billion parameters (2025).
If this pattern continues, a Billion parameter model can be revealed by the end of 2025, marking a significant leap in AI capacities.
In addition, it believes that its solution can reduce the number of GPUs needed to run large AI models, significantly by switching some of the treatment tasks away from GPUs to the largest SSDs, and this approach could bring training costs down to only 3% of the current projections (97% savings) or less than 1/25 of the usual operating expenses.
Phison has already done that Collaboration with Maingeear to launch AI work stations run by Intel Xeon W7-3455 CPUs, signaling its commitment to reshaping AI-Hardware.
Since companies seek cost-effective ways to train massive AI models, innovations in SSD technology could play a crucial role in the gain External HDD settings remain relevant to long-term data storage.
The pressure for cheaper AI training solutions gained momentum after Deepseek made headlines earlier this year when its Deepseek R1 model demonstrated that advanced AI could be developed into a fraction of the usual costs, with 95% fewer chips and reportedly only required $ 6 million for training.
Via Tweactown
