[ad_1]
AI has the potential to generate $280 billion in economic benefits for Australia by 2030, and it’s clear that its transformative capabilities are key to the Future Made in Australia agenda, driving efficiency, innovation and sustainability, critical for industries like manufacturing. However, AI requires an incredible amount of data, and as the use of AI increases, the raw data requirements are accelerating well beyond what legacy environments can handle. Both enterprises and hyperscalers must find sustainable ways to scale within space, power and budget constraints. Otherwise, the new era of AI may stumble.
Even before the emergence of generative AI, data was the key to unlocking innovation. As Seagate’s Multicloud Maturity Report demonstrates; organisations with a high level of maturity in storing and managing data can bring new solutions to market six times faster than those with low maturity.
The problem is in being able to get ahead of the acceleration to achieve maturity. Two factors are at play here, as Rosalina Hiu, Seagate Global VP of Brand Strategy, explained:
One is the raw data itself. Seagate estimates that generative AI (GenAI) alone will produce 100 zettabytes (ZB) of data within the next four years.
The other is that the way that AI is used, fundamentally changes how organisations need to handle their existing data.
“With GenAI, retention needs to be for longer periods of time because AI can hallucinate and this can cause problems if it’s not managed,” Hiu said.
“The solution is that the historical truth of things needs to be preserved, because if we don’t have anything to refer back to, to determine what pieces are information are true, it will become hard to tell the difference with what is fake.
“Based on that, we believe that retention of data will be for longer periods of time, and perhaps indefinitely, depending on the kind of data, and this is going to create new storage pressures.”
And, since tomorrow’s AI tools will be able to derive yet-unimagined insights from yesterday’s data, organisations need to keep as much data as possible. However, data retention costs often lead to companies deleting data.
Hui expands, “No company has an unlimited budget. While some GenAI applications require speed, data preparation and training need tiered storage. Price per terabyte can be at least 6:1 flash to hard drives. Hyperscalers store 90% of their data on hard drives, a model enterprise likely follows. Here at Seagate, we are more like 99% hard drive based.”
Areal Density: The Solution to AI Storage Challenges
Hui shares how Seagate’s recent work has been focused on developing storage solutions that meet these new demands of AI. This can be broken down into three categories: capacity, efficiency and space optimisation.
- Increasing Capacity with Mozaic 3+
To address the burgeoning raw data needs, Seagate has been investing into expanding storage capabity. By advancing areal density, measured by capacity per disk, Seagate has been able to achieve density levels previously inconceivable within the same form factor of 3.5-inch hard drives, and with the same high levels of performance and reliability.
As Hiu shares, this brings Total Cost of Ownership (TCO) savings both in terms of acquisition and operating cost, and places less demand on the world’s resources. It has also enabled Seagate to offer customers 30+ terabyte (TB) storage solutions, with expectations to achieve 40+ TB and 50+TB in the near future.
- Enhancing Efficiency
AI is incredibly power-hungry, and the application of data is a big part of that. A 2024 presentation by SNIA on the “hidden” costs of AI highlighted the full extent of this: It takes around 100.5 mWh of power usage over 24 days within the server environment to train a 70B parameter model.
On the consumption end of AI, meanwhile, Carnegie Mellon University research showed that generating a single GenAI image required the equivalent amount of power as one smartphone charge.
As Hiu said, addressing some of this power draw through efficiency is a critical component of Seagate’s strategy. Its Mozaic 3+ drives are designed to conserve energy, boasting a 45% reduction in power usage and 55% reduction in embodied carbon reduction per terabyte. This not only helps reduce operational costs but also addresses some of the environmental concerns associated with large-scale data storage.
- Optimising Space with Areal Density
Critically, given the growing costs of physical space for data centres, the physical size of storage solutions also contributes significantly to the cost of AI environments. “We believe that one of the ways to really help businesses scale their operations and preserve data is by increasing areal density,” Hiu said. In short, more data can be stored in the same physical space, mitigating at least some of the challenges related to raw space demands.
Here, Seagate’s breakthrough with its Mosaic 3+ platform incorporating Seagate’s unique implementation of HAMR and a breakthrough collection of nanoscale technology to achieve 3 TB+ per disk platter, allows for single drives with up to 30 TB capacity and beyond. “If you move from 16 TB in your data centre rack to 30 TB, you can essentially double the capacity within the same space,” Hiu said.
Hiu said that looking ahead, Seagate is committed to pushing the boundaries of storage technology to keep pace with AI’s rapid evolution. The company is already prototyping even higher capacity drives with 4 TB and 5 TB per disk, which it said will help organisations keep pace with their growing capacity needs.
As AI continues to permeate various sectors, from hyperscalers through to edge environments and endpoint devices, there is going to need to be a fundamentally changed approach to how storage is structured and managed. By focusing on capacity, efficiency and space optimisation, Seagate aims to overcome the most pressing, practical concerns of CIOs and allow them to leverage AI while maintaining control over operational expenditure.
Hiu concludes, “Storage is the backbone for AI. No storage, no data, no AI.
“Business leaders will soon realise that their data storage and management strategies are a make-or-break driver of AI success. Companies that are not strategic about it, may have difficulties to scale.”
To learn more about how your business can harness AI storage, click here.
[ad_2]
Source link
