The AI Challenge: Meeting Data Demands with Efficient Solutions

By Dave Sierra, Efficiency Evangelist, Solidigm

Artificial intelligence (AI) is transforming industries, but its rapid growth comes with a significant challenge: an unprecedented demand for compute power. By 2028, over 50 percent of data center critical IT power is expected to be driven by AI, while data centers already account for more than 2 percent of global electricity consumption — a figure projected to rise to 4.5% by 2030.¹ This creates a dilemma for enterprises: How can they meet AI’s escalating resource needs while adhering to long-term power efficiency objectives?

At least a partial answer may lie in how we approach data storage. Traditional hard disk drives (HDDs) have long been the go-to solution for their cost-effectiveness, but they come with significant drawbacks — high power consumption, large physical footprints, and reliability challenges. In contrast, ultra-high-capacity solid-state drives (SSDs) like the Solidigm™ D5-P5336 122 terabyte (TB) SSD, offer a compelling alternative, significantly reducing both storage energy and space requirements. This article explores how high-capacity SSDs can help reconcile AI’s data demands with the urgent need for sustainable practices, offering a path toward more responsible data centers.

The Problem with Hard Disk Drives: Old Tech, New Problems

HDDs just turned 69 years old. IBM introduced the world to the hard disk in 1956 in the form of the 350 Disk File. HDDs have been a staple in data centers in the many decades since, but their mechanical design is increasingly at odds with the demands of AI and analytics workloads. HDDs are inherently slower, less reliable, and physically much larger than SSDs, creating inefficiencies that ripple across the data center ecosystem. Hard drive vendors themselves acknowledge that raw performance is dropping as drive capacities increase,² a fact that’s at odds with the growing AI data and performance imperatives.

A case in point: You’d need at least four cutting-edge 30TB HDDs to match the capacity of a single 122TB SSD. However, HDD challenges go beyond the physical “size” of your data. To ensure reliability, many storage designers employ tactics like 3x replication — storing the same piece of data on three separate drives to mitigate the risk of failure.

Just as it sounds, this triples the number of drives required, along with their power and space demands. Other techniques like short-stroking — using only a portion of an HDD’s capacity to improve performance — can further inflate the number of drives deployed. Accelerated HDD replacement cycles can further exacerbate the issue. Over a five-year period, these mitigation techniques can vastly increase the number of HDDs purchased, driving up their cost and environmental impact.

TABLE 1: Comparing HDD vs. SSD Footprint

The AI-Driven Power Crunch: Every Watt Counts

The power demands of modern data centers are a growing concern, and AI is a major contributor. Storage, while often overlooked, plays a critical role in this equation. Every drive in a data center consumes energy, and as AI workloads scale, the inefficiencies of legacy storage solutions become more pronounced.

Consider a 50 petabyte (PB) Network Attach Storage (NAS) solution designed for performance and reliability. Using 30TB HDDs, you’d need 7,143 drives, each consuming an average of 9.8 watts during active read-write operations. That totals more than 70 kilowatts (kW) of power for the storage devices alone. If you applied this reasoning to the build out of data centers at scale, it could mean saving enough electricity to power an entire community.

In contrast, the same solution using the Solidigm 122TB SSDs requires just 857 drives, each using 16 watts during active reads, for a total of 13.7 kW. That’s an 80 percent reduction in power consumption — 56 kW saved on storage devices alone. The benefits extend further: the HDD setup requires 596 servers to house the drives, while the SSD solution needs only 36, further reducing power demands for compute and cooling.³

Table 2: Comparing HDD vs. SSD Solution-Level Energy Consumption

In an era where data center energy consumption is doubling, these savings are significant. High-capacity SSDs not only reduce the direct power footprint of storage but also alleviate strain on the broader data center infrastructure making them a critical tool for sustainable scaling.

Space Efficiency: Reducing the Physical Footprint of Storage

AI’s data demands don’t just strain power grids — they also require physical space, potentially lots of it. For a hypothetical 5exabyte (EB) NAS deployment, the difference between HDDs and SSDs is stark. A solution using 30TB HDDs would require 775 racks, requiring over 290,000 square feet of data center space.³ The same 5EB with 122TB SSDs needs just 92 racks and 36,000 square feet — an 8.4:1 NAS space reduction. A single 122TB SSD can enable up to 3 petabytes in a 2U server,⁴ showcasing density that HDDs simply can’t match.

Table 3: Physical Building Requirements

This space efficiency has profound environmental implications. Constructing new data center facilities requires materials like concrete and steel, both of which carry a high carbon cost. Collectively the production of concrete and steel account for over 15 percent of global greenhouse gas emissions.⁵ By reducing the physical footprint of storage, high-capacity SSDs can help reduce embodied concrete and steel-based emissions by more than 80 percent compared to HDDs. A smaller storage footprint also reduces cooling requirements, a major energy demand in data centers. The ripple effect is convincing: Optimizing storage space doesn’t just save square footage, it can help curb the broader environmental impact of AI’s physical infrastructure.

Table 4: Greenhouse Gas Impact of HDD vs. SSD Storage Footprint

The Less Obvious Benefits of Solid-State Storage

Beyond power and space, SSDs offer less obvious but equally important efficiency advantages. Unlike HDDs that rely on mechanical components that are prone to wear, SSDs are solid-state, resulting in greater durability and fewer failures. The Solidigm 122TB SSD, for example, offers unlimited random write endurance over five years,⁵ reducing the need for replacements. In a 50PB NAS solution, the HDD solution requires 7,143 drives, while the SSD solution needs just 857.

Over a five-year deployment, that’s thousands fewer devices to manufacture, ship and dispose of, significantly reducing storage e-waste.³ By the same equation, builders could reduce construction emissions by nearly 500 tons of carbon. Applying this logic to new data center construction at scale would be the equivalent of taking thousands of cars off the road every day.

Table 5: HDD vs. SSD End-of-Life Disposition Impact 

Then, there’s the ripple effect. Fewer, smaller SSDs also generate less heat than HDDs, lowering solution-wide cooling demands — a substantial energy sink in data centers. Far fewer expected SSD failures⁶  translate to a smaller manufacturing and shipping impact, both of which contribute to carbon emissions. From production to end-of-life, high-capacity SSDs offer a more sustainable approach to storage, aligning AI’s data needs with a smaller environmental footprint.

Summary: A More Manageable Path for Data Center Storage

As AI continues to drive innovation, its environmental impact cannot be ignored. Traditional storage solutions like HDDs are becoming increasingly unsustainable with their high-power consumption, large physical footprints, and frequent replacement cycles. Solidigm™ D5-P5336 122TB SSD offers a powerful alternative, reducing power usage by up to 80 percent, shrinking data center footprints by a factor of eight, and minimizing e-waste through greater performance, efficiency and durability.

Earth Day prompts a timely moment to reflect on how we can support AI’s growth without compromising our planet’s future. High-capacity SSDs provide a practical solution, enabling enterprises to meet the extreme data demands of AI while advancing overall efficiency. By rethinking storage, we can propel AI’s transformative potential and simultaneously align with environmental responsibility — proving that innovation and sustainability can go hand-in-hand.

_{Solidigm does not control or audit third-party data. You should consult other sources to evaluate accuracy.
Some results have been estimated or simulated using internal Solidigm analysis or architecture simulation or modeling, and provided to you for information purposes only. Any differences in your system hardware, software or configuration may affect your actual performance.} 

_{Nothing herein is intended to create any express or implied warranty, including without limitation, the implied warranties of merchantability, fitness for a particular purpose, and non-infringement, or any warranty arising from course of performance, course of dealing, or usage in trade.}

_Appendix:

¹ _{Source – Semianalysis, March 2024: https://semianalysis.com/2024/03/13/ai-datacenter-energy-dilemma-race/}

² _{Source – Western Digital, May 2023 https://blog.westerndigital.com/dual-actuators/}

³ _{Source – Solidigm. Some results have been estimated or simulated using internal Solidigm analysis or architecture simulation or modeling, and provided to you for information purposes only. Any differences in your system hardware, software or configuration may affect your actual performance.}

⁴ _{Source – Solidigm. 3 petabytes based on 24 SSDs in a 2U server times 122.88TB capacity per drive} 

⁵ _{Source – Solidigm. Unlimited random write endurance based on Solidigm™ D5-P5336 122.88TB SSD 32K random write performance and 134 petabytes written (PBW) specifications}

⁶ _{HDD AFR Source – Backblaze, 2024 end-of-year drive stats: https://www.backblaze.com/blog/backblaze-drive-stats-for-2024/. SSD AFR figure is based on 24-month AFR actuals for Solidigm’s 4th generation QLC products.}