The storage industry is watching a bold claim: Pure Storage says it will stop selling new HDD-based systems after 2028.
If true, it implies that all new HDD capacity purchased in 2029 would instead be replaced by SSDs, raising a critical question:
Can global NAND manufacturing capacity produce enough flash memory to replace new HDD demand by 2029?
This article evaluates that possibility from the perspective of NAND manufacturing fundamentals, temporarily setting aside TCO comparisons.
HDD and SSD Price Trends #
HDD shipments have been declining for more than five years as SSDs replace them across laptops, desktops, and many enterprise workloads. SSDs benefit from:
- No mechanical seek delays
- Higher IOPS
- Faster response times
- Continuous density improvements (shrinks, more 3D NAND layers, TLC/QLC adoption)
Meanwhile, SSD prices per TB have fallen faster than HDDs:
Enterprise SSDs remain more expensive, but the price gap has narrowed:
Some analysts suggest IT buyers will switch to SSDs once the price difference reaches ~5×. Today, the delta remains ~9–10×, but falling.
Pure Storage argues that QLC SSDs already offer a lower 5-year TCO than nearline HDD-storage arrays — and that flash density improvements will continue pushing costs down.
However, cost alone is not the only barrier.
The bigger question is manufacturing capacity.
NAND Manufacturing Capacity: Can It Scale? #
Micron’s Colm Lysaght stated in 2019:
The exabytes required to fully replace nearline HDDs are too large for the NAND industry to handle…
The capital investment required is immense.
Other analysts disagree. For example, Wikibon’s David Floyer predicted that SSDs could be cheaper than HDDs by 2026, assuming sufficient NAND scaling.
Pure’s 2023 position aligns more with Floyer: they believe NAND supply will not be the limiting factor.
So, who is right?
To explore this, we built a simplified NAND capacity model.
Modeling NAND Capacity Through 2029 #
Key Inputs (2022 Baseline) #
- HDD shipments: 1320 EB
- SSD shipments: 277 EB
- NAND manufacturing output: ~538 EB (2021)
Data Growth #
IDC’s Datasphere suggests a 21.2% CAGR in data generation from 2026–2029.
Applying this rate:
- HDD-equivalent demand in 2029: 5071 EB
- NAND demand for existing SSD growth in 2029: 2067 EB
- Total NAND needed: 7137.8 EB
Manufacturing Growth Assumptions #
Jim Handy notes:
- 3D NAND layers roughly double every 2 years
- ~40% annual effective output growth from higher layers
- QLC adds 33% more bits per die than TLC
Combining these:
- Additional layer-driven capacity: 3375.9 EB
- QLC uplift: 33.3%
Final projected NAND output: ~7543 EB
This is 405 EB short of the 7138 EB required — a 5.7% gap, which is within model error margins.
A more conservative assumption (30% layer growth) increases the shortfall to ~2648 EB.
Vendor Perspectives #
Dell #
HDD demand will continue beyond 2028; innovation in low-cost HDDs remains strong.
Seagate (CTO John Morris) #
It is impossible for NAND to replace HDDs.
NAND cannot reach the cost per bit required, and HDD+SSD will coexist long-term.
He also argues analysis should focus on nearline HDD vs enterprise SSD, where HDD volume is ~7× larger.
Pure Storage #
Total NAND output is 2.5× larger than SSD-only NAND usage.
Non-SSD NAND (phones, tablets, automotive, USB) could shift toward SSD markets.
Data reduction, higher performance, and better durability mean SSD replacement is not 1:1 with HDDs.
Conclusion #
Based on simplified modeling, NAND manufacturing can come close to supporting full HDD replacement by 2029, with:
- A 5.7% shortfall assuming aggressive (but historically reasonable) 40% yearly layer growth
- Capacity potentially becoming sufficient when accounting for Pure’s argument that global NAND output is far larger than SSD consumption
The reality will depend on:
- continued 3D NAND layer scaling
- expansion of QLC/PLC adoption
- new fab investments
- whether nearline HDD demand declines faster than expected
For now, the numbers suggest NAND capacity is not an absolute barrier, and replacing new HDD shipments by 2029 appears technically feasible, though far from guaranteed.
We’ll continue tracking developments as the industry evolves.