Dell Technologies CEO Michael Dell has issued a stark warning to the global tech industry, projecting that memory demand for AI infrastructure will explode 625 times by 2028.
Speaking at a Bank of America event on April 8, 2026, Dell described a “structural mismatch” where the speed of AI adoption is far outstripping the years-long cycle required to build new semiconductor fabrication plants.
1. The “25 × 25” Logic
Michael Dell explained that the 625-fold surge is the compounding result of two simultaneous “25-fold” increases in the AI hardware ecosystem.
- 25x Capacity per Unit: In 2022, a flagship AI accelerator like the NVIDIA H100 featured roughly 80GB of High Bandwidth Memory (HBM). By 2028, next-generation accelerators are expected to pack up to 2TB of memory each—a 25x increase in per-chip density.
- 25x Deployment Scale: Simultaneously, the sheer volume of these accelerators being deployed in data centers is expected to grow by another factor of 25 to meet the training and inference needs of trillion-parameter models.
- The Result: $25 \text{ (per-unit capacity)} \times 25 \text{ (deployment scale)} = 625\text{x total demand}$.
2. Supply: The Multi-Year Bottleneck
While demand is vertical, supply is horizontal. Dell warned that the “worst is yet to come” for memory prices due to historic underinvestment.
- Capital Expenditure Cuts: Memory giants like Samsung, SK hynix, and Micron significantly cut spending in 2023 during a market downturn. This created a “supply hole” just as the AI boom took off.
- Infrastructure Lag: Building a new “Fab” (semiconductor factory) takes 3 to 5 years. Consequently, the industry cannot simply “turn on” more capacity to meet the 2026–2028 surge.
- Pricing Power: Dell believes memory manufacturers now hold total “pricing power,” and enterprises may soon have to pay “whatever price is demanded” to secure the hardware needed for survival.
3. Impact on Consumer Electronics
The “AI memory vacuum” is already starving the consumer market, as manufacturers prioritize high-margin data center chips over standard PC and smartphone RAM.
| Device Category | Projected Price Increase (by late 2026) | Reason |
| Laptops/PCs | 20% to 35% | Higher component costs for RAM and SSDs. |
| Gaming Consoles | 15% to 20% | Tight availability of specialized GDDR memory. |
| Enterprise Servers | High Volatility | Shift to “hourly pricing” models for some hardware. |
4. The “Sovereign AI” Engine
A major driver of this demand is Sovereign AI—nations building their own autonomous AI clusters to avoid dependence on foreign tech.
- Top 25 GDP Countries: Dell noted that nearly all the world’s leading economies are currently advancing state-funded AI infrastructure projects.
- Geopolitical Stakes: Nations now view compute and memory as “strategic reserves,” similar to oil or food, further intensifying the global scramble for chips.
5. Dell’s Verdict: “No Choice but to Invest”
Despite rising costs, Michael Dell argues that businesses cannot afford to wait for a price drop.
- The Productivity Trap: He famously remarked, “It doesn’t make sense for a knowledge worker earning $100,000 a year to use an outdated PC.” * The Strategy: For enterprises, the question is no longer “whether to buy,” but “when to buy” to avoid being locked out of the next three years of supply.
“A structure is forming where total memory demand is increasing at a rate the world has never seen,” Dell stated. “The industry is moving from a ‘buy-as-needed’ model to a ‘buy-whatever-is-available’ model.”