Klaba’s warning shot, fired via a recent post on X, highlights a critical bottleneck: the shift in memory production towards High Bandwidth Memory (HBM) for GPUs, leaving less capacity for the RAM and NVMe drives that underpin standard cloud infrastructure.
The problem isn’t just about AI needing more memory; it’s about where that memory is going. As AI workloads explode, memory manufacturers are prioritizing HBM production, leading to a squeeze on the supply of traditional memory components. This creates a ripple effect, driving up prices across the board.
“This in turn increases price pressure on all types of RAM and NVMe drive components, not just those for AI,” Klaba stated, painting a picture of across-the-board cost increases.
Analysts Sound the Alarm
Klaba’s prediction isn’t an isolated one. The memory market has been flashing warning signs for months. TrendForce, a memory-centric analyst firm, reported a dramatic surge in spot prices, with 1Gx8 DDR4 memory jumping 158% and DDR5 2Gx8 skyrocketing by 307% since September 2025. Counterpoint Research has gone even further, predicting prices will double.
Adding fuel to the fire, Samsung, a major memory manufacturer, has reportedly already increased prices by 60 percent, signaling a new era of higher memory costs.
So, what does this mean for cloud users? Klaba believes that existing stockpiles of components might delay the inevitable until around June 2026. However, he anticipates server costs to increase by 15-25% between December 2025 and early 2026, leading to a 5-10% price hike for some cloud services between April and September 2026.
“These are estimates based on the information we have as of November 2025,” Klaba cautioned. “This could accelerate.”
The million-dollar question is whether these price increases will trigger a mass exodus from the cloud. While some companies, like Grab and 37 Signals , have successfully repatriated workloads to on-premise infrastructure, analysts remain skeptical about a widespread trend.
The reality is that managing AI-intensive workloads requires specialized hardware and expertise that many organizations lack. Hyperscale cloud providers often have preferential access to the latest GPUs, making them an attractive option for AI deployments, regardless of price.
The coming months will reveal whether rising cloud costs will force a fundamental shift in how businesses approach their infrastructure, or whether the allure of AI will outweigh the financial burden. One thing is clear: the AI revolution is reshaping the cloud landscape in more ways than one.
