SSD and DRAM prices to remain high through 2030 per TrendForce

The immediate shift in manufacturing capacity toward AI-optimized hardware is creating a structural deficit for standard enterprise storage components. For streaming infrastructure providers, this translates to significantly higher capital expenditures for content delivery networks and data center expansions that rely on high-density SSDs. As DDR5 becomes the baseline for modern server architecture, the fivefold price hike forces a difficult choice between delaying hardware refreshes or absorbing massive cost increases. Watch for whether major cloud providers adjust their storage egress or instance pricing in response to these sustained hardware premiums through 2030.

The memory supply crunch extends well beyond consumer pricing. TrendForce's analysis reflects a broader industry pattern where manufacturers are reallocating wafer capacity toward high-bandwidth memory and advanced packaging for AI accelerators, leaving conventional NAND and DRAM segments undersupplied. In its June 2025 Mobility Report, Ericsson noted that AI-native workloads are fundamentally shifting the nature of network traffic, with uplink-intensive applications such as AR-embedded AI agents flooding networks with video and sensor data. That traffic growth directly increases storage and processing demands at the edge and in core data centers, compounding the cost pressure from elevated SSD and DRAM prices for streaming operators planning capacity expansions. The competitive dynamics around memory allocation are intensifying as AI infrastructure spending absorbs an ever-larger share of fab output. At MWC 2026, Ericsson's networks chief Per Narvinger highlighted that AI is reshaping connectivity requirements across the telecom stack, noting that uplink traffic patterns will change as AI agents move from text-based interactions to camera and video-heavy workloads. This shift means streaming and CDN operators will need denser, faster storage arrays precisely when NAND flash pricing is at its most punishing. The structural mismatch between AI-driven demand for advanced memory and the commodity NAND needed for video caching and content delivery creates a multi-year budgeting challenge for infrastructure teams. On the technical side, the agentic AI wave is adding new compute and storage layers to network operations themselves. Blue Planet and Telefónica Deutschland completed a proof of concept using agentic AI for 5G network slicing, demonstrating that AI-driven orchestration can compress slice design from weeks to minutes. Meanwhile, Ericsson's agentic AI framework processes data from over 60,000 KPIs to identify 20 distinct classes of network issues, claiming an 80 percent reduction in analysis and decision-making time. These AI-native operational tools require their own storage and memory resources, adding another layer of demand on the same constrained NAND and DRAM supply chains that streaming infrastructure depends on. The net effect is a feedback loop where AI both drives the price increases and creates new consumption of the very components becoming scarcer.