HBM Memory Prices Expected to Climb Further in 2027 as AI Demand Keeps Pressure on Supply
High-bandwidth memory prices may continue rising into 2027 as the artificial intelligence boom keeps demand elevated and gives memory suppliers more leverage in contract negotiations, according to market research firm TrendForce.
The rapid expansion of AI servers, data centers, and advanced GPUs has already tightened the memory market. HBM, a key component used in powerful AI accelerators, remains one of the most important technologies behind modern AI infrastructure. As companies prepare for the next generation of AI chips, memory makers are expected to push for higher HBM pricing in upcoming supply agreements.
TrendForce says the current pricing structure for memory contracts has limited how quickly manufacturers can react to market changes. Many memory supply deals are based on annual contract pricing, which means suppliers may not be able to immediately raise prices even when demand surges or supply becomes constrained. With memory prices rising quickly after the AI boom, this contract model has prevented some manufacturers from fully matching their prices to real-time market conditions.
As negotiations shift toward HBM4 memory for 2027 supply, memory companies are expected to use the current tightness in the broader DRAM market to strengthen their position. TrendForce’s analysis suggests that certain DDR5 memory products have recently become more profitable on a per-wafer basis than HBM.
The firm compared HBM and DRAM profitability using factors such as die size, yield rates, and per-gigabit pricing. Based on this analysis, HBM revenue per wafer fell below that of DDR5 64GB RDIMM modules in the first quarter of the year. This means some high-capacity DDR5 server memory products may currently offer better profitability than HBM, despite HBM’s critical role in AI hardware.
That shift could have major consequences for HBM pricing. If buyers resist higher HBM prices, memory manufacturers may choose to allocate more limited production capacity toward profitable DRAM products instead. This would further reduce available HBM supply, creating even more pressure in a market already dealing with strong demand from AI chipmakers.
In other words, memory suppliers may have two advantages at once: rising DRAM profitability and sustained HBM demand. This combination could allow them to negotiate higher HBM prices for future contracts, especially as AI hardware companies compete to secure enough supply for next-generation accelerators.
TrendForce expects HBM demand to remain strong throughout 2026 and 2027, although the main drivers may shift over time. In 2026, custom AI chips are expected to play a major role in boosting HBM consumption. By 2027, demand could be pushed even higher by NVIDIA’s Rubin Ultra platform and other advanced AI GPU launches.
The continued growth of AI infrastructure is likely to keep memory supply chains under pressure. Cloud providers, enterprise AI developers, and chip designers all require massive amounts of high-performance memory to support increasingly powerful AI models. Since HBM production is technically complex and capacity expansion takes time, supply may struggle to keep up with demand.
For the broader technology market, rising HBM prices could influence the cost of AI servers, GPUs, and data center deployments. AI hardware is already expensive, and higher memory costs may add further pressure to companies building large-scale AI systems.
Overall, the outlook points to a memory market where HBM remains in high demand, DRAM profitability strengthens supplier bargaining power, and contract negotiations for 2027 could lead to another round of price increases. As AI adoption continues to accelerate, high-bandwidth memory is likely to stay one of the most strategically important and closely watched segments of the semiconductor industry.
