Why are RAM prices so high right now?

The surge stems from a confluence of three memory makers—Samsung, SK Hynix and Micron—being forced to allocate most of their wafer output to high‑bandwidth memory (HBM) and massive DRAM orders for AI giants. Micron’s CEO Sanjay Mehrotra warned that a single HBM stack consumes three times the wafer area of a DDR5 chip, turning every HBM module into a three‑fold DRAM opportunity cost. Data‑center projects such as Elon Musk’s xAI “Colossus” in Memphis, with 555,000 NVIDIA GPUs, are slated to consume roughly 112 million GB of RAM, an amount equivalent to the memory in 14 million iPhone 17s. A separate OpenAI‑AMD pact to build 6 GW of AI compute includes a 1 GW site using Instinct MI450 GPUs, each supporting up to 432 GB of RAM, which alone could tie up another 194 million GB. DigiTimes reports that all RAM capacity slated for 2027 has already been contracted, confirming that the supply chain is effectively booked out for years.

The RAM market, which for decades grew predictably alongside PC and smartphone cycles, is now being reshaped by a single, cash‑rich demand class. Historically, manufacturers could match incremental upgrades in OS requirements or modest data‑center growth with modest fab expansions. Today, AI firms are buying entire years of production, leaving consumer‑grade DDR5‑4800 kits that cost $100 in early 2025 to sell for $400, and high‑speed DDR5‑6000 32 GB sticks that were $250 now fetching $1,200. This “RAMpocalypse” mirrors earlier commodity squeezes (e.g., the 2021 GPU shortage) but is amplified by the wafer‑intensive nature of HBM and the sheer scale of AI compute. The result is a permanent upward pressure on memory prices, with Deloitte projecting relief only after 2030, and both SK Hynix and Micron executives echoing that demand will outstrip supply well beyond 2027.

Looking ahead, the persistence of high RAM prices will force OEMs to redesign products around lower‑capacity configurations or absorb cost increases, potentially eroding margins on mid‑range laptops and smartphones. Investors should monitor any announced fab expansions from the three memory giants, as well as policy moves that could accelerate domestic wafer production. A sudden contraction in AI funding could release some inventory, but given the multi‑year contracts already in place, a rapid price correction appears unlikely.