Adata's chairman has projected that the global DRAM shortage will continue for at least ten more years, dismissing concerns about an AI-driven bubble in memory markets until "2040 or 2050," in remarks carried by Tom's Hardware.

The projection places the current supply tightness in a structural rather than cyclical frame. The chairman's bubble timeline of "2040 or 2050" extends roughly fifteen to twenty-five years out, well beyond the planning horizon of current fab construction cycles, equipment lead times, and node transitions.

What exactly did the chairman say?

The remarks, as reported by Tom's Hardware, contain two central claims. First, the DRAM shortage has a remaining duration measured in decades, not quarters. Second, AI-driven memory demand is not on track to produce a bubble in the near or medium term.

The quoted phrasing "2040 or 2050" sets the implied bubble window at the far end of every forecast range currently in circulation.

Who is the Adata chairman addressing?

Tom's Hardware carries the report directly. Adata is best known as a memory module and storage integrator, particularly across NAND SSDs and DRAM modules. Its commentary on spot pricing and contract dynamics carries weight beyond its own wafer demand because module partners supply the channel that distributes the major chip makers' output.

Why does a DRAM shortage matter beyond memory makers?

DRAM is the volatile working memory that feeds CPUs, GPUs, and AI accelerators across the entire compute stack. Tightness in this segment lifts prices for consumer DIMMs, server memory, mobile LPDDR, and graphics memory simultaneously, because capacity decisions at the wafer level cascade through every downstream channel.

A ten-year shortage projection therefore implies structurally higher memory content costs for PCs, servers, and mobile devices, with limited substitution options at the system level.

What does "no bubble until 2040" actually mean?

A bubble in memory markets traditionally refers to a price spike driven by speculative over-ordering or capacity overshoot, followed by a sharp correction that idles fabs and burns inventory. Past examples include the 2017–2018 server DRAM spike and the 2021 gaming and crypto-driven consumer DRAM rally.

The chairman's framing rejects that pattern for AI demand, suggesting that supply additions will track — but not outpace — actual compute deployment through the next two decades.

How does this shift planning assumptions?

If the projection holds, several downstream dynamics follow:

• Module and DIMM pricing stays elevated relative to pre-2023 baselines through multiple product cycles
• HBM (high-bandwidth memory) for AI accelerators continues absorbing a growing share of leading-edge wafer output
• Server and hyperscaler buyers lengthen qualification cycles for alternative memory architectures, including CXL-attached DRAM pooling
• Consumer-channel buyers face persistent allocation pressure through peak build seasons

What's next for buyers and module makers?

The chairman's remarks provide the most concrete public signal to date from Adata on the cycle outlook. Subsequent industry commentary — at events such as CES, MWC, and the semiannual earnings calls of the three primary DRAM suppliers — will test how broadly the no-bubble framing is shared across the supplier base.

For module integrators and channel buyers, the operational signal is unchanged: prioritize supply relationships over spot purchasing, and price systems around the assumption that DRAM content cost will not relax through the rest of the decade.