Nothing CEO Carl Pei Warns of Rising Smartphone Prices as DRAM Costs Surge

Nothing CEO Carl Pei Warns of Rising Smartphone Prices as DRAM Costs Surge
Memory has become the single most expensive component in a smartphone, according to Nothing co-founder and CEO Carl Pei — a claim he made publicly on X on 12 June 2026, citing the cost trajectory of the Nothing Phone 4A as a concrete example.
Pei's post described a cost escalation that will be familiar to anyone who has watched DRAM and NAND markets cycle through supply crunches before, but the numbers he cited are stark. Memory pricing for the Phone 4A doubled between the point at which the device's bill of materials was locked in during the design phase and its commercial launch. It then doubled again after launch. That is a 4× cost increase across a single product's lifecycle — a figure that compresses an entire commodity supercycle into the span of one SKU.
The practical consequence is straightforward. When a component doubles and then doubles again while a product is already in market, a manufacturer faces three options: absorb the margin hit, reprice the device mid-cycle, or pass the accumulated cost into the next generation. For a company like Nothing — which has built its positioning around accessible price points relative to Samsung and Apple — none of those options is painless.
Pei did not specify in the post whether Nothing has already adjusted retail pricing or intends to. What he did, by going public with the cost data, is shift the framing around any forthcoming price increases from a brand decision to a supply-chain reality. That is a recognizable communications move: get the supply constraint on the record before the price tag changes.
The broader memory market context supports the direction of what Pei is describing. DRAM and NAND flash have historically been among the most volatile commodity markets in consumer electronics, subject to sharp price swings driven by capacity investment cycles, yield improvements, and demand surges. The current AI/ML infrastructure buildout has added a demand vector that did not exist in previous cycles: HBM (high-bandwidth memory) and high-capacity LPDDR5X are being pulled hard by data-center and on-device AI workloads, tightening supply for the same DRAM fabs that produce smartphone memory. Smartphone OEMs are, in effect, competing for wafer capacity with hyperscalers.
Worth flagging here: Pei's post is a single first-person account from an interested party, and he has not released the underlying cost data or named specific suppliers. The claim that memory is now the costliest line item — above the application processor, display, or camera assembly — is a significant one. It is plausible given current spot pricing, but until other OEMs or supply-chain analysts corroborate the specific ranking, the assertion should be read as Nothing's internal view of its own BOM rather than an industry-wide audited figure.
That caveat aside, the directional signal is credible. Memory's share of smartphone BOM cost has been climbing for several years as average DRAM densities per device have risen — driven by camera buffer requirements, on-device ML model storage, and consumer expectations around multitasking. A mid-range device shipping with 8 GB or 12 GB of LPDDR5 today carries a meaningfully larger memory footprint than a flagship did five years ago. More gigabytes at higher per-bit prices compounds quickly.
For product managers and procurement teams at other Android OEMs reading this: if Pei's cost curve applies beyond Nothing's specific supplier agreements, the industry is heading into a period where memory alone could force a structural reset in how mid-range devices are positioned. Either specs get trimmed, or price floors rise. Neither outcome is good for the segment that has driven global smartphone volume growth over the past three years.
The longer-term question is whether on-device AI requirements — which demand more LPDDR and faster NAND — will accelerate or moderate that pressure. Right now, they are accelerating it. The market dynamic that makes AI features a selling point is the same one that is making the hardware to run them more expensive. That tension is not resolved by a single CEO's post on X, but Pei has at least put a concrete number to it.


