Cloud and AI company demand has pulled memory, SSD, and GPU supply out of the consumer market. A CNET analysis argues the build versus buy math has flipped across budget, midrange, and high end tiers.
For most of the last decade, the smart move for anyone who wanted a fast computer was to build it. Pick a case, a motherboard, a power supply, the right graphics card, and the savings over a comparable prebuilt were real and reliable. As of June 2026, that math no longer holds in the United States, and it is not because prebuilt manufacturers suddenly got better at their jobs.
The shift, documented in a CNET analysis by ten-year DIY builder Mark Knapp, comes down to who is buying the same components hobbyists need. Hyperscalers, the cloud and AI companies running large model training and inference fleets, have absorbed a large share of the DRAM, NAND flash, and high-end GPU supply that consumer builders used to count on. The result is a component crunch that has hit memory (RAM), solid-state storage (SSDs), and graphics cards, in that order of pain.
Knapp anchored the claim with a side-by-side build comparison across multiple price tiers. In the budget segment, a Sam's Club HP Omen 16L at $1,199 cost $345 less than an equivalent DIY build ($1,544). In the midrange, a Sam's Club Asus ROG G700 at $1,749 was $284 cheaper than the comparable DIY configuration ($2,033). At the high end, a Micro Center PowerSpec G757 beat DIY by $632, and a Micro Center Alienware Aurora ACT1250 at $2,399 undercut DIY by $532.
The story the components market is telling is not a normal post-launch price dip or a tariff blip. AI compute capacity is being added on a multi-year buildout, and the suppliers that make DRAM, NAND, and advanced GPUs are allocating capacity to those customers first. OEMs and boutique builders, with their bulk purchase agreements and direct vendor relationships, can absorb the allocation in ways a single buyer walking into a Micro Center cannot. The consumer's leverage point in the market has moved.
It is worth flagging the basis for the claim. The CNET analysis is a single-source snapshot, and the comparisons are specific builds rather than a market-wide index. The DRAM and NAND price moves are widely reported across the industry, but the specific inversion Knapp describes — that DIY now loses to prebuilt across tiers — has not yet been confirmed against an independent OEM-versus-DIY cost index. Treat the conclusion as a credible signal, not a settled market fact.
For anyone planning a new PC in 2026, the calculus has changed. The old default, that any reasonably informed DIY build would beat a prebuilt on price, is no longer reliable across tiers. Prebuilts and boutique builders have a structural cost advantage in this market, and the gap is widest precisely on the parts that drove DIY savings in the first place: RAM, storage, and the GPU. The build-versus-buy decision has not gone away. It has just moved to a different spot on the price chart, and the spot is unfavorable to anyone who assumed DIY was always the cheaper answer.