Loading today’s stories
Product1 distinct publisher3 min readUpdated
Denser NAND and a fall from more than $8,000 per gigabyte to under a dollar made small drives pointless to build. Kingston's DataTraveler line now starts at 64GB.
The Product Desk · Product desk
Compiled by The Product DeskSomething wrong?How this is made
The consumer floor for USB flash drives has settled around 64GB: most of Kingston's current DataTraveler models start there, with some reaching 512GB [1]. For anyone provisioning devices or shipping media in volume, that floor is a budgeting fact, not a marketing one, and it follows from the cost structure of the product rather than from anybody's preference for bigger numbers.
Start with what has changed and what has not. A flash drive still needs a USB connector, a controller, a circuit board, a casing and one or more memory chips [2]. What collapsed is the price of the chips: USB flash storage went from more than $8,000 per gigabyte in its early years to 94 cents per gigabyte by 2013, according to Engadget's account [3], a decline of roughly 8,500-fold [4]. In early 2002, JMTek's USBDrive and its up-to-1GB capacity were considered notable [5]. The floor is now about 64 times that [6].
The density work is the mechanism. NAND makers shrank cells, packed multiple bits into each one, and eventually stacked them vertically with 3D NAND [7]. A clean illustration came in 2005, when Toshiba and SanDisk announced an 8Gb NAND chip holding 1GB on a single die, less than 5 percent larger than their previous-generation 4Gb part but with twice the capacity [8]. The companies said it would become the "production workhorse ... bringing significant cost reductions" for their flash products [9]. TLC NAND, which puts three bits in each cell, is now common in cost-sensitive consumer storage such as USB drives [10].
That is why the small SKU stopped making sense. Shrinking a 64GB drive to 1GB eliminates none of the controller, connector, board, casing or packaging, and none of the assembly, shipping or shelf space [11]. Once high-density NAND is cheap at volume, cutting capacity does not cut the finished product's cost by anywhere near as much [12]. The premium sits at the top of the range instead: a 2.1GB Verbatim Store 'n Go went for $250 in 2004, about $119 per gigabyte [13][14], and Kingston's 512GB DataTraveler cost $1,750 in 2013, roughly $3.42 per gigabyte, or about 3.6 times the average price per gigabyte that year [15][16][17].
There was no threshold at which the economics flipped and no single day when 8GB retired 1GB; the industry drifted [18]. Demand drifted too. Microsoft's Windows installation-media tool requires a blank drive with at least 8GB, which disqualifies a 1GB stick from one of the most ordinary jobs a spare drive gets [19].
The supply side now runs the other way for older parts. Global MLC NAND capacity is expected to fall 41.7 percent year over year in 2026 as major suppliers cut or halt output and shift resources to newer processes [20], leaving roughly 58 percent of the prior year's capacity [21], with the result that an older NAND type can get more expensive as supply shrinks [22].
Small capacities do survive where a fixed hardware configuration, compatibility or endurance outranks capacity: Delkin sells industrial USB drives starting at 1GB, and Apacer goes down to 256MB [23].
Two things to watch. First, whether the MLC contraction pushes legacy and embedded pricing up enough to change replacement plans on fielded hardware [20][22]. Second, resist reading capacity as throughput on spec sheets: larger is not automatically faster, and the NAND, controller and USB interface all decide the result [24].
Follow any of these and your For You feed starts watching them — no settings page required.
Ranked by verification strength, evidence, and original report placement.
Most of Kingston's current DataTraveler models start at 64GB, with some reaching 512GB.
A flash drive still requires a USB connector, controller, circuit board, casing and one or more memory chips; the basic recipe has not changed much.
USB flash storage went from more than $8,000 per gigabyte in its early years to 94 cents per gigabyte by 2013.
In early 2002 the industry was abuzz about JMTek's USBDrive and its up-to-1GB capacity.
For decades NAND manufacturers have made memory cells smaller, packed multiple bits into each cell, and eventually stacked cells vertically with 3D NAND.
In 2005 Toshiba and SanDisk announced an 8Gb NAND chip capable of storing 1GB on a single chip; it was less than 5 percent larger than the previous-generation 4Gb part but offered twice the capacity.
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
Concrete figures, single unverified source
The cluster rests entirely on one consumer-tech explainer. Its historical anchors are specific and internally consistent (per-gigabyte endpoints, 2005 chip dimensions, 2004 and 2013 retail prices), which raises checkability, but there are no primary vendor documents, filings or analyst reports in the supplied material, and the one forward-looking figure is unattributed.
Product-line evidence, no market-share data
Adoption of the higher capacity floor is shown through observable product lineups and a mainstream tooling requirement rather than shipment or share data: Kingston consumer drives starting at 64GB, industrial vendors still listing 1GB and 256MB parts, and an 8GB minimum for Windows media creation. That is real but narrow vendor-level evidence, with no unit volumes or channel data to size the shift.
Slight overclaim on causality
The framing that small drives died specifically of bill-of-materials arithmetic is plausible and consistent with the cited price collapse and fixed component list, but no cost breakdown quantifies NAND's share of a finished drive, and the source itself admits there was no identifiable threshold or cutover moment. The remaining claims are stated proportionately and hedged, so the overstatement is mild rather than systemic.
No incentive disclosures in supplied material
The supplied source contains no sponsorship, affiliate, vendor-relationship or funding disclosure, and no company in the story is shown to have supplied or reviewed the claims. There is no basis to score incentive pressure without inferring facts the material does not provide.
Moderate on history, weak on the forecast
Confidence is moderate: the historical and structural claims are specific, arithmetically coherent and easy to verify against public product history, and the industrial-niche and Windows-tooling details are concrete. It is held down by total dependence on one publisher, the absence of any cost model behind the causal thesis, and one unattributed forward projection that carries the story's main market implication.
build
A UDP packet is now enough: IKEEXT RCE moves from patch queue to fire drill1 distinct publisher
product
Nebius funds $4.5bn of AI capacity on terms that pay lenders mostly in stock2 distinct publishers
invest
Behind-the-meter gas is the data center buildout's real cost: 318 Mt a year1 distinct publisher
build
Grok 4.6 lands in Copilot two days after launch, and the model picker becomes a procurement problem1 distinct publisher
Distinct publishers with included, body-backed reporting in this cluster.
1 article · August 17, 2026