Loading today’s stories
Build1 distinct publisher3 min readPublished
HostingSift normalised 711 live plans from 39 hosts to dollars per gigabyte of RAM and found a 123x spread around a $6.00 median. That is only a buying signal if memory is the resource you run out of first.
The Engineer · Build desk
Compiled by The EngineerSomething wrong?How this is made
Dollars per gigabyte holds exactly one variable constant. Divide a plan price by its RAM figure and you have said nothing about vCPU share, disk type, disk size, bandwidth allowance, or how many other tenants sit on the box. The number transfers to your workload only when memory is the resource you exhaust first. A JVM heap you cannot shrink qualifies, as does a Postgres instance sized by shared buffers. A CPU-bound API tier does not, and for that workload the table sorts on an axis you are not constrained by.
The rules around the survey are doing more work than the ranking. HostingSift dropped Railway's Pro plan because its 1 TB figure is an account ceiling rather than provisioned memory [10]. It left Fly.io's storage and bandwidth out of the RAM price, since both bill on usage [11]. And it priced a 24-month Contabo term at its advertised per-month figure [9]. That third rule carries the top of the table: the $0.47 gigabyte is a two-year commitment wearing a monthly price tag [6].
The headline arithmetic holds. $58.00 divided by $0.47 is 123.4, which is the 123x [1]. The $6.00 median is 12.8 times the cheapest reading [2].
The fixed basket is where best-plan ratios stop being useful. SiteGround's Jump Start, $100 for 8 GB, works out at $12.50 per GB [4], the same figure as SiteGround's cloud entry point [12]. HostingSift calls that 20x Contabo's rate for the same memory [19]. Take the multiple literally and Contabo's 8 GB price sits near $5.00, or about $0.63 per GB, a third above its own headline rate [5]. DigitalOcean shows the effect from the other side: $48 for 8 GB is $6.00 per GB, the market median exactly, against a best-plan figure of $5.25 [6][12]. A provider's cheapest gigabyte usually lives on a rung you were not planning to buy.
That is the part I would act on. HostingSift's own explanation for the width of the spread is that buyers read the plan name and the monthly total [22], and it says its signup ranking falls apart in the renewal column [21]. Both failures have the same shape. The comparison gets made on the figure the seller chose to display.
So unit price earns a place as a screen rather than a verdict. Sort the market on dollars per gigabyte, keep three candidates, then re-price each one at the exact RAM size you will run, the term you can actually commit to, and the rate you will be paying in month thirteen. In my context the value of a 711-plan table is that it tells you which three are worth the re-pricing.
Ranked by verification strength, evidence, and original report placement.
Prices were converted to USD at the publisher's stored rate for the eight EUR-billed hosts in the set: Contabo, Hetzner, Netcup, OVHcloud, Private Hosting, Diploi, Evolution Host and Time4VPS.
Pricing is monthly-equivalent, so a 24-month Contabo term shows as its per-month figure, which is how the provider advertises it.
HostingSift priced 711 active VPS and cloud plans from 39 providers, converting EUR prices to USD and normalising everything to dollars per gigabyte of RAM per month, from nightly price readings taken on 30 and 31 August 2026.
The median price came out at exactly $6.00 per GB per month, with half the market above that line and half below.
A gigabyte of RAM on a virtual server costs $0.47 per month and also costs $58.00 per month, same unit and same billing period, both prices live on 31 August 2026.
Distinct publishers with included, body-backed reporting in this cluster.
dev.to
1 article · August 31, 2026
Follow any of these and your For You feed starts watching them — no settings page required.
build
752 Nigerian institutions, shipped as a repo instead of an endpoint1 distinct publisher
build
The First Firewall Rule Is a Cutover: One Allowlist Entry, One Dead Production App1 distinct publisher
build
OVHcloud reprices most of its catalog as its own RAM index hits 6x, with 9x due in September1 distinct publisher
product
The sandbox teams fled vm2 for now has its own guest-to-host escape1 distinct publisher
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.
Rules published, rows withheld
HostingSift discloses more than most price roundups: in-stock plans only, RAM figure required, VPS and cloud categories, and two named exceptions — Railway's 1 TB account ceiling and Fly.io's usage-based storage — where the exception cuts against the more sensational result. What it withholds is the part that would let anyone rerun it: the stored exchange rate applied to eight euro-billed hosts, and the 711 rows themselves. The arithmetic a reader can check does hold, and it is a small fraction of the arithmetic being asserted.
Offers, not purchases
Everything here describes what 39 hosts were willing to advertise on two nights in August. Nothing describes what buyers provisioned: no volumes, no market share, no sense of how many customers take the IONOS first-year rate and are still there when $11.00 becomes $44.00, and no indication that anyone moved because Time4VPS placed second. A price census is a supply snapshot, and reading demand into it would be invention.
The headline picks the two ends
The 123x figure is honest arithmetic between two rows that are not the same product, and the piece concedes it two-thirds of the way down — dedicated-core hosts price badly per gigabyte because memory is a side effect of selling you a core. The gap is one of billing, not accuracy: the extremes lead, the $6.00 median and the CPU, network and support caveats follow, and the behavioural explanation for why the spread persists arrives with no data at all. The dek's own qualifier does most of the correcting the headline needs.
The comparison site sells comparison
This is a hosting comparison business publishing the rankings that are its product, and the piece routes readers to its own true cost calculator and two of its own explainers along the way. Rankings that crown Contabo, discover an unknown Lithuanian host in second place and single out SiteGround as indefensible are exactly the output that draws hosting shoppers. No affiliate or commercial relationship with any ranked host is disclosed either way. The counterweight is real — the excluded Railway plan would have topped the chart — but the reader is still being asked to trust an unauditable scraper run by an interested party.
Plausible, unaudited, single-voiced
Nothing in the reporting looks fabricated — the checkable ratios reconcile, the method is stated, the caveats about oversubscribed cores and metered transfer are the ones a practitioner would raise. But the entire story rests on one interested publisher's private dataset from two nights in August, no provider was asked to confirm its own renewal price, and no second outlet priced the same market. That is enough to act on for screening, not enough to quote as market fact.