La formule de vitesse d'hébergement bon marché (2026)

Une licence cPanel Premier coûte USD à une société d'hébergement 69.99 un mois pour courir 100 comptes clients. Cela équivaut à USD 0.70 par site, par mois, avant l'électricité, Matériel, bande passante ou un seul ticket d'assistance. Now look at the plan you’re being offered for USD 1.99. The arithmetic resolves two ways: put more sites on the box, or charge you much more later. Most budget hosts do both.

Réponse rapide: Don’t guess what slow hosting costs you. Work out your break-even lift instead. Divide the monthly price difference between your cheap plan and a faster one by your site’s monthly revenue. If that number lands under 1%, upgrade today. For almost every site that earns money, ça fait. The number that should worry you isn’t speed, it’s the renewal multiplier. IONOS sells its bestselling plan at USD 1/mo, then charges USD 14/mo from month 13.

Dernière révision: septembre 2026. Prix ​​et fonctionnalités vérifiés.

How We Built This Cost Model

Every price below came off the provider’s own pricing page on 17 septembre 2026, read in a browser where the page injected prices with JavaScript. Nothing here is quoted from a review aggregator or a coupon site, because those carry stale numbers and we’d have no way to date them.

Three exclusion rules shaped what made the cut. Première, we dropped any figure a provider didn’t publish itself. Verpex shows renewal prices in its plan table but no promotional rate. So we report the renewal and leave the discount blank rather than guess. Deuxième, we ignored plans where the advertised price requires more than a 48-month prepayment. La troisième, the conversion math draws on one study only: the Google-commissioned research by Deloitte and the agency 55. It’s the one large dataset that names server latency among the metrics it measured.

The weighting reflects the question this article asks. Entry-to-renewal ratio matters most, followed by what the cheap tier withholds (serveur Web, CPU share, CDN), then published resource caps. Uptime badges got no weight at all. They’re self-reported and unfalsifiable.

Ce que nous n'avons pas fait: we ran no synthetic load tests and measured no TTFB on live customer accounts. We also have no way to see how many accounts a host puts on one machine. Our speed claims are limited to what providers publish about their own tiers. Where a provider declined to state a renewal price anywhere on its site, we say so instead of estimating.

La formule de vitesse d'hébergement bon marché

Most articles on this topic try to predict your losses. They’ll take a conversion statistic, multiply it by your traffic, and hand you a scary number. That approach is broken, because the underlying research models speed logarithmically. You can’t multiply a 0.1-second result by eight and call it a 0.8-second result.

So flip the question. Instead of predicting the gain, work out how small a gain would justify the spend. Three steps:

• Étape 1. Monthly revenue through the site (R) = monthly visitors x conversion rate x value per conversion.
• Étape 2. Monthly upgrade cost (U) = renewal price of the faster plan minus renewal price of your current plan. Use renewal prices. Promotional prices will lie to you here.
• Étape 3. Break-even lift (B) = U divided by R, as a percentage.

B is the conversion improvement the faster plan has to deliver just to pay for itself. Now you have something to judge. The Deloitte study measured an 8.4% conversion lift on retail mobile from a 0.1-second improvement. That’s the optimistic ceiling, from one vertical, under conditions we’ll pick apart below. Read your own B against it:

• B under 1%: upgrade now. The faster plan pays for itself if it moves conversions by a rounding error.
• B between 1% et 8%: check your TTFB first. Worth doing only if your server response time is genuinely poor.
• B above 8%: hosting isn’t your bottleneck. Spend the money on anything else.

Here’s what surprised us when we ran real numbers. B almost never lands above 1%.

Take a consultancy site: 8,000 visiteurs par mois, 2% send an enquiry, 12% of enquiries close at USD 900. Value per enquiry is USD 108, so R is USD 17,280. Moving from HostArmada’s Start Dock to Speed Reaper costs USD 9.80 plus par mois au renouvellement, and it’s the jump from NGINX to LiteSpeed. B works out at 0.057%. Pas 5%. Zero point zero five seven.

Or a small WooCommerce store: 12,000 visiteurs, 1.1% conversion, USD 48 commande moyenne. R is USD 6,336. Going from IONOS Essential to IONOS Plus at renewal costs USD 6 more a month and buys five times the server resources plus a CDN. B is 0.095%.

Now a local brochure site: 900 visiteurs, 3% appel, USD 40 par appel. R is USD 1,080. That same USD 9.80 upgrade gives a B of 0.91%. Still under 1%, but only just, and that’s a site earning USD 1,080 un mois.

The pattern is blunt. If your site earns anything at all, the upgrade is too cheap to agonize over. The whole difference is USD 72 en USD 118 une année. And if your site earns nothing, then slow hosting costs you nothing either, and you should stop reading advice that says otherwise.

Where a Cheap Server Actually Loses Time

“Overselling” gets blamed for everything, and it’s unverifiable from the outside. You can’t count how many accounts share your machine. What you can do is read what the cheap tier openly withholds, because providers publish this themselves.

The web server is often the upgrade itself. HostArmada exécute NGINX sur Start Dock et Web Warp. LiteSpeed, the web server most budget hosts market as their speed advantage, only appears on Speed Reaper. So a buyer picking HostArmada for its speed reputation, then choosing the cheapest plan to save money, doesn’t get the thing they came for. That’s the trap in miniature.

IONOS is blunter still, and it’s the most useful pricing page in budget hosting for exactly that reason. Its plans are graded by “access to server resources”: Essential gets Standard, Starter gets Extended, Plus gets 5x, Ultimate gets 10x. That’s a provider telling you in plain text that the cheap plan gets a smaller slice of the CPU. IONOS also gates its CDN. “Performance boost with CDN” appears on Starter and up, never on Essential. Buy the cheapest IONOS plan and your visitors reach one origin server, wherever it happens to be.

Published caps tell the rest. Verpex lists its CloudLinux limits per plan. Bronze gets 1 Go de RAM, 1 processeur virtuel et 30 processus d'entrée (how many requests it serves at once). Silver and Gold double the CPU to 2 vCPU. But look at the disk figures, because they don’t move: all three tiers cap at 1024 IOPS et 50 Mo / s. Paying 2.5x more takes you from USD 6/mo to USD 15/mo and buys zero extra disk throughput. For a database-heavy WooCommerce store, that’s the limit that’ll actually bite.

Ultahost puts a number on it directly. Its Shared Starter plan is sold for roughly 10,000 visites mensuelles. That’s not a throttle, it’s a sizing guide, and it’s more honest than the “illimité” language most of the category uses.

The threshold to measure all this against is Google’s, not a host’s. Google’s web.dev guidance was last updated on 18 novembre 2025. It puts a good Time to First Byte (TTFB, the delay before your server sends its first byte) à 0.8 secondes ou moins. Anything above 1.8 seconds counts as poor. TTFB isn’t itself a Core Web Vitals metric. It just sits underneath Largest Contentful Paint, so a slow server caps how good your LCP can ever be. Test your own before you spend anything.

The Renewal Multiplier: The Cost That Isn’t Speed

This is where cheap hosting genuinely costs more, and it has nothing to do with milliseconds. We calculated the entry-to-renewal multiplier for every plan we verified. Divide the renewal price by the promotional price and the category splits into three groups.

• Plat. Ultahost Shared Starter: USD 3.44/mo promotional, USD 3.44/mo on renewal. A multiplier of 1.0.
• Modérer. IONOS Starter at 1.7x (USD 6 en USD 10) and Ultimate at 1.8x. IONOS Essential and Stablepoint’s Go Web both sit at 2.0x. InterServer reaches 2.8x (USD 2.50 premier mois, puis USD 7).
• Raide. Hostinger Cloud Startup at 3.3x (USD 7.99 en USD 25.99). Hostinger Premium at 3.7x (USD 2.99 en USD 10.99). Hostinger Unlimited at 4.3x (USD 3.99 en USD 16.99).
• Extrême. Every HostArmada shared plan at exactly 5.0x, and IONOS Plus at 14.0x.

Two of those deserve a closer look. HostArmada’s 5.0x isn’t approximate. Start Dock goes USD 1.99 en USD 9.95, Web WarpUSD 3.29 en USD 16.45, Vitesse Reaper USD 3.95 en USD 19.75. Trois plans, one flat 80% launch discount applied across the board. Once you see it, their pricing page reads differently: the promotional price carries no information about the product, only about the discount.

IONOS Plus is the one to watch, because IONOS labels it “Bestselling Plan”. It’s USD 1/mo for a 1-year term, puis 14 USD/mois. Run that over three years and it costs USD 348. Ultahost’s flat USD 3.44 plan costs USD 123.84 au cours des mêmes trois années. The plan advertised at USD 1 finit 2.8x plus cher than the plan advertised at USD 3.44. That comparison is invisible at checkout, quel est le point.

There’s a reason the whole category drifted this way, and it’s not greed alone. cPanel raised license prices on 1 janvier 2026, taking Premier to USD 69.99 un mois pour 100 accounts plus USD 0.49 per extra account. Every cPanel-based budget plan sold today therefore carries a higher cost floor than the same plan did in 2025. Réglez cet USD 0.70 per account against a USD 1.99 promotional price and the panel license alone eats about 35% de revenus. Contre un USD 9.95 renouvellement, c'est 7%. The renewal isn’t a penalty. It’s the price at which the business works.

Stablepoint published its own version of this on a pricing update page. Go Web moved from GBP 1.99 en EUR 3.98 un mois, Medium Web from GBP 7.49 en EUR 13.98. Both apply “from your next renewal”. Most hosts change prices quietly. Publishing the old and new figures side by side is rare, and it’s worth more than an uptime badge.

When Cheap Hosting Is the Correct Choice

The honest answer most articles in this category skip: souvent.

If your site doesn’t convert visitors into money, the entire formula collapses to zero. R is zero, so B is undefined, and no speed upgrade can pay for itself. Un blog personnel, un portefeuille, a club site, a project you’re learning on. Buy the cheapest plan that runs current PHP and gives you a free SSL certificate, and spend the saved money on something that matters. Anyone telling a hobbyist that cheap hosting is costing them business is selling something.

Cheap also stops meaning slow once you read the spec instead of the price. Ultahost holds USD 3.44/mo flat with 30 Go de stockage NVMe. HostArmada’s Speed Reaper carries LiteSpeed at USD 3.95 promotionnel. Verpex puts 30 GB of NVMe behind a published USD 6/mo renewal. Price and performance are correlated, but they’re separate axes. NVMe storage and LiteSpeed now appear well below USD 5/mo. Our wider roundup of les meilleurs plans d'hébergement mutualisé bon marché goes deeper on which budget tiers hold up.

Where cheap hosting does break down is concurrency, pas de vitesse brute. One visitor on a 1 vCPU plan is fine. Forty at once on a plan capped at 30 entry processes is not. That’s the signal to look at Hébergement VPS, and it’s a traffic decision rather than a budget one.

What the Speed Studies Actually Say

Three claims circulate constantly in this category. Two are wrong and one is misreported.

Myth: site-level Core Web Vitals. A cluster of SEO blogs spent this year describing a March 2026 mise à jour. It supposedly aggregates Core Web Vitals across your whole domain, then adds a new signal called Engagement Reliability. None of it appears in Google’s documentation. Google’s own Core Web Vitals page was last updated on 10 décembre 2025. It still lists the same three thresholds: LCP sous 2.5 secondes, INP sous 200 millisecondes, CLS sous 0.1. There is no Engagement Reliability metric. Don’t re-plan your hosting around a signal nobody at Google has described.

Misreported: 0.1s equals 8.4%. This one’s real research, badly quoted. The study was commissioned by Google and run by Deloitte and the agency 55, revêtement 37 brands and 30 million user sessions over four weeks. The retail figure of +8.4% conversion and +9.2% average order value came from 20.5 million of those sessions. But the 0.1-second improvement wasn’t server time. It was a combined 0.1-second improvement across four metrics at once: Max Server Latency, First Meaningful Paint, Estimated Input Latency and Observed Load. Server response was one of four inputs, not the whole cause. The model was also logarithmic and built on naturally occurring speed variation, not a controlled test. So treat 8.4% as a ceiling for retail mobile, and never as a per-0.1s rate you can multiply.

Folklore: bad server neighbors. The shared-IP spam theory gets repeated endlessly, without Google ever confirming it as a ranking mechanism. What is confirmed is far more boring: your server response time feeds LCP, and LCP is one of three Core Web Vitals. A cheap host that returns bytes in 400 ms poses no ranking problem. An expensive host that takes 2 seconds does.

How to Decide What to Spend

Run the formula, then match your result to one of these.

Revenue under USD 100/mo. Stay on the cheapest plan that runs PHP 8.3 and includes SSL. Your B will be enormous, which is the formula telling you the truth. Ultahost’s flat USD 3.44 is the pick here purely because it removes the renewal decision. Skip anything at a 5x multiplier: HostArmada’s Start Dock at USD 1.99 looks cheaper today and costs USD 9.95 from year four.

Revenue USD 1,000 to 10,000/mo. Budget USD 15 to USD 25/mo at renewal and stop optimizing the hosting line. IONOS Plus at USD 14 renewal buys 5x server resources and a CDN. HostArmada Speed Reaper at USD 19.75 renewal buys LiteSpeed. At the top of this bracket, break-even lands under 0.1%. At the bottom it’s nearer 1%, which is still a yes. Don’t buy IONOS Plus for the USD 1 titre, bien que. Buy it knowing year two is USD 14.

WooCommerce, 500+ commandes mensuelles. Disk throughput is your constraint, pas de processeur. Verpex caps every shared tier at 50 MB/s regardless of price. Paying up from Bronze to Gold buys storage and sites, not database speed. This is the point where shared hosting stops being the right shape and dedicated eCommerce hosting gagne son prix.

Au-dessus de 50,000 visites mensuelles. Ultahost sizes its Starter tier at around 10,000 visites mensuelles. You’re 5x past that. Entry-process caps will queue your visitors during peaks even while your average response time looks fine. Move to a VPS or managed platform before you buy another year of shared.

Not sure which bracket you’re in? Measure TTFB first, on a Tuesday afternoon and again on a Sunday evening. Cheap shared hosting degrades under neighbor load, so a single reading tells you nothing. If both readings sit under 800 SP, your host is not your problem. If both readings sit above 1.8 secondes, the fix is a host with LiteSpeed and a CDN, not another caching plugin.

Questions fréquemment posées

Is cheap hosting bad for SEO?

Not by price. Google ranks on measured performance, and the threshold that matters is a Largest Contentful Paint under 2.5 secondes. Your server controls only the first slice of that, so a fast USD 3.44 Ultahost plan leaves you plenty of room to spend elsewhere. Un dollar 20 plan that takes two seconds has already eaten most of the budget. What cheap hosting does risk is inconsistency: response times that are fine at 3am and poor at peak. Test twice at different times before you blame or clear your host.

How much traffic can a USD 3 shared hosting plan actually handle?

Ultahost sizes its USD 3.44 Shared Starter at roughly 10,000 visites par mois, which is the clearest published figure in the budget tier. The real limit isn’t monthly traffic though, it’s simultaneous visitors. Verpex caps Bronze at 30 processus d'entrée et 1 vCPU, so a burst of concurrent requests queues long before you hit any monthly total. A steady 10,000 c'est bien. Le même 10,000 arriving after one newsletter send is not.

Is it cheaper to buy three years of cheap hosting or one year of better hosting?

Depends entirely on the multiplier, and the answer is often the opposite of what the sticker suggests. IONOS Plus at USD 1/mo costs USD 348 sur trois ans, because months 13 à 36 bill at USD 14. Ultahost en USD 3.44 flat costs USD 123.84 sur la même période. The plan that looks 3.4x cheaper is 2.8x more expensive. Always multiply the renewal price by the months you’ll really stay.

Will moving to a faster host improve my Google rankings?

Only if your current server is the thing holding back your Largest Contentful Paint. If your TTFB is already under 800 SP, moving hosts changes nothing, and your LCP problem is images, fonts or render-blocking scripts. If TTFB is above 1.8 secondes, which Google classes as poor, a LiteSpeed plan like HostArmada’s Speed Reaper can fix it in a weekend. Measure before you migrate, because migration risk is real and ranking gains are not guaranteed.

Verdict final

The framing this topic usually gets is wrong. Cheap hosting is not expensive because it’s slow. For any site earning real money, the speed upgrade is trivial next to revenue. The break-even lift lands under 1%, so just buy it. The consultancy example above needed a 0.057% conversion improvement to justify a better plan. That’s not a decision, it’s a formality.

Cheap hosting is expensive because of arithmetic you don’t see at checkout. HostArmada applies a flat 5.0x renewal across every shared plan. IONIQUE sells its self-declared bestseller at USD 1 and renews it at USD 14. That 14.0x step turns the cheapest-looking plan into a USD 348 engagement sur trois ans. Hostinger sits between 3.3x and 4.3x. Contre ça, Ultahost détient des dollars américains 3.44 flat and Verpex prints its renewal price in the plan table. Both do something the rest of the category avoids.

So use the formula twice. Once to decide whether speed is worth paying for, where the answer is almost always yes and the amount is almost always trivial. Then again on the renewal price, where the real money is. Multiply the renewal by the months you’ll actually stay, not the months of the discount. Do that and the cheap hosting problem mostly solves itself.

For the buying side, notre guide pour the cheapest hosting in the USA carries current entry and renewal figures for every provider named here. Outgrown shared hosting? Sociétés d'hébergement cloud and VPS providers are the next step up. Both change the concurrency math, not just the price.