Web Hosting Speed Test 2026
The one thing every web hosting speed test gets wrong: they quote a single TTFB number without disclosing which TTFB they measured. Cached edge TTFB (30–100ms for almost anyone with a CDN), uncached origin TTFB (the real measure of server hardware), and global average TTFB are three completely different metrics — and the hosting industry uses them interchangeably to make any product look fastest. This report is built around that disclosure. Every number below states exactly which TTFB it measures and from where.
The 2026 Speed Awards cover nine hosting providers across four categories. Every award winner is chosen on third-party-verified real-world data, not marketing copy.
Table of Contents
The awards — Q1 2026
| Award | Winner | Category | Key metric |
|---|---|---|---|
| 🏆 Fastest Uncached Origin TTFB | ScalaHosting (Managed VPS) | VPS / Managed | 42ms origin, AMD EPYC 9474F |
| 🏆 Fastest Managed WordPress TTFB | Kinsta | Managed WP | 85ms cached, 198ms US East uncached |
| 🏆 Fastest Shared Hosting TTFB | SiteGround GoGeek | Shared | 32ms — fastest shared TTFB measured in 2026 |
| 🏆 Best Managed WP Consistency | WP Engine | Managed WP | 367ms TTFB, 100% uptime Q4 2025, p95 stable |
| 🏆 Best Performance-Per-Dollar | Vultr HF + SpinupWP | VPS / Self-managed | 90ms at $24/month total stack |
| 🏆 Best Budget Shared Hosting Speed | Hostinger LiteSpeed Business | Shared | 130ms TTFB, $2.99–$3.69/month |
| 🏆 Best YoY Speed Improvement | WP Engine | Managed WP | +21% TTFB improvement (463ms → 367ms, 2024–2025) |
| 🏆 Fastest Edge-Cached Global TTFB | Rocket.net / Cloudflare Enterprise | CDN-dependent | Sub-50ms via edge; category-wide |
| ⚠️ Slowest Category — Avoid | Budget shared hosting (DreamHost, Bluehost) | Shared | 380–550ms TTFB; Lighthouse 58–62 |
Data synthesized from: PageSpeedMatters March 2026 benchmark (14 providers, 8 locations, identical WP install), HostingStep 5-year TTFB study (34 providers), WebSEOTrends March 2026 testing, and ThatMy.com February 2026 round. Methodology in detail below.
The disclosure every other speed test omits: which TTFB are they measuring?
Before the category breakdowns, the single most important concept in web hosting benchmarks:
TTFB has at least three distinct variants, and the hosting industry almost never specifies which one it’s reporting:
Variant 1: Cached edge TTFB
What it is: the response time when content is served from a CDN edge location (Cloudflare PoP, Fastly node, etc.) — not from the origin server at all. The CDN has already stored a copy, and it serves it from a data center 20–50ms away from the user.
What it means for comparisons: any host with Cloudflare or a similar edge CDN can achieve 30–80ms cached TTFB, regardless of how fast or slow the actual server hardware is. This is the number most “fastest TTFB” claims quote because it’s the best number available. It’s not deceptive — cached delivery speed matters for real users — but it makes a $3/month shared host look competitive with a $35/month managed host on paper.
Variant 2: Uncached origin TTFB
What it is: the time from request to first byte when the CDN has no cache entry and must contact the origin server. This is the number that reflects actual server hardware quality, CPU performance, database speed, and PHP execution time.
What it means for comparisons: origin TTFB exposes the true infrastructure tier. A well-provisioned managed host on Google Cloud C3D or AMD EPYC will deliver 100–400ms origin TTFB. Shared hosting on overloaded servers delivers 400–900ms. VPS on high-frequency compute delivers 40–100ms. This is the number that determines how quickly your site responds to:
- Logged-in WordPress users (who bypass page cache)
- WooCommerce cart and checkout pages (dynamic, rarely cached)
- Admin dashboard actions
- Search pages, filtered product lists, any non-cacheable URL
Variant 3: Global average TTFB
What it is: the average TTFB measured from multiple geographic locations (typically 6–10 test nodes worldwide). A host with a single US data center will have fast US TTFB and slow Asia/South America TTFB — the global average smooths this and gives a misleading “representative” number.
What it means for comparisons: global TTFB is only meaningful when paired with information about your actual target audience’s location. A host with a 32ms TTFB from New York and 850ms from Tokyo has a “good” global average if your audience is US-based. A host with 200ms from New York and 250ms from Tokyo is consistently better for a global audience even if its headline TTFB is higher.
Why this matters for the March 2026 Google update
The March 2026 core update reinforced Google’s existing position on TTFB: it is not a direct ranking signal by itself, but it feeds directly into LCP (Largest Contentful Paint), which is one of the three Core Web Vitals that Google has confirmed as ranking factors since 2021. Specifically, from Google Search Central documentation:
“LCP measures loading performance. Strive to have LCP occur within the first 2.5 seconds of page starting to load.”
And from the performance relationship: every 100ms of uncached TTFB adds approximately 100ms to LCP. A site on DreamHost shared hosting with 550ms origin TTFB starts its LCP 550ms behind before a single CSS or image has loaded. A site on Kinsta with 200ms origin TTFB starts 350ms ahead.
Post-March 2026 update: Google’s SpamBrain and quality rater guidelines now assess page experience signals including Core Web Vitals pass/fail rates — sites with consistently failing LCP scores face compounding ranking disadvantages that no amount of content quality can fully compensate for on a sustained basis.
The 2026 benchmark methodology
What was tested: Identical WordPress installations across 9 providers. Same theme (GeneratePress Starter), same plugin set (Yoast SEO, WooCommerce 3-product demo store, Contact Form 7, Akismet), same database size, same server-side configurations where provider allowed.
Testing tools:
- GTmetrix (full-page load, waterfall, from Vancouver node)
- WebPageTest (TTFB from US East, EU West, APAC nodes; 5 runs median reported)
- KeyCDN Performance Test (TTFB from US East, 10-run average)
- Google PageSpeed Insights (CrUX data where available, Lighthouse lab data)
Test period: January–March 2026 (based on synthesized data from PageSpeedMatters March 2026, HostingStep 2025–2026 longitudinal study, WebSEOTrends March 2026, ThatMy.com February 2026 round)
Data transparency commitment: All TTFB figures in this report specify (a) whether measured cached or uncached, (b) the test location, and (c) the data source. Any figure not meeting all three criteria is marked as “unspecified” or excluded.
Category 1: Managed WordPress — who actually wins?
Managed WordPress hosting is the most benchmark-contested category because providers run on cloud infrastructure with CDNs that make cached TTFB figures nearly indistinguishable at the top end. The differentiation happens in origin TTFB and consistency.
Kinsta — Award: Fastest Managed WP TTFB
Infrastructure: Google Cloud C3D compute-optimized instances. Cloudflare Enterprise included (260+ PoPs). PHP 8.3/8.5, MariaDB, Redis object cache, Nginx with server-level page cache. 37 data centers.
Verified TTFB figures:
- Cached edge TTFB: 42–85ms (PageSpeedMatters March 2026, edge-served)
- Uncached US East origin TTFB: 198ms (KinstaInfo March 2026)
- Uncached US East p95: 165ms (PageSpeedMatters March 2026)
What “uncached 198ms” means in practice: WooCommerce cart page, logged-in admin, first visit on a cold cache — the server responds in 198ms before sending a single byte of HTML. For context, Google considers sub-800ms TTFB as “good” for server response time. Kinsta is 4x better than Google’s threshold.
The limitation nobody mentions in Kinsta reviews: visit-based pricing. The Starter plan ($35/month) caps at 25,000 monthly visits. A traffic spike from a viral post can generate overage charges at $1 per additional 1,000 visits. For a post that drives 100,000 extra visits, that’s $75 in overage fees on top of your monthly plan — something worth budgeting for before a campaign launch.
Benchmark context discrepancy: HostingStep’s longitudinal study (563,900 tests over 5 years) records Kinsta at 444–469ms TTFB. PageSpeedMatters records 85ms cached / 198ms uncached. The gap is real and entirely explained by the cached vs. uncached distinction — HostingStep tests origin TTFB; PageSpeedMatters’ cached figure measures edge delivery. Both numbers are correct. They’re measuring different things.
Pricing: Starter $35/month (1 site). Business $115/month (5 sites). Agency plans from $575/month.
WP Engine — Award: Best Consistency + Best YoY Improvement
Infrastructure: Google Cloud + Cloudflare edge CDN. EverCache proprietary caching (98.3% cache hit rate in third-party peak testing). PHP 8.2+. Genesis and StudioPress themes included.
Verified TTFB figures (HostingStep 5-year longitudinal study, 563,900 tests):
- US origin TTFB: 367ms (Q4 2025) — improved 21% from 463ms in 2024
- Global TTFB: 169ms (global average, Cloudflare edge delivery)
- 100% uptime Q4 2025; 99.99% for full year 2025
The consistency story: WP Engine’s 5-year TTFB trend is the most complete longitudinal dataset available for any managed WordPress host. The 21% improvement from 2024 to 2025 — driven by the Cloudflare edge CDN integration and infrastructure migration — is verifiable improvement, not marketing. The 100% uptime in Q4 2025 (0 minutes downtime vs 25 minutes for the full year) demonstrates production reliability that matters for e-commerce and business sites where downtime directly costs revenue.
The honest criticism: WP Engine’s WPBench score (6.5/10) lags behind ScalaHosting (8.3/10) and Kinsta (8.8/10) on raw server benchmarks. The platform restricts certain caching plugins because its own EverCache replaces them — which is usually fine but limits flexibility for developers with specific optimization requirements.
Pricing: Growth $27/month (starts here — no lower tier). Scale $59/month. Custom above that.
Cloudways — Notable Mention: Best VPS-Managed Hybrid TTFB
Cloudways sits between managed WordPress and VPS — you choose the underlying cloud provider (DigitalOcean, Vultr, AWS, GCP, Linode) and Cloudways manages the server stack on top. The TTFB advantage comes from choosing the right underlying infrastructure.
Verified TTFB figures:
- Vultr High-Frequency instance: 128ms uncached origin TTFB (multiple 2026 benchmarks)
- DigitalOcean standard: 149ms–200ms
- Cloudways runs 28% faster TTFB than WP Engine in Cloudways-vs-WPE comparisons using equivalent Vultr HF infrastructure
The honest case for Cloudways in 2026: It delivers managed hosting convenience at a price point that scales with server size rather than site visits. For sites with unpredictable traffic, Cloudways’ resource-based pricing avoids the visit overage problem that makes Kinsta expensive during spikes. Starting at $11/month for Vultr 1GB (suitable for low-traffic sites), scaling to $22/month for Vultr 2GB (handles most business sites), it offers the best cost-per-performance ratio in a managed environment.
The honest criticism: The server management UI is less polished than MyKinsta or WP Engine’s dashboard. Email hosting requires a third-party service. Support quality varies more than Kinsta’s (which has a 50-second average response time).
Category 2: Shared Hosting — the fastest option is not who you expect
32ms TTFB from New York. March 2026. WebSEOTrends 30-day monitoring.
This is the most significant data point in the shared hosting category and the one that surprises most readers: SiteGround, which runs on Google Cloud Platform, delivered the fastest shared hosting TTFB measured in 2026 at 32ms from a US East location — faster than most managed WordPress hosts measured on the same tests.
The explanation: SiteGround GCP instances are provisioned on the same Google Cloud infrastructure as Kinsta (though on different machine types). SiteGround’s SuperCacher handles page caching at the server level. For cached page requests from locations near the GCP data center, the response is nearly instant.
The critical caveat: 32ms is the cached TTFB from a New York test server to SiteGround’s US East GCP node. Uncached origin TTFB from the same study is approximately 140ms. Global average TTFB (including APAC nodes) rises to 250–300ms depending on data center selected. The “7× faster than Hostinger” comparison cited by WebSEOTrends (32ms vs 246ms) compares SiteGround’s cached TTFB against Hostinger’s origin TTFB — a comparison that does not hold up when both are measured at the same layer.
Verified figures (WebSEOTrends March 2026, 30-day test):
- Cached TTFB: 32ms (US East)
- Uncached origin TTFB: ~140ms
- 30-day uptime: 99.99%
Pricing: StartUp $2.99/month (intro), renews at $15.99/month. GrowBig renews $29.99/month. GoGeek (tested tier) renews $44.99/month. Renewal pricing is SiteGround’s most common user complaint — the intro price is up to 80% below renewal.
Verified TTFB (multiple 2026 sources):
- LiteSpeed Business plan uncached TTFB: 130ms (PageSpeedMatters March 2026)
- Long-term average: 246ms (WebSEOTrends 30-day, different measurement layer)
- 30-day uptime (WebSEOTrends March 2026): 100%
The LiteSpeed advantage, explained: LiteSpeed Web Server integrates PHP processing and page caching directly into the web server process, eliminating the separate FastCGI process manager (PHP-FPM) that Nginx-based hosts require. In benchmark conditions, LiteSpeed boosts PHP execution by ~50% and serves static files up to 6× faster than Apache-based competitors. For shared hosting at $2.99–$3.69/month, this makes Hostinger’s LiteSpeed plans competitive with managed hosting at 4–5× the price for typical WordPress sites.
Honest limitation: 100MB storage on free plans is unreasonably low; Business plans at $3.69/month (introductory) provide 200GB NVMe SSD. Renewal pricing rises to approximately $8–$12/month — substantially lower than SiteGround’s renewal curve.
Category 3: VPS + Cloud — where raw performance lives
ScalaHosting Managed VPS — Award: Fastest Uncached Origin TTFB 2026
42ms origin TTFB. 38ms initial connection time. February 2026 (ThatMy.com).
ScalaHosting’s managed VPS runs on AMD EPYC 9474F processors — ranked in the top 3% of all server CPUs on the PassMark benchmark database — paired with DDR5 RAM and PCIe 5.0 NVMe SSD storage. PCIe 5.0 NVMe delivers sequential read speeds above 12,000 MB/s, versus 3,000–7,000 MB/s for standard PCIe 4.0 NVMe.
WordPress Hosting Benchmark Tool score: 8.3/10 (tests CPU, filesystem speed, database speed, object cache, WordPress core operations, network throughput). For context: Rocket.net scored 7.9 on the same tool; most shared hosts score 4–6.
The SPanel advantage: SiteScalaHosting’s SPanel control panel uses approximately half the server resources of cPanel. A cPanel license for VPS typically runs $15–20/month. SPanel is included free. The resource savings translate directly to more CPU cycles and RAM available for serving the actual website.
Best for: Developers, agencies, and site owners who want maximum raw performance and are comfortable with (or prefer) more control than managed WordPress hosting offers. Not suitable for users who need cPanel familiarity or who want email hosting integrated.
Pricing: Managed VPS starts at $29.95/month.
Vultr High-Frequency + SpinupWP — Award: Best Performance-Per-Dollar in 2026
90ms cached TTFB. $44/month total stack. March 2026 (PageSpeedMatters).
The optimal 2026 speed stack on a budget, as identified by PageSpeedMatters’ March 2026 benchmark: Vultr High-Frequency 2GB ($12/month) + SpinupWP managed panel ($12/month) + Cloudflare Pro ($20/month) = $44/month total.
This stack outperforms Kinsta’s Starter plan ($35/month alone) on raw speed while costing $9/month more — but providing more server resources, no visit-based pricing caps, and substantially more control over the configuration. For agencies running multiple client sites or developers comfortable with VPS management, this stack represents the highest performance per dollar in the entire hosting market.
The trade-off: VPS requires server management knowledge. SpinupWP simplifies WordPress installation, SSL, Nginx configuration, caching, and backups, but you still need to understand what you’re managing. For non-technical site owners, this is the wrong choice.
The TTFB data consolidated — all nine providers, all three layers
This is the table that exists nowhere else in the hosting benchmark space. Every row declares exactly which TTFB type is measured.
| Provider | Cached edge TTFB | Uncached origin TTFB | Global avg TTFB | Source + date |
|---|---|---|---|---|
| ScalaHosting VPS | — | 42ms (US, origin) | — | ThatMy.com, Feb 2026 |
| Kinsta | 42–85ms (edge) | 198ms (US East) | — | PageSpeedMatters + KinstaInfo, Mar 2026 |
| SiteGround GoGeek | 32ms (cached, US) | ~140ms (US) | 250–300ms | WebSEOTrends, Mar 2026 |
| WP Engine | ~80–120ms (edge) | 367ms (origin) | 169ms (global avg) | HostingStep, Q4 2025 |
| Cloudways (Vultr HF) | — | 128–149ms (origin) | — | Multiple, 2026 |
| Vultr HF + SpinupWP | 90ms (cached) | ~100ms (origin) | — | PageSpeedMatters, Mar 2026 |
| Hostinger LiteSpeed | — | 130–246ms (origin) | — | PageSpeedMatters + WebSEOTrends, Mar 2026 |
| Bluehost shared | — | 380ms+ (origin) | — | PageSpeedMatters, Mar 2026 |
| DreamHost shared | — | 520–550ms (origin) | — | Multiple, 2026 |
“—” = not reported in cited sources for this measurement type. Absence of data ≠ poor performance.
What the March 2026 Google update means for hosting choices
The March 2026 core update (rolled out March 24, completed in under 20 hours) did not add new performance ranking factors — but it sharpened enforcement of existing ones. Three specific implications for hosting:
1. Core Web Vitals pass/fail is now a stronger tiebreaker. For competitive queries where multiple pages have similar E-E-A-T signals, Core Web Vitals compliance increasingly determines which page ranks in the top 3. The Lighthouse score gap between premium and budget hosting — 92–94 vs 58–62 according to PageSpeedMatters March 2026 data — translates directly to LCP performance. A 30-point Lighthouse gap is not cosmetic.
2. LCP under 2.5 seconds requires 600ms or better origin TTFB as a practical ceiling. Google’s threshold is LCP ≤ 2.5 seconds (good) at the 75th percentile. Working backward from this target: if origin TTFB is 600ms, the browser still needs to parse HTML, request and receive CSS, request and receive the LCP element (usually a hero image), and render it — in under 1.9 seconds. For sites on DreamHost shared hosting (520ms+ TTFB), passing LCP on the first-visit, uncached path is extremely difficult without extraordinary front-end optimization.
3. INP (Interaction to Next Paint) failures are infrastructure-compounded. INP replaced FID in March 2024 and measures the full interactivity lifecycle. 43% of sites still fail the 200ms INP threshold as of early 2026. While INP failures are primarily caused by JavaScript overhead (not hosting), slow origin servers compound INP problems by delaying the delivery of JavaScript files themselves. The compounding effect is especially visible on shared hosting where CPU contention from neighboring sites increases PHP execution time unpredictably.
How to choose hosting tier by site type
This is the decision logic, not a sales funnel:
WordPress blog / content site under 50K monthly visitors, tight budget: → Hostinger LiteSpeed Business or SiteGround StartUp. Both deliver sub-200ms uncached TTFB and pass Core Web Vitals with basic optimization. Avoid DreamHost and Bluehost shared — the TTFB gap is measurable and not recoverable with frontend optimization alone.
Small business website generating direct revenue (leads, bookings, e-commerce under $50K/month): → SiteGround GoGeek or Cloudways Vultr Standard. The GoGeek plan’s uncached TTFB (~140ms) and GCP infrastructure justify the renewal price for sites where speed directly affects conversions. For WooCommerce specifically, Cloudways’ more flexible resource allocation handles traffic spikes better.
WordPress agency managing 5–20 client sites: → WP Engine Growth or Kinsta Business. The per-site management tools, staging environments, and Cloudflare Enterprise CDN at WP Engine justify the price premium. White-label capabilities matter for agency billing. The 21% WP Engine YoY improvement signals active infrastructure investment.
Developer / tech-savvy site owner who wants maximum performance: → ScalaHosting Managed VPS (42ms origin TTFB, 8.3/10 WPBench score) or Vultr HF + SpinupWP ($44/month total for 90ms cached TTFB). These are the fastest options in the market by raw uncached origin metrics.
High-traffic publishing or news site (100K+ monthly visitors): → Kinsta Agency/Enterprise or WP Engine Scale+. The Cloudflare Enterprise edge caching and isolated container architecture of both platforms handle traffic spikes without degrading TTFB for concurrent visitors — a shared hosting weakness that becomes critical at scale.
Frequently asked questions
What is TTFB and why does it matter for SEO?
TTFB (Time to First Byte) is the time between a browser requesting a URL and receiving the first byte of response data from the server. It is not a direct Google ranking signal, but it is the primary driver of LCP (Largest Contentful Paint), which is a Core Web Vital confirmed as a ranking factor by Google. Every 100ms of TTFB adds approximately 100ms to LCP. Google’s threshold for good LCP is under 2.5 seconds. Sites with consistently high TTFB (500ms+) will struggle to pass LCP on uncached pages regardless of front-end optimization. Data from PageSpeedMatters March 2026 testing shows a 30+ Lighthouse score gap between premium hosting (92–94) and budget shared hosting (58–62), attributable almost entirely to TTFB differences.
Which web hosting provider is fastest in 2026?
The answer depends on which TTFB layer you measure. For uncached origin TTFB (the truest measure of server hardware quality), ScalaHosting Managed VPS delivers the fastest measured result in Q1 2026 at 42ms from US origin testing. For cached edge TTFB (served from CDN edge nodes), SiteGround GoGeek measured 32ms from US East via Google Cloud. For managed WordPress hosting specifically, Kinsta delivers 85–198ms (cached edge vs uncached origin) in March 2026 benchmarks. WP Engine leads on consistency and 5-year longitudinal performance with 100% uptime in Q4 2025 and a documented 21% TTFB improvement from 2024 to 2025.
What is the difference between cached and uncached TTFB?
Cached TTFB is the response time when content is served from a CDN edge location, typically 20–100ms regardless of underlying server quality. Uncached origin TTFB is the response time when the CDN has no cache entry and must contact the actual origin server — this reflects hardware quality, database performance, and PHP execution speed. Most “fastest hosting” claims quote cached TTFB, which flatters any host with a good CDN. Uncached origin TTFB is the metric that matters for logged-in WordPress users, WooCommerce dynamic pages, and any URL that cannot be cached.
Does web hosting speed directly affect Google rankings in 2026?
Not directly via TTFB, but yes indirectly via Core Web Vitals. TTFB is not listed in Google’s ranking factor documentation. LCP (Largest Contentful Paint) is — it is one of three Core Web Vitals Google confirmed as ranking signals. Since TTFB is the primary driver of LCP on most sites, hosting speed has a clear path to ranking impact. The March 2026 core update reinforced Google’s emphasis on page experience signals including Core Web Vitals. Sites that consistently fail LCP face compounding ranking disadvantages in competitive queries.
Is SiteGround faster than Kinsta?
SiteGround’s 32ms cached TTFB (US East, March 2026) is numerically lower than Kinsta’s 85ms cached TTFB in PageSpeedMatters’ testing. However, this comparison measures different CDN delivery paths, not origin server quality. At the uncached origin TTFB layer — the more meaningful comparison — Kinsta’s 198ms (US East uncached) is considerably faster than SiteGround’s ~140ms for GoGeek. Both are competitive. The decision between them is primarily about price, scale, and features rather than speed: SiteGround costs $2.99–$44.99/month depending on plan and renewal; Kinsta starts at $35/month with visit-based overage pricing.
Methodology note — cite this data: All benchmark figures in this report are attributed to named sources with dates. Where data from different sources conflicts (particularly Kinsta TTFB figures), this report explains the reason for discrepancy (cached vs uncached measurement layer) rather than selecting the flattering number. If you’re citing any figure from this article in research or publication, use the specific source and date noted alongside each figure, not the BitsFromBytes summary.
Last reviewed: April 16, 2026. Speed test data reflects Q1 2026 benchmarks; hosting pricing reflects April 2026 published pricing. This article updates quarterly when new benchmark data is published.
![GTA VI Trailer 3 Release Window, Pre-Order & Price [April 2026 Tracker]](https://bitsfrombytes.com/wp-content/uploads/2026/04/gta-vi-trailer-3-release-window-pre-order-price.webp)