What are the biggest companies in Europe?

By revenue, the ten biggest in Fortune’s 2026 Europe 500 are Volkswagen, Shell, Glencore, BP, TotalEnergies, Stellantis, BMW Group, Mercedes-Benz Group, Banco Santander and BNP Paribas. Volkswagen leads at $363.1 billion. The full top 50 is in the table above.

Which of Europe’s biggest companies has the fastest website?

On Google’s mobile lab test, Shell scores highest at 97 — but that score was taken on a pre-rendered snapshot served only to Google’s tools, and real visitors to Shell’s site fail Core Web Vitals. Setting Shell aside, Volkswagen is fastest in the lab at 77. For real visitors, 21 of the 47 measured sites pass all three Core Web Vitals.

Which has the slowest website?

In the lab, Bunge at 26 and Lloyds Banking Group at 29 score lowest. For real visitors the picture differs: Lloyds passes Core Web Vitals, while E.ON’s visitors wait 7.8 seconds for the largest element and BBVA’s wait 1.7 seconds for a tap to respond.

Why are the lab and real-visitor numbers so different?

The lab loads each page once, with an empty cache, on a simulated mid-tier phone over a throttled connection. Real-visitor data averages actual Chrome users over 28 days, many on fast devices and repeat visits. The lab is the better guide to what to fix; the real-visitor data is what Google uses to assess Core Web Vitals.

Why three runs per site?

Because a PageSpeed score moves on its own. Crédit Agricole returned 33, 46 and 61 on three consecutive runs of an unchanged page, and 16 of the 47 sites moved ten points or more. The median of three is far more stable than any single run.

Is serving a pre-rendered page to Google cheating?

Not in Google’s rules, as long as the content matches what visitors get; Google calls dynamic rendering a workaround rather than a recommended solution. But it means a lab score measures a page real visitors never receive, so it should not be quoted as the site’s speed.

Our company is in the table and a figure looks wrong. What now?

Tell us and we will re-measure and correct it, with the date of the correction noted. Every figure comes from a script run against Google’s public API, and the method above is enough to reproduce it.

← All resources

— Original research

The 50 biggest companies in Europe, and how fast their websites really are.

Every homepage from the top 50 of Fortune’s 2026 Europe 500, put through Google’s PageSpeed Insights three times and checked against what real visitors experience — on the same morning.

By Level UIPublished 5 October 202623 min read

Europe’s 50 biggest companies took in $5.8 trillion in revenue between them, and every one of them has a corporate homepage that investors, journalists, job applicants and customers open first. So we measured them. We took the top 50 of Fortune’s 2026 Europe 500, ran each homepage through Google’s PageSpeed Insights on the mobile test three times, and set the lab score against Google’s own data on what real visitors experienced over the previous 28 days. The median score is 54 out of 100. Only one site reached Google’s green band, and that one turned out to be serving Google’s tools a different page from the one its visitors get. Real visitors fare better than the lab suggests, but fewer than half of these sites pass Google’s Core Web Vitals for them. The full table, the method and every caveat are below.

PageSpeed Insights speed score, mobile

Every bar is one corporate homepage on Google’s PageSpeed Insights mobile test, the median of three runs, scored out of 100. Google paints anything below 50 red and anything below 90 amber. Select a bar for its page weight and what real visitors experience.

Shell

Scored on a pre-rendered snapshot served only to Google’s tools; visitors get an empty page that builds itself with JavaScript.

Detail

No. 2 by revenue. Runs: 92 / 97 / 97.
Lab LCP 2.2 s; real-visitor LCP 3.0 s at the 75th percentile.
200 KiB across 12 requests. Real visitors: fails Core Web Vitals on LCP and INP.

Median of three PageSpeed Insights runs per homepage, mobile strategy, Lighthouse 13.5.0, measured on 2 October 2026 between 08:07 and 08:35 UTC. Three sites could not be measured and are not shown; Gazprom has two valid runs instead of three.

All 50 companies

Company		Country
Shell	2	UK	97 (pre-rendered)	2.2 s	3.0 s	200 KiB	12	Fail
Volkswagen	1	Germany	77	5.0 s	2.5 s	4,054 KiB	26	Fail
Electricité de France	15	France	70	10.2 s	2.5 s	1,995 KiB	50	Pass
Glencore	3	Switzerland	70	5.6 s	2.6 s	2,753 KiB	19	Fail
Bosch Group	22	Germany	69	7.1 s	2.2 s	2,303 KiB	55	Pass
Groupe BPCE	40	France	68	5.3 s	1.6 s	2,499 KiB	47	Pass
TotalEnergies	5	France	68	10.1 s	1.8 s	1,952 KiB	42	Pass
Gazprom	16	Russia	65	4.8 s	2.2 s	3,045 KiB	34	Pass
BNP Paribas	10	France	64	6.3 s	2.0 s	2,311 KiB	64	Fail
EP Group	43	Czechia	64	5.3 s	2.7 s	8,197 KiB	87	Fail
Christian Dior	29	France	63	7.9 s	2.1 s	1,969 KiB	51	Pass
Société Générale	23	France	63	7.2 s	1.9 s	5,530 KiB	116	Pass
Banco Bilbao Vizcaya Argentaria	33	Spain	61	7.1 s	2.2 s	888 KiB	61	Fail
BP	4	UK	61	13.4 s	1.6 s	2,309 KiB	55	Pass
Barclays	44	UK	60	5.7 s	1.9 s	9,469 KiB	83	Pass
HSBC Holdings	11	UK	60	10.3 s	1.8 s	1,774 KiB	39	Pass
UBS Group	41	Switzerland	60	10.2 s	1.4 s	1,458 KiB	95	Pass
BMW Group	7	Germany	59	11.8 s	2.7 s	3,858 KiB	132	Fail
Mercedes-Benz Group	8	Germany	57	9.2 s	2.7 s	1,193 KiB	106	Fail
Allianz	14	Germany	56	15.6 s	2.1 s	2,905 KiB	104	Pass
Banco Santander	9	Spain	55	9.5 s	2.4 s	10,080 KiB	113	Pass
REWE Group	21	Germany	55	18.7 s	1.4 s	6,844 KiB	71	Pass
Stellantis	6	Netherlands	55	16.7 s	1.5 s	3,795 KiB	53	Pass
JBS	34	Netherlands	54	26.9 s	2.1 s	6,180 KiB	78	Pass
Tesco	24	UK	53	20.8 s	2.1 s	6,434 KiB	57	Pass
Deutsche Bank	46	Germany	52	11.6 s	2.4 s	2,033 KiB	61	Pass
Engie	39	France	52	18.8 s	2.1 s	8,060 KiB	75	Fail
Royal Ahold Delhaize	20	Netherlands	52	18.7 s	2.0 s	3,190 KiB	46	Pass
E.ON	31	Germany	51	15.8 s	7.8 s	2,926 KiB	213	Fail
Siemens	32	Germany	50	4.9 s	3.2 s	4,449 KiB	130	Fail
ING Group	45	Netherlands	48	39.6 s	3.7 s	13,543 KiB	100	Fail
AXA	18	France	47	20.0 s	3.8 s	4,211 KiB	73	Fail
DHL Group	28	Germany	47	15.8 s	2.5 s	2,490 KiB	103	Pass
Rosneft Oil	26	Russia	47	16.0 s	3.1 s	3,130 KiB	117	Fail
Crédit Agricole	37	France	46	13.4 s	5.6 s	2,115 KiB	42	Fail
Equinor	19	Norway	46	11.3 s	2.8 s	2,196 KiB	63	Fail
Roche Group	42	Switzerland	43	12.5 s	3.4 s	1,926 KiB	51	Fail
BASF	47	Germany	42	15.6 s	2.9 s	2,765 KiB	117	Fail
Deutsche Telekom	13	Germany	42	7.6 s	1.3 s	1,102 KiB	79	Fail
Enel	30	Italy	42	13.4 s	4.1 s	2,664 KiB	61	Fail
ENI	27	Italy	41	16.3 s	2.8 s	3,599 KiB	115	Fail
Nestlé	17	Switzerland	40	14.8 s	6.1 s	2,543 KiB	82	Fail
Airbus	38	Netherlands	39	10.6 s	2.2 s	5,635 KiB	51	Fail
Carrefour	25	France	38	7.4 s	2.3 s	993 KiB	65	Fail
Orlen	48	Poland	37	12.4 s	2.3 s	4,491 KiB	40	Fail
Lloyds Banking Group	35	UK	29	13.3 s	1.3 s	2,269 KiB	57	Pass
Bunge Global	49	Switzerland	26	15.7 s	2.8 s	2,421 KiB	75	Fail
Sberbank	12	Russia	—	—	—	—	—	—
Vinci	36	France	—	—	—	—	—	—
VTB Bank	50	Russia	—	—	—	—	—	—

Fortune rank is by revenue in the Fortune Europe 500, 2026. Speed score, lab LCP, page weight and requests: median of three PageSpeed Insights mobile runs on 2 October 2026. Real-visitor LCP and the Core Web Vitals verdict come from the Chrome UX Report for the whole site (origin), 75th percentile over the previous 28 days; “Pass” means LCP, INP and CLS are all in Google’s good range. A dash means the homepage could not be measured.

What we measured, and how

The list is the top 50 of the Fortune Europe 500 for 2026, published on 16 September 2026 and ranked by revenue. Volkswagen leads it at $363.1 billion, followed by Shell, Glencore, BP and TotalEnergies. For each company we tested the corporate homepage — volkswagen-group.com rather than a car configurator, group.mercedes-benz.com rather than a showroom — because that is the one site every company on the list has and the one it controls centrally.

Each homepage went through Google’s PageSpeed Insights on its mobile test three times, and we kept the median. Lighthouse, the engine behind it, simulates a mid-tier phone on a mobile network, and its score moves between runs on its own: Crédit Agricole returned 33, 46 and 61 on three consecutive runs of a page that did not change, and 16 of the 47 sites moved ten points or more between their best and worst run. A single run is a sample, not a measurement. We only counted a run when its timestamp was new, because PageSpeed sometimes re-serves a cached result.

The same response carries a second, very different set of numbers: the Chrome UX Report, Google’s record of what real Chrome users experienced on the whole site over the previous 28 days. That is the data Google uses for its Core Web Vitals assessment, and it is the column most studies leave out. We report both, because they disagree more than anyone would expect.

Everything was measured on the morning of 2 October 2026 from Google’s own servers, with Lighthouse 13.5.0. Before trusting any score, we checked that Google had received the real homepage and not an error page, a bot challenge or a cookie wall — which is how we found the one result in this study that does not mean what it says.

The headline numbers

The median homepage of Europe’s 50 biggest companies scores 54 out of 100 on Google’s mobile test. Seventeen of the 47 we could measure sit in Google’s red band below 50, 29 sit in the amber band, and one reaches green. The lowest scores belong to Bunge at 26, Lloyds Banking Group at 29, Orlen at 37, Carrefour at 38 and Airbus at 39.

The median homepage weighs 2,753 KiB across 64 requests, and in Lighthouse’s simulation its largest element takes 11.6 seconds to appear. Google’s target for that measure, Largest Contentful Paint, is 2.5 seconds. In the lab, exactly one homepage meets it.

The biggest companies in Europe do not have fast websites, and it is not for lack of budget. These are businesses with entire digital departments and agency rosters. Whatever is making their homepages slow, it is not that nobody could afford to fix it.

The 97 that isn’t: Shell

Shell tops the chart with 97, the only green score in the study, on a homepage of 200 KiB and 12 requests. That is so far from everybody else that we checked what Google had actually been sent. A normal phone browser asking for shell.com receives 9 KB of HTML with an empty body and not a single link: the page builds itself afterwards with JavaScript. The same request with Lighthouse’s user agent, or Googlebot’s, receives 95 KB of finished HTML with 53 links and a response header from Prerender.io, a service that renders pages in advance for crawlers.

So Shell’s 97 is the score of a pre-rendered snapshot that only Google’s tools are served; visitors never get that page. Google itself calls this technique, dynamic rendering, “a workaround and not a recommended solution”, and recommends server-side rendering, static rendering or hydration instead. It is not cloaking as long as the content matches, and there is no sign here that it does not. But it does mean the lab score measures the wrong page.

The real-visitor data agrees. Over the previous 28 days, Shell’s site failed Google’s Core Web Vitals for real visitors, with the largest element taking 3.0 seconds at the 75th percentile and taps taking 229 milliseconds to respond, both outside Google’s good range. We checked all 50 homepages the same way. Shell is the only one that sends Lighthouse a different page. Leave it out, and the best score in the study is Volkswagen’s 77.

The lab says 11.6 seconds. Real visitors say 2.3

The most striking thing in the data is how far apart the two kinds of measurement are. In Lighthouse’s simulation the median homepage needs 11.6 seconds for its largest element. Real visitors to the same sites wait 2.3 seconds at the 75th percentile. Seventeen companies have a lab LCP over ten seconds and a real-visitor LCP inside Google’s good range of 2.5 seconds.

Both numbers are honest; they measure different things. The lab loads the page once, with an empty cache, on a simulated mid-tier phone and a throttled mobile connection. The Chrome UX Report averages real Chrome users across the whole site over 28 days, and the people who open a corporate homepage are disproportionately investors, analysts and journalists on recent phones and office connections, many of them on their second or tenth visit. Lloyds Banking Group shows the gap at its widest: 29 in the lab, the second-lowest score in the study, yet its real visitors pass all three Core Web Vitals.

Even on the friendlier measure, most of these sites fall short. Twenty-one of the 47 pass all three Core Web Vitals for real visitors — 45%. Twenty-eight are in the good range for loading, 38 for responsiveness and 38 for layout stability. Of the ten biggest companies by revenue, four pass: BP, TotalEnergies, Stellantis and Banco Santander. Volkswagen, number one by revenue and best in the lab after Shell, misses on loading by 48 milliseconds.

Bigger companies do have faster sites, slightly

There is a relationship between size and speed, and it runs the way you would hope. Ranked by revenue and by speed score, the 47 companies show a rank correlation of 0.49: a moderate tendency for the bigger company to have the faster homepage. The ten biggest have a median score of 62.5; the other 37 a median of 51.

It is a tendency, not a rule. Glencore, third by revenue, scores 70; Nestlé, seventeenth, scores 40; Groupe BPCE at number 40 scores 68 and passes every real-visitor test. Whatever decides how fast these homepages are, it is something their web teams choose, not something their size hands them.

Page weight is not what slows them down. JavaScript is

The heaviest homepages are very heavy. ING’s weighs 13.2 MB, Banco Santander’s 9.8 MB and Barclays’ 9.2 MB; ten of the 47 are over 5 MB. The lightest real ones, setting Shell aside, are BBVA at 888 KiB, Carrefour at 993 KiB and Deutsche Telekom at 1,102 KiB.

Yet across these 47 sites, page weight barely predicts the score at all — what predicts it is how long the main thread is blocked by JavaScript. The rank correlation between weight and score is −0.08, close to nothing. Blocking time is where the slowest homepages stand out: Orlen’s main thread is blocked for 976 milliseconds, Airbus’s for 803, Siemens’ for 798, BASF’s for 787 and Nestlé’s for 763. Lighthouse treats 200 milliseconds as the threshold for good. Big images cost bandwidth; scripts that run before the page can respond cost the score.

This matches what we found when we speed-tested 40 agency websites: weight is the easiest thing to measure and rarely the whole story. The work that moves a score is in the server and the scripts, and the full method for getting to 100 is in how to get a 100 PageSpeed score.

Pages that jump, and taps that wait

Two of the three Core Web Vitals are about what happens after the page appears. Cumulative Layout Shift measures how much the layout jumps while it loads; Google’s good range is 0.1 or less. In the lab, Carrefour’s homepage scores 0.63, Bunge’s 0.54, Deutsche Telekom’s 0.46 and Lloyds Banking Group’s 0.44 — the kind of jump that moves a button just as someone goes to tap it.

Interaction to Next Paint measures how long the page takes to respond to a tap or a click, in real use; good is 200 milliseconds or less. Most of these sites manage it, but some do not by a long way. Real visitors to BBVA’s site wait 1,698 milliseconds for a response at the 75th percentile, more than eight times the target. Nestlé’s visitors wait 390 milliseconds and Siemens’ 387. On a corporate site that is mostly reading, a slow tap usually means a heavy script busy doing something else when the visitor reaches for the screen.

What the three metrics measure, and why Google chose them, is in Core Web Vitals explained.

By country and by sector

Splitting 47 sites by country leaves small groups, so treat this as a description rather than a league table. France’s ten companies have the best median score at 63. The UK’s six score 60, Germany’s twelve 53.5, the Netherlands’ five 52 and Switzerland’s five 43. The single Czech, Norwegian and Polish companies score 64, 46 and 37, which says something about EP Group, Equinor and Orlen and nothing about their countries.

By sector, the two biggest groups are almost level: fourteen financial companies have a median of 58 and fourteen energy companies 56.5. Food and drink is weakest, with Nestlé, JBS and Bunge at a median of 40. The pattern that matters is not between countries or industries; it is between individual web teams, which is where the decisions are made.

The three we could not measure

Three homepages did not produce a result, and we report them rather than quietly dropping them. Sberbank’s page never painted anything in Google’s test, three times out of three. VTB Bank’s could not be loaded at all from Google’s servers. Both are Russian banks, and both were tested from the same place as every other site, so we record them as unreachable from Google’s test machines, not as slow.

Vinci, the French construction group, is the odd one out. Its homepage loads normally from our office, in a browser and with Lighthouse’s user agent. Google’s servers were told, three times, that the page was not HTML at all — most likely bot protection treating Google’s data-centre addresses differently. The site may well be fast; from where Google measures, it is not there.

Gazprom produced two valid runs rather than three, and its median is the middle of two. We left it in, marked in the chart footnote, because two runs is still more than most published speed comparisons use.

Why a corporate homepage’s speed matters at all

None of these companies sells anything on its corporate homepage, so it is fair to ask whether any of this matters. It matters in three smaller ways. The homepage is the first impression for the people a company most wants to impress — investors, journalists, graduates deciding where to apply — and a page that takes eleven seconds on a phone in the lab is one some of them will close. Speed is also one of the signals Google uses for page experience: a tie-breaker in the results rather than a lever, but a tie-breaker on some of the most searched company names in the world.

The third is less obvious. Two of these homepages, Shell’s and Tesco’s, send almost no content in their HTML and build the page with JavaScript. Five more — Allianz, E.ON, Bunge, REWE Group and Nestlé — answered a plain request from our office with a bot challenge instead of the page. Google renders JavaScript and is let through. Many of the new AI crawlers do not run JavaScript at all, which is the subject of what AI crawlers actually send back. A page that is empty without scripts depends on every reader being able to run them.

For a smaller business the lesson is the reassuring one. If the biggest companies in Europe ship a median of 2.7 MB and fail real-visitor Core Web Vitals more often than not, a small company’s site can be faster than theirs with a few weeks of careful work. A fast site is a series of choices, not a budget line — the same thing we found in what popular websites have in common.

How to check your own site the same way

Open [PageSpeed Insights](https://pagespeed.web.dev/), enter your homepage and run it three times; take the middle score. Then look at the top section, “Discover what your real users are experiencing”. If your site has enough Chrome traffic, that is the Chrome UX Report, and it is the verdict Google uses for Core Web Vitals. If the two disagree, believe the real-visitor data about your visitors and the lab data about what to fix.

If the score is surprisingly high, check what Google was served. Open the “View treemap” or the final screenshot in the report and make sure it is your page, not a cookie wall, an error or a snapshot made for crawlers. If it is low, look at Total Blocking Time before page weight: in this study, scripts explained far more than bytes did.

If you want somebody else to do it, a technical audit covers exactly this ground, page by page. Or ask us about web development if the answer turns out to be a rebuild — it often is not.

Method, tools and sources

Every score in this article was measured on 2 October 2026 between 08:07 and 08:35 UTC through the PageSpeed Insights API, mobile strategy, three distinct runs per homepage, and can be reproduced with the tools below. The company list and ranks are Fortune’s.

Fortune Europe 500, 2026 ↗
The ranking of Europe’s largest companies by revenue, published 16 September 2026. The top 50, their revenue in US dollars, country and sector are taken from it; Volkswagen is first at $363,090 million.
Fortune — The 2026 Fortune Europe 500 ↗
Fortune’s article accompanying the 2026 list.
Google — About PageSpeed Insights ↗
Lab data simulates a mid-tier device on a mobile network; field data reports real users’ experience over the previous 28 days and can be shown at page or origin level. This study uses origin-level field data for every site.
Chrome UX Report — Methodology ↗
Who contributes to the Chrome UX Report (opted-in Chrome users on desktop and Android, not iOS), the popularity threshold for inclusion, and how origin-level data aggregates all pages of a site.
web.dev — Web Vitals ↗
Google’s thresholds for a good experience at the 75th percentile: LCP within 2.5 seconds, INP of 200 milliseconds or less, CLS of 0.1 or less.
Lighthouse performance scoring ↗
How the 0–100 performance score is weighted from its metrics, and why it varies between runs of the same page.
Google Search Central — Dynamic rendering as a workaround ↗
Google: dynamic rendering “is a workaround and not a recommended solution”; it recommends server-side rendering, static rendering or hydration, and does not treat dynamic rendering as cloaking while the content is similar.
Google Search Central — Understanding page experience ↗
Where page experience, including Core Web Vitals, sits among Google’s ranking signals.

Was this helpful?

— FAQ

Frequently asked questions

— Keep reading

More resources.

Guide

Dark Patterns: 12 Examples, the EU Law and the Cost

Twelve named dark patterns with real examples, the EU rules each one collides with, what they have cost the companies caught using them, and what to build instead.

Guide

How to Build a Website: A Step-by-Step Guide

Thirteen steps from first idea to launch, with real tools, real prices and honest time estimates — written by an agency that will tell you when you do not need one.

Top

What Is Docker Used For? 10 Creative Ways

A short answer to what Docker actually is, then ten uses past the obvious one — ranked by what they take off your week, each with its honest limit.

Guide

Google Ads vs Facebook Ads: What Each Really Costs

Both bill whatever budget you set. What matters is the price of a click in your market, what a lead turns into, and how much you spend before either tells you anything.

Guide

Google Business Profile Optimization

Google says local results come down to relevance, distance and prominence. Here is what that means in practice, what the tests show, and which popular tips are a waste of an afternoon.

Price guide

Link Building Cost: What Backlinks Cost in 2026

A link can cost nothing or a thousand dollars, and the price tells you very little about what it is worth. Here is what the market charges, what Google and the law allow, and where the value actually is.

— Keep reading

More resources.

Guide

Dark Patterns: 12 Examples, the EU Law and the Cost

Twelve named dark patterns with real examples, the EU rules each one collides with, what they have cost the companies caught using them, and what to build instead.

Guide

How to Build a Website: A Step-by-Step Guide

Thirteen steps from first idea to launch, with real tools, real prices and honest time estimates — written by an agency that will tell you when you do not need one.

Top

What Is Docker Used For? 10 Creative Ways

A short answer to what Docker actually is, then ten uses past the obvious one — ranked by what they take off your week, each with its honest limit.

Guide

Google Ads vs Facebook Ads: What Each Really Costs

Both bill whatever budget you set. What matters is the price of a click in your market, what a lead turns into, and how much you spend before either tells you anything.

Guide

Google Business Profile Optimization

Google says local results come down to relevance, distance and prominence. Here is what that means in practice, what the tests show, and which popular tips are a waste of an afternoon.

Price guide

Link Building Cost: What Backlinks Cost in 2026

Want to know how your site compares?

We will measure it the same way — lab and real visitors — tell you what is slowing it down, and give you the fixes in priority order.

Start a project WhatsApp