According to DEV Community, a developer measured what inline SVG icons actually cost on the homepage of pub-trivia.app. Inline SVG means the icon's drawing code sits directly inside the page's HTML, so it needs no separate network request. The author argues the cost is rarely measured because it does not show up in the Network tab or the JavaScript bundle.
The homepage HTML was 170,427 bytes and held 48 SVG elements. Together they came to 31,623 bytes, or 18.6 per cent of the document. Of the 48, 43 were small icons averaging 375 bytes each. The other five were copies of a hand-drawn mascot averaging 3,098 bytes each. One 240-byte check mark appeared 22 times.
The author then removed every SVG from the page and compressed both versions. According to the article, 31.6 KB of SVG markup came to about 1.6 KB on the wire. That turns 18.6 per cent of the raw page into 11 per cent of the compressed one. Repeated icons like the check mark compress almost to nothing after the first copy.
The author notes that inlining is not free in other ways. The 48 elements are still DOM nodes (the browser's in-memory page structure) that must be parsed and possibly drawn. The mascots compress less well because each needs a unique clip path id. A shared set of symbol definitions with references would cut about 15 KB raw to about 3 KB, but the team kept the current setup because each mascot animates independently.
On content pages the count is much lower. Every page under /guides ships five SVGs totalling 1,778 bytes. Three are hover-only arrows that cost 72 bytes after compression. The author suggests identical counts across a page cluster make odd pages easy to spot.
Why it matters
Inline icons are often blamed for page bloat, but this measurement suggests that for small, repeated icons the transfer cost is low. Compression handles repetition well, so raw HTML size can overstate what visitors download.
The bigger cost is likely to come from large or unusual graphics, such as illustrations repeated with unique ids. Those compress less well and add more work for the browser. The article is one site's data, so your own numbers may differ.
Source: DEV Community. This brief summarises their reporting; read the original for the full detail.