Ask HN: Video streaming is expensive yet YouTube "seems" to do it for free. How?

Hacker News Discussion ↗

YouTube’s seemingly “free” video streaming rests on massive ad revenues, deep integration with Google’s global infrastructure, and economies of scale that drive bandwidth, storage, and encoding costs down to fractions of a cent per gigabyte. Commenters describe how Google’s own data centers, custom encoding hardware, edge caches inside ISPs, and favorable peering deals make costs that would crush a startup manageable, while ads (and increasingly, paid Premium subscriptions) attempt to cover huge egress bills and creator payouts. Many conclude that at internet scale video can be profitable, but only for a handful of giants with existing ad businesses and networks, making true competitors to YouTube very hard to sustain.

YouTube’s revenue and unclear profitability

  • Reported $31.5B in ad revenue in 2023; some say this alone likely covers costs, others note revenue ≠ profit and that payout to creators (40–60%) plus infra may leave thin margins.
  • Historical reports suggested YouTube was break-even or loss-making; more recent analyst commentary suggests some parts (e.g., YouTube TV) are reaching profitability.
  • Several commenters think Google may keep YouTube near break-even for tax and strategic reasons; exact P&L is unknown because Alphabet does not break it out.

Cost structure: what’s expensive vs cheap

  • Consensus: bandwidth/egress is the dominant variable cost at scale; encoding and storage are significant but comparatively smaller.
  • Conflicting claims on bandwidth: some say “bandwidth is effectively free” at large scale via peering; others with infra experience say it is still “real money” and a primary cost driver.
  • Storage is cheap per TB but huge volumes (exabyte scale) plus replication and performance make it non-trivial; long-tail, low-view videos complicate tiering and deletions.
  • Encoding is expensive in aggregate but manageable with idle compute and hardware acceleration; quality is tuned down vs “home-rip” quality to save cycles.

Infrastructure, peering, and caching

  • Google owns dark fiber, a global backbone, and has cache nodes/CDN-like infrastructure in IXes and inside/near many ISPs.
  • Popular videos are heavily cached near users; long-tail content may be served from fewer regions or colder storage, sometimes causing slower starts.
  • This global infra is shared with other Google services (search/ads), so video “rides along” on already funded network capacity.

Advertising model and user experience

  • Ads are the main revenue source; many commenters feel ad load and intrusiveness have increased sharply, making non-Premium use “unpleasant.”
  • Some suspect ad-blocking users get worse recommendations; others just see overall recommendation quality decline.
  • YouTube also monetizes via Premium and Music; pricing suggests per-user serving costs are modest but non-zero.

Strategic value beyond direct profit

  • YouTube gives Google massive control over video distribution, user profiling for ads, and a large corpus for AI training.
  • This strategic value may justify low margins or cross-subsidy from search ads.

Why competitors struggle

  • New entrants face:
    • Much higher per-GB bandwidth and CDN costs.
    • No massive ad sales machine or targeting graph.
    • Network effects: creators go where the audience is; audiences go where the content is.
  • Some smaller/niche players (e.g., corporate, live, adult, specialty) survive via narrower focus, own infra, or cheaper hosting/CDN strategies.

Technical tactics and optimizations

  • Heavy use of custom ASICs/FPGAs/GPUs for transcoding; multiple bitrates/resolutions per video.
  • Intelligent caching and storage tiering, possibly re-encoding older/low-traffic videos more aggressively.
  • Some discussion of P2P, Cloudflare R2, and DIY CDNs as ways small players can cut costs, but these add complexity and don’t erase YouTube’s scale advantage.