Your $25 flat maintenance fee is now an unhedged bet on edge-request volume

A rate breakdown published on dev.to puts figures on a shift most agency owners already feel in their bank balance: Vercel Pro includes 10 million edge requests and then charges $2 per additional million, includes 1 million function invocations and then charges $0.60 per million, and meters active CPU at $0.128 to $0.221 per CPU-hour depending on region ^[6][8][9]. The billing model went from fixed to metered; the standard maintenance retainer did not, so the agency is now the party absorbing traffic variance it does not control.

The old arithmetic was almost embarrassingly good. A $50-a-month reseller or dedicated cPanel box held 30 client WordPress sites, each billed a flat $25 a month ^[1]. That is $750 of revenue against $50 of infrastructure, about $1.67 of hosting cost per site and roughly 93 percent gross margin on the line ^[2]. More to the point, the cost side did not move. A capped VPS simply drops excess requests when it runs out of headroom; serverless auto-scales, and the invoice scales with it ^[5].

The per-project floor is where the model breaks quietly. Supabase bills a dedicated instance 24/7, from Micro at $10 to Small at $15 and Medium at $60 a month ^[10]. Thirty client projects on Micro is $300 a month, six times what the single shared server cost, before a request is served ^[22]. One Medium instance consumes 2.4 clients' entire monthly fees ^[23]. Neon takes the other approach, billing per compute-unit-hour and suspending to $0 compute when idle ^[11], which is some protection against the staging project nobody paused, one of the failure modes the post flags alongside running more than one project per org ^[17].

Seats are the other fixed leak. Vercel charges $20 per deploying member per month on Pro, with read-only Viewer seats free and unlimited, and extra paid seats carry no additional usage credit ^[12]. Netlify, as of 14 April 2026, charges a flat $20 a month for the whole team with unlimited seats ^[13]. Five deploying engineers is $100 on Vercel against $20 on Netlify, before any usage ^[24]. Contractors and client stakeholders left on paid tiers after a project ships are, according to the post, a routine drain ^[18].

The scale check in the piece is worth redoing by hand. Vercel has published that 100 million edge requests, before compute, would cost roughly $180 after the included allowance and Pro's $20 credit, against a lower figure Cloudflare quotes for comparable work on its own edge network ^[14]. But 90 million billable requests at $2 per million is exactly $180, which suggests the $20 credit is not actually netted out of the number as stated ^[20]. Either way, $180 of edge requests alone is 7.2 months of one client's $25 retainer ^[21], and 200 GB above Vercel's included 1 TB of Fast Data Transfer adds $30 ^[25]. The post describes its figures as verified rates for mid-2026 and notes that exact numbers move as pricing pages update ^[19][14].

Three things to watch. Whether edge-request volume and cache hit ratio get their own line in client reporting rather than a footnote under hosting, given that unoptimised media, poor caching and bot traffic are the named overage triggers ^[15]. Whether database and staging projects acquire an expiry date at the moment they are created ^[17]. And whether the flat retainer survives its first viral campaign or unthrottled crawler, since uncached SSR and un-rate-limited middleware sit on the compute meter ^[16]. The post's own prescription is auditing, governance and renewal tracking ^[26]; the pricing question underneath it is simpler, which is who pays for the second standard deviation of traffic.