The structural shift: capex (server + sysadmin) → opex (managed service, administration included)

Synthesis: The cost-floor story is not only about cheaper unit prices. It is also a structural shift from capex + ongoing ops labor to pay-as-you-go opex with the provider absorbing administration.

Then: dedicated server + colocation rack space + bandwidth + a sysadmin to patch / backup / fail over the machine. Capex on the box, opex on the labor.

Now: managed hosting + serverless compute (AWS Lambda previewed November 13, 2014; GA April 9, 2015 — code with no servers to provision) + managed databases (Amazon RDS announced October 2009 (MySQL first); GA May 31, 2011 — managed DB absorbs admin/backup/failover) convert that into pay-as-you-go opex.

Practical SMB delta: A hobbyist or SMB can now run, for ~$5–$50/month, a managed-DB-backed app that in ~2005 implied a dedicated server (Pre-cloud dedicated/colocation pricing 2004–2007 — managed dedicated $250–$1,400/month; budget $89–$199; 1U colo ~$50/month) plus part of a sysadmin's salary (Network & Computer Systems Administrator BLS median: ~$51K–$58K (early 2000s, needs-verification) rising to $96,800 (May 2024)).

Source: Synthesis of the pre-cloud + post-cloud entries above.

Confidence: Industry-consensus.

Why this matters for Candid: The unit-price drop (storage, compute, bandwidth) is the headline. The structural shift is what actually moved the SMB into reach. Both belong in any cost-floor article.