A quote to translate a ten thousand word site into one more language lands somewhere between seventeen hundred and five thousand dollars. Standard rates, no surprises. That number is real, and it is also the smaller half of the question. It prices the words already on the page. It says nothing about the build that has to exist before a single translated word renders in front of a visitor.
The build starts with a decision nobody puts in the translation quote: where does the second language actually live. Google lays out four options. A country specific domain sends the clearest signal but costs the most to run. A subdomain is easy to stand up, though visitors often cannot tell if it means a language or a country. A subdirectory is the cheapest to maintain, since it stays on one host. A query parameter is the one Google says outright not to use, because it is too hard to separate cleanly. Whichever gets picked, hreflang tags become mandatory the moment separate URLs exist, so search results send people to the version they can actually read.
There is a shortcut a lot of teams reach for first: swap the page content by browser setting, or redirect visitors automatically once their language is detected. It feels efficient. Google is specific about why it fails anyway, it might never crawl the other version at all, because there is no distinct URL for it to find. A page that only exists behind a toggle is, as far as search is concerned, a page that does not exist.
Once routing is settled, the translation number gets more useful. General content runs ten to thirty cents a word, technical or legal copy forty to sixty. An in house team costs two hundred to three hundred thousand dollars a year once salary, software and management are counted, and that only makes sense for a company publishing new pages daily. Everyone else is choosing between an agency billed against the actual project (roughly forty five hundred dollars for that same ten thousand words across three languages at fifteen cents each) or a machine first pass a person cleans up afterward.
None of that is what catches people out. The build is a cost you pay once. Keeping two languages in sync is a cost that starts the day the site goes live and never really stops. Every page added from here on has to ship in both languages, or the second one quietly falls behind the first within a year. A sentence that runs longer in French or German than it did in English still has to fit a layout it was never designed to overflow. That is a design problem, not a translation problem.
Translation memory and an API tied into the CMS can cut that ongoing cost by a real margin. But only when it is built in from day one. Bolted on after launch, it is just another migration.
The word count invoice is the easy number to get right. The routing, the hreflang, the layout that has to hold in two languages, and the year of pages that follow launch: that is the actual cost of a second language. It is the conversation worth having with whoever is building the site, before the translation quote lands, not after.