It was not only the MS upgrade / new licensing terms & cost increases that hit some of them pretty hard, it was also the end of life for MS Windows versions (XP, servers) and that many have outsourced their IT services / departments to 3rd party providers like CapGemini, Accenture, CGI etc.
This year e.g. with the EOL of Windows 2003 server, we have heard the word "extortion" quite a few times too often with some of the agencies.
To start with, the migration of a MS Win 2003 server to e.g. MS Win 2008 R2 or 2012 can easily take 20+days.
Now multiply this with £1K p/d, plus 3rd party service uplifts, training etc.
When you look at a larger scale data-intensive agency (10k+ staff) you are talking 800-1k+ servers running MS Windows. Only a fraction of those requiring an upgrade - e.g. 70-100 - already means £Millions. The licensing cost within that are of course marginal compared to the overall TCO. What is hitting the hardest is that the 3rd party suppliers will treat the new OS versions as a "different" OS. Such migrations (MS 2003 server, like with XP to Win7) require that quite some of the legacy software running on those servers has to be upgraded / needs new service contracts / does not immediately run / has to be fully re-tested. Plus new HW of course (not included in above cost).
Similar can be said for the XP to Win7 upgrade. While there are some automation tools to test your end user application estate (larger agencies have / had 800+ "standard" applications installed) a good estimate for such a multi-year upgrade programme is in the region of £3-5M+ for 12-15k end user computers. Windows licensing costs within these figures are marginal (e.g. 5-10%) or even included within the new HW that will cost on top of this migration (est. £600 * 8-10k+ -> £6M+ computers over 3-5y). These numbers go up substantially if you are upgrading high-security environments.
These figures are for mid-size / larger UK agencies with annual IT budgets in the region of £200-300M+ - the largest UK agencies / government departments have £1B+ annual IT budgets. The situation with local government is completely different - they have to calculate with "every penny", often have only a few £M annual budgets, not having these "favourable" MS licensing terms (heard about massive increases in the last years for local governments), are already overstretched in their other budget areas - as a result of IT cost increase they will often have to cut services elsewhere. FYI, MS is much better in this arena - as for example - Oracle. There was a case documented this year when Oracle sent a UK local government a £600M annual licensing bill (increase)- Oracle, after this "shock therapy" then normally negotiates & settles for a much lower sum. Oracle claims to have an almost 80% profit margin with these kind of "licensing deals".
Hope these real-world figures help to better understand where the issues are. The main cost - IMHO - are with the way how IT is delivered & serviced (3rd party at very high cost), constant changes in licensing terms for smaller to mid-size organisations, application licensing for specific MS Win versions, end of life of core IT building blocks vs continuity and gradual upgrades. If only a fraction of these amounts go into improving (end-user) open source solutions, we will have a boost in usability & features with most of these solutions. Additionally in these cases those improvements can then be shared across all government agencies, in the UK & world-wide. In the moment each improvement or application adaptation is always charged for multiple times (within the larger gov.agency scenario often by the same 3rd party suppliers).
It would be great if someone in those government agencies would document these actual cost and Central Gov would collect this data e.g. for base-lining against e.g. Linux (there are e.g. already thousands of Linux / Unix servers running within UK government agencies).
As with the Munich example mentioned a few time in this discussion, the above numbers / ballparks would have to be used for true comparison as Munich would have had to upgrade from XP to Win7 and also face the whole server upgrade scenarios. Instead much lower MS upgrade costs have been used and most importantly - upgrades required to all the (end user) applications, forms, macros & services that would most likely also have been required within the MS Win upgrades XP -> Win7, are now being told as Linux specific cost.