These are my estimates, not from BT. I combined Andrew Ferguson's mapping results with cost estimates from several industry sources. Carriers use DSL instead of fiber because it's much cheaper, with likely costs of $100-$300 for the gear and installation. Even so, I was amazed when I calculated the BT cost for 10M homes servable. It's probably less than ~$300M/year for four years.
That's a heckuva of money to you and me, but a very small investment for a company the size of British Telecom. ~$300M is ~1% of revenue and 10% of the capital budget. It's probably less than BT is currently spending on DSL at slower speeds and will not require increased investment. The cost will be less if you make some plausible assumptions.
The original plan, FTTdp, would probably have cost five times as much. Running fiber close to everyone would cost more than the equipment. Going to the existing cabinets saves that cost, although the change drops many people's speeds from 500-750 megabits to 100-300 megabits. Thanks to Andrew Ferguson at ThinkBroadband, I can calculate the likely costs of the cabinet build.
96 ports of DSLAM at $250 is $24,000. Multiplied by 40,000 cabinets you come to $960,000,000. Add 25% for overhead and cost overruns raises that to ~1,200,000,000, about 1% of BT's sales in the four years. A telco looking for a higher figure than $250 for a press release would come much higher. A vendor looking to bid low enough to guarantee the contract would probably come in much lower. Both could find senior economists who will confirm their figures for a fee.
Andrew's mapping concludes BT will have to upgrade 25,000-40,000 cabinets to offer 10M Brits 100 megabits or more. Ferguson also points out that the cabinets will be served with no more than 96 ports initially, which only works if fewer than ~30% take high speeds. Putting just 96 ports in each reduces the number of DSLAMs needed by 2/3rds or more, bringing the initial capex down further. If more than 96 homes want high speeds, the extra they will pay will more than cover the cost of an additional DSLAM.
Someone from Nokia Alcatel or Huawei may call me saying they can't price at $24K installed. That may be true in 2017 but will come far down by 2019 & 2020. For ten million lines, they can find a way.
A few weeks ago, BT's stock rose ~ $5B in a few days. That could have to bring 500 megabits to a majority of Brits, I calculated, going to the distribution points. It could alternately finance fiber home to much of the country.
That's inoperative because BT's stock has come down $15B in the two days since Brexit, amidst a $40B rout in the prices of European telcos.