OG Xbox was sold for a 183.10$ loss adjusted to 2019, 360 was sold for a $164.28$ loss adjusted to 2019. Xbox One was actually MS's first console sold for a profit (in theory because BOM doesn't represent 100% of costs). After the trauma which was the One launch, I'm pretty doubtful that MS is going to go with that strategy again.
Selling to the core in the first year makes or breaks a console. MS can talk about services all day long, but if they are launching the same console at the same price point as Sony, they will get run over again. Maybe not as bad as last time, but still. Losing money on their hardware on launch isn't something that will anger shareholders, losing money the first year is a very standard thing in the industry. Gaining grounds, that's what important in the first year, not how much money you make. You want a lot of people buying your console fast so when the casuals come, all their expert friends already own your console, you want that landslide effect. The two years are too important, this gen they killed the One beyond repair, last-gen they built the 360.
People who would play in 1080p instead of 4K for 100$ less aren't counted on one hand, it's actually the other way around. We talk resolutions and tech all day long in our Era echo-chamber but in real life Wii outsold everything, Switch outsold everything and PS4 outsells Pro by 3:1. If at launch there is a next-gen console that plays everything next-gen but in 1080p for 100$, it will probably be a very popular SKU, just not amongst the Era crowd.
Regarding the power difference, if the CPU is the same, the memory is the same (with adjustments because 1080p requires less volume and bandwidth) and the GPU instruction set is the same, you will get the same game only in lower resolution or with some lower graphical settings. PC gamers had it for decades, there is a reason why NV sells both a 6.45TF 2060 RTX and a 16.3TF Titan RTX. Nintendo Switch is another example, you pull it out of the dock and you lose 50% of its GPU power, so the console cuts resolution and you get the exact same game running on a handheld because the CPU and memory setup is the same.
Regarding lowering the BOM, if we are going for an X2.5 less powerful GPU in the 1080p Lockheart, you can (let's assume we have a 5700 XT in the Anaconda):
- Cut the CU count from 40 to 20.
- Lower the clock from 1755Mhz to 1404Mhz.
- Cut the 320-bit interface to a 192-bit memory interface.
- Move from 14Gbps GDDR6 to 13Gbps GDDR6 (which will result in 312GB/s which is a lot for a 1080p console).
- Cut memory from 18GB to 12GB (if we assume ~18GB from Scarlett's X10 mixed chips design).
- Cheaper cooling.
- Cheaper power supply.
- Less I/O connections.
- Smaller form factor.
- Cheaper case (materials and finish).
- Cut SSD from 1TB to 500GB (makes sense because game sizes will be smaller and the target audience is more casual).
We are reducing the BOM by way more than 100$ here, I would even say ~150$, depends on how expensive Scarlett's SSD, memory and APU are.
There are three companies that I can think of that can actually make a good streaming infrastructure out of the gate because they are knee-deep in the server world and that's Amazon, Google and MS. There are three companies that can make a good gaming service out of the gate because they are knee-deep in gaming and that's Sony, MS, and Nintendo. The only company that appears at the intersection of that Van diagram is MS, so I wouldn't place Xcloud in the same boat as PSNow or Stadia.
High clocks are awesome, they also draw a lot of power and generate a lot of heat. The 5700 XT is (if we ignore the Vega VII rush-job) AMD's first 7nm card so I'm sure they will improve their 7nm performance by the end of 2020, but 5700 XT has shown us that 40CU at 2Ghz is way too much for a console right now in 2019. Who knows, maybe they will be able to achieve that, but you shouldn't take the Flute and Gonzalo leaks at face value. Getting a peek at two random lab testing out of hundreds doesn't really mean anything, just that AMD had tested the GPU at 1Ghz, 1.8Ghz, and 2Ghz at some point. It doesn't mean much regarding the final chip, just that at some point a guy at the lab forgot to pull the Ethernet cable out of the test machine while he conducted a 2Ghz benchmark.
Here is a random thought exercise - Can you think of any console that has a higher clock then it's PC counterpart? Because both 1.8Ghz and 2Ghz with 40CU are faster clocks than the 5700XT during gaming (according to AMD, gaming clock is 1755Mhz).
