Instead of throwing around percentages with little basis, I believe it is better to think in terms of net present value:
First step: make a forecast of difficulty.
Second step: make the choice: if (current price that a miner can be sold for) > (forecasted net present value of the mining profit from the miner), then SELL IT.
Of course, this relies on forecasting difficulty somewhat accurately. More importantly, if there are currently space/power limitations on deploying the miners, it is very likely that the choice of selling a large amount of miners immediately is the most profitable one.
With some more numbercrunching, one could also work with a probability distribution around the forecasted difficulty, and use portfolio theory to hit the risk/reward sweet spot.
First step: make a forecast of difficulty.
Second step: make the choice: if (current price that a miner can be sold for) > (forecasted net present value of the mining profit from the miner), then SELL IT.
Of course, this relies on forecasting difficulty somewhat accurately. More importantly, if there are currently space/power limitations on deploying the miners, it is very likely that the choice of selling a large amount of miners immediately is the most profitable one.
With some more numbercrunching, one could also work with a probability distribution around the forecasted difficulty, and use portfolio theory to hit the risk/reward sweet spot.
Well, my point is that there's sure to be a perfect ratio that would generate maximum profit at any given point in time, and I'm pretty sure that, until all other ASIC manufacturers get up to speed, mining in the short term at much higher than 10% of global hashrate is better since difficulty is lower now than it will be in the future, and we seem to have a definitive leg up on the other manufacturers... but I have no idea what the optimal ratio actually is today, let alone in the future. It is, after all, a moving target... you're absolutely right that it depends entirely on total global hashrate (i.e. difficulty) and cost for devices per TH/s in the open market versus the going exchange rate for fiat. If BTC goes either up or down in value versus fiat, that will affect the optimal ratio of mining to selling by AM, as the prices people are willing to pay for devices, even in BTC, is affected by how valuable the BTC being generated are against their preferred fiat. If exchange rates are flat and hardware is not due for a generational update (as in changing the die size to a smaller process width), then it makes more sense to sell shovels. But, if we have a leg up in technology, it makes sense to keep that new technology (the "new shovels", so to speak) to ourselves until someone else comes out with a good competing product, while unloading the old shovels to the market. Only when there's a balance of power should we primarily focus on selling shovels, since the effect of our own addition to the hashrate would be balanced by the other manufacturers. Whenever we have the advantage, we displace our own in-house shovels with the new ones until it makes more sense to sell them, since our addition would have a greater effect on the difficulty levels and we would be the only ones able to take advantage. You are also very much correct when it comes to physical capacity, since we can't mine if there's no juice/space/cooling to run the machines.
But I see that as an exception to the formula given the amount of space contracted recently.I think this is mostly a time-to-market problem, really. The faster we can get the new shovels to market compared to the other ASIC manufacturers, the more we should hold back sales to be able to use them ourselves and to encourage the highest prices possible for the new shovels that we *do* sell. And, to circle around to my original point, if we commit to a certain percentage of the global hashrate, we have an advantage against the other manufacturers (on the basis of hardware sales) on two fronts: establishing market trust in the ability to deliver quality goods (we mine with our own shovels, after all) and stability in supply because everyone knows we're not only transparent about how many we're keeping for ourselves, but we're also depending on our development of new technology for our own mining efforts, meaning that we are going to push the envelope for our own good (which translates to the good of the buyers as well.) This is in the same general spirit as the Ferrari model, where if global demand for a car is 10,000 per year, they make 9,999 of them and command top dollar.
You may be right, but I think you are overthinking it. For instance, the BTC exchange rate is of little importance, since the choice is between mining and selling - both of which are paid in BTC.
After you have a difficulty forecast, you easily estimate many BTC a particular device will generate over its lifetime, for example, 50 BTC. Then say someone is willing to buy the device for 55 BTC. What do you do? Sell it, of course - you'll make more money!
And you can simply repeat this procedure for every device until the highest of the two values is below the marginal cost of producing the miner - at that point production is no longer profitable.
Of course some of the profit needs to go to R&D for next generation technology, but even then the basic principle remains the same. And you have to be good at forecasting difficulty and always keep the forecast up-to-date. And you need to know how much people are willing to pay for the devices. I assume AM has a difficulty forecast, and with the recent auction, AM has also learned more about what people are willing to pay for the devices.
PS: Even if your mining earnings estimate for the device is 50 BTC and someone is willing to buy it at 50 BTC, you should sell - because having 50 BTC in your wallet.dat today is much better than maybe having 50 BTC at some point in the future (because you eliminate uncertainty and you may re-invest it the 50 BTC from the sale).
