The fundamental solution is simple:
Don't compete on the basis of tax discounts.
It's obscene that some companies are offered "incentives" while local businessmen can't enjoy those same low tax rates. Fundamental fairness says the tax rate should be equitable across the board.
One major retailer built a warehouse near me. Now the locations for a warehouse serving retail stores is relatively limited -- figure a four hour driving radius so your drivers can make the journey back and forth in a day. Yet in my case three states competed by offering tax discounts to the corporation -- despite the fact the corporation logically had to locate in one of those states.
Part of the deal keeps the town from collecting full taxes on the facility for 20 years. Of course in 20 years that depreciation on the facility and especially the automated handling equipment inside will be substantial -- and the corporation will simply be shopping around for somewhere to build a new facility instead of dealing with the interruption of overhauling an existing plant while maintaining productivity there. Complete with new tax discounts, of course. The existing building, just as it gets to be taxed at full value will see the major corporation pull out and be replaced by a bunch of small businesses looking for low rent and having a fraction of the taxable assets that a corporation installing all new, highly automated machinery does.
It's a completely crazy way to operate, soak the local businessmen with the highest taxes, and provide the national corporations who will come and go with the biggest discounts.
>if their pile of servers in California becomes a problem, whether due to taxes, energy costs or
>just shifting user patterns, they can stick some or all of those containers on the back of trucks
I'm not sure that's the really compelling cost argument for them.
I think the advantage, really, is along the lines of pre-fab houses: It's far cheaper to build them in one central facility, then on site. You can have specialists at the "data center factory" who just specialize in assembling them, and then they're shipped to be assembled on site.
With an obsolesence of what, 5 or 6 years, why bother moving existing containers as-is? Break down / ship / setup labor costs will exceed tax costs in most cases. Instead as modules reach their functional obsolence, pull it out, send to the factory for a refurb with new servers, ship the refurbed unit to the new location. Within a few years you've accomplished shutting down the old site without the headaches of coordinating one big move.