I read the article in the Minneapolis Star Tribune on tax breaks for data centers the other day and was moved to respond. This is an area in which I am particularly qualified to comment, as I was an economic development program administrator forty years ago, and later became a software entrepreneur. I know a lot about location decisions, economic policy and data centers. In my economic development role, I made recommendations on numerous property tax abatement requests, and formulated plans for tax increment financing programs in downtown areas.

There are quite a few points to make here. You are abating sales taxes on the equipment purchased when the data center is getting built as well as on electricity and other purchases incurred during the operation of the data centers. Let me fill you in on a few things.

1. The computers and networking equipment that is installed in data centers is generally not purchased in Minnesota. It is purchased direct from the manufacturer and shipped here. As such it may be subject to use tax, but I suspect your use tax collection rates are pretty low. Businesses are not generally known for being honest on their use tax returns.
   
2. Computers and related equipment are capital expenses, as such their cost does not impact a businesses’ profitability in the year they are purchased. Instead, they are depreciated, which means their cost is spread over five years. That means when you purchase a million dollars of computer equipment, you spread an average of $200k of cost over the subsequent five years. As a result, a 10% reduction in the cost as a result of a tax break does not have a lot of bottom line impact, and is not going to drive anybody’s decision making. Sure there’s a small impact on cash flow but the big data center operators are rather flush with cash.
   
3. On the other hand, electricity is an operating expense, and a projection of its cost and availability over a period of years is going to be an important part of the siting calculation. The same is true for other operating costs, which is probably going to include the cost of bandwidth and the cost of water for cooling. Data centers do use a lot of water. Internet bandwidth is the most critical factor, though. A data center is pretty useless if the data that gets processed there can’t get out to users. This is why the initial data center industry grew up around the main internet connection points, where the big bandwidth providers had their interconnects.
   
4. One does not need many employees to operate a data center, once it is built and put into operation. Perhaps a security guard and a small staff of technicians to replace components that fail. I can manage multiple data centers around the world from my office in Bloomington. The instrumentation is highly refined, and failure prediction and identification was one of the earliest AI applications. If there’s a problem that requires on-site response, a service request ticket generally gets opened automatically and routed to the team that can address it.
   
5. All other things being equal, Minnesota should look like an attractive place to build data centers because
   • There appears to be a reliable supply of electricity at reasonable cost
   • The risk of geophysical and weather events that could take a data center off line is relatively low compared to other locations in the U.S.
   • Geopolitical risk is lower in the U.S. than in other countries.
   • Its centrally located. It may come as a surprise, but distance from the server to the device interacting with the server matters for many processes. Long distance data transmission causes enough latency in data transmission to reduce quality of service.
6. As a result of the above, the benefit of data center siting may not be the direct employment impact of the data center but the ancillary benefits of having related businesses establishing facilities nearby; however this is something that you need to do some analysis on to determine if it has really happened in the years since the tax break was initially established.
   
7. There’s some risk that the proliferation of data centers as a result of the AI binge is going to drive up the cost of electricity for all users, as well as make it more difficult to reduce reliance on fossil fuel based sources of energy. Beware unintended consequences!

This is all to say that tax breaks for data centers may not be such a smart public policy. You can’t definitively say that they are responsible for location decisions and you can’t definitively say that having them will provide employment opportunities without increasing operating costs for other Minnesota businesses.

When I was in the business of giving away tax base in exchange for business location, it surely seemed like a zero sum game. The surrounding localities and states were all giving away tax base, and it seemed like a competition of who could give away more, which created a true race to the bottom.

I would suggest that if you are going to get into industrial policy you take a broader view of the types of business development you want to promote, and create a set of programs and incentives that build on the state’s strengths and address any negative perceptions you encounter to achieve a set of positive outcomes for the state.