The Economic & Financial Impact
This is Part II of a 3 part series on data centers. See Part I
Why The Economics Matters
The economics and financing matter because this represents the use of scarce resources, scarce resources that God intends us to steward wisely. My argument has never been with those who claim we are to be wise stewards of God’s Creation. My argument is how to achieve this worthy goal. A voluntary cooperative free market using freely formed prices is the best way to allocate scarce resources and thus be a wise steward on behalf of the Lord.
The Economic and Financial Dimensions
It seems clear that given all the data centers being built that they are sucking up an enormous amount of resources. It does not seem to matter that these are costly build outs or whether the companies building them are not close to being profitable. As the costs of these centers rise it does not seem that the builders of them care about the financial expenditures involved. The question arises as to whether these rising costs are a genuine market price signal or whether they are a facet of the fact that these companies are getting cheap, made-up money that acts as a subsidy. In short, is this a monetary bubble like the housing market up until 2007?
To start, let’s review one of the insights of one of the greatest economists, Frédéric Bastiat. Bastiat, in his magnificent That Which Is Seen and That Which Is Not Seen counsels us to look past what we see and consider what we do not see. What is seen is the data center campus with the obligatory ribbon cutting officiated by local potentates. Also seen are the construction jobs created by the building project. What is unseen is the thing that might have been built instead of the data center. Perhaps it would have been an iron smelter or a steel mill. Maybe it would have been an auto manufacturing facility. It may have been much more mundane: warehouses or small business parks, or really any imaginable use of the capital that was used on a data center. It may very well have been that this marginal capital expenditure was because the data center was economically advantaged over other uses.
The insight of economists like Ludwig von Mises is that prices only function properly when they reflect the real scarcity that exists in a market, and therefore the real preferences of market participants. Every tax break, abatement, tariff exemption, grant or “incentive” program warps this price signal thereby distorting useful information.
Also, in an inflationary (as in money creation) environment such as exists in the U.S. today, we must consider the so-called Cantillon Effect which sees benefits of money creation accrue to the first receivers of the newly created money. Whoever gets this newly created money sooner will be able to bid for resources at the older (pre-inflation) prices, giving them a decided advantage. Clearly large data center builders with strong balance sheets and private equity partners are obviously toward the front of the line.
This money creation sets in motion the business cycle. This artificially cheap money and credit draws resources into capital-intensive, long-horizon projects thereby creating a boom. The viability of these projects is predicated on these rates staying low, which is unlikely since they were set artificially low when the project began.
The Obvious Subsidies
Prior to the pause (discussed last week) in many state’s data center incentive programs at least 38 states had offered incentives. These were primarily sales and use tax exemptions on equipment. These also include property tax abatements. In the work Money Lost to the Cloud (2016) it was estimated that 11 deals averaged $1.95 million per job. The largest was a center built by Apple in North Carlina that came in at $6.4 million/job.
Virginia passed up $1.94 billion in sales tax revenue to create 1,610 jobs in fiscal year 2025. This works out to be $1.2 million per job. The state estimates losing 52 cents for each subsidy dollar handed out.
There is a counter claim here, however. The Virginia Department of Taxation does claim that this subsidy produces an 11% annual return. Also, a study claimed $6.10 of labor income per subsidy dollar. Even if this “accounting” is correct, it is still a gross distortion of prices which results in an irrational allocation of scarce resources.
Texas lost about $1 billion in revenue due to incentives with no disclosure of jobs created or even where these projects are located.
In the town of Alabama New York, Stream U.S. Data Centers put together a $19.4 billion project at the STAMP industrial park, with $1.46 billion in subsidies from Genesee County that would garner 125 permanent jobs at a whopping $11.7 million per job, which may be the most on record nationally for a data center. The spokesperson of the Genesee County Economic Development Center (GCEDC) Jim Krencik admitted that without the subsidy they would not be there. The Investigative Post calculated that an employee would need to work for 122 years to pay back the $11.7 million it cost to create their job. It is worth mentioning that the GCEDC stands to collect $140 million in fees from Stream in this deal. A payoff to the entity granting the subsidy is a clear conflict of interest.
At the federal level the so-called One Big Beautiful Bill Act (OBBBA) put together a series of incentives that subsidize the creation of data centers. This occurred via a couple of means.
§168(k) of the Internal Revenue Code (IRC) was amended by OBBBA. This restored a 100% bonus depreciation for qualified properties, after January 19, 2025. This now extends this tax treatment to data centers. This means an immediate full deduction on servers, as well as HVAC and electrical systems.
§163(j) of the IRC was amended by OBBBA as well. This permanently ended depreciation and amortization from a firm’s adjustable taxable income. This represents a material loosening of the interest-deduction cap for those builders using leverage.
Now it is true that these tax breaks do apply to any business in the country that meets this criteria, not just data centers. However, the benefits of these changes accrue disproportionately to the most capital-intensive and most heavily leveraged industries. This is clearly encouraging the movement of capital toward this sector even if data centers are not mentioned.
A clear direct subsidy is “Rural Opportunity Zones”. This offers a 30% step-up after five years, triple the standard 10%. This has the effect of offering a significant tax incentive to build data centers.
Now a free market supporter might argue that this is just allowing businesses to keep more of their own money, so what could be wrong with that? As a general principle that is fine. Yet, these tax breaks are not a generalized reduction of the tax burden. These are selective tax reductions dependent upon the production of goods and services that the government deems valuable, not the market. It is clearly a case of “you can keep your money if you do as we tell you”. This is antithetical to a cooperative free market. This results in more of what the government wants produced and less of what the market wants produced, which as stated above remains unseen.
All of these together create a serious use of government power to push investment toward data centers.
The Not So Obvious Subsidies
As was pointed out last week on average data center impacts on electric rates tend to drop them. This is because more usage across an existing built out infrastructure lowers the average cost for other ratepayers. However, this is not always the case.
In many instances large data center builders are getting public utility ratepayers to subsidize the operation of data centers. One way this happens is the creation of confidential “special contracts”. These are designed to be priced below cost with the difference made up by ratepayers.
Another way this occurs is that the Federal Energy Regulatory Commission will allocate transmission costs between utilities. They then get divided up among ratepayers and in the process costs get shifted to other ratepayers.
Yet another means of achieving this cost shift is to overbuild infrastructure in response to inflated load forecasts. Developers will often shop the same project to multiple utilities, creating a sort of “phantom load” phenomenon. In the end, this encourages utilities to build real capacity that if not used by data centers must be recovered from their current customers.
It is worth noting that 23 states have approved “large-load tariffs” as of May of this year, with 7 more states pending. This is intended to shift costs back to data center owners. However, a review by Harvard indicated that all of this is impossible to verify.
One counterweight to all this is the increasing prevalence of states insisting or, in the case of my home state of Nebraska, allowing data center builders to secure their own power generation. This is a healthy development as it internalizes the costs needed to build out these facilities and makes the calculation as to whether to build them a bit more rational.
Then there is the case of monetary subsidies. It is true that rates are higher than a few years ago before the Fed had to start dealing with the price inflation it caused. Yet, the capital commitments that were made by the data center developers were done back when rates were lower. It is during this time period that all the business assumptions were made. Also, the accumulated monetary base does not just disappear because rates have risen. Most crucially, one must keep in mind that given central bank monetary manipulation rates will almost always be lower than they would be otherwise, thus distorting the capital flows in the economy.
Some things to consider. According to Morgan Stanley global data center investment is at a pace for $2.9 trillion investment level by 2028. Yet, there is only $1.4 trillion in data center operating cash flow. This leaves a $1.5 trillion financing gap.
There is likely a bubble that has formed in this sector of the economy. According to the Bank for International Settlements (BIS) hyperscalers corporate bond issuance went over $100 billion in 2025. These bonds carry a maturity of 5+ years for the most part. This is a potential problem as these maturities may be longer than the economic lifespan of the servers themselves, which means the asset stops earning money before the debt matures.
Also, much of the debt of these data center buildouts is off of corporate balance sheets, according to the Bank for International Settlements. This is typically done via a “special purpose vehicle”. According to a report the “Big Five” technology companies held $969 billion of undiscounted future lease commitments at the end of last year. $662 billion of these lease commitments have not yet started and therefore remain unrecognized on these corporate balance sheets.
All of this off the book, and unrecognized debt will make for a nasty surprise to many shareholders when the bills come due. This is especially true if these projects remain unprofitable.
Much of the financing of these projects is provided by vendors and chip makers. The BIS rightly calls this “circular financing” and recognizes this as a structural worry.
None of this is fraud. This is what a credit driven unsustainable boom looks like from the inside. Debt drifts out of sight off the company balance sheet, collateral starts being valued at projected (usually overly optimistic) outcomes and vendor financing becomes a loop between chipmakers, model developers and hyperscalers. These are the visible signs of a capital structure that is being warped beyond what actual voluntary savings support. In short, a classic case of the Austrian Business Cycle Theory boom—to be followed by the inevitable bust. The important thing to understand here is that all this is a government sponsored monetary subsidy courtesy of massive monetary infusions by the Federal Reserve.
Then there is a whole array of regulatory and property rights subsidies to consider. One is the use of eminent domain to force property owners to sell in order to create a data center. The federal government has offered selective expedited regulatory permitting. Many libertarians get confused here because they see a company being allowed to build. Yet, they are forgetting that other companies are not allowed to build because they are politically disfavored.
Along with this is preferential access to public power. The New York Power Authority (NYPA) offers low-cost hydro power allocations driven by political, not market, considerations.
This mishmash of subsidies, overt and hidden as well as the increasingly politically driven allocation of capital on top of a massively inflationary policy by the Federal Reserve has created serious distortions in the data center space. So much so that it is nearly impossible to tell how overbuilt this market really is. This is the true scandal of the data center phenomenon.
Tying It All Together
First, environmentalists have a point. They are correct in their view that something is wrong here. However, they are aiming at the wrong target. As was shown in last week’s post the environmental concerns are really minimal and largely overblown. The truly illegitimate aspect of the data center boom is fiscal, regulatory and monetary.
Second, the industry is engaging in blatantly self-serving behavior to maintain their subsidies. They claim, correctly, that their footprint is small, like an office building or industrial plant. They also claim that without the incentives the projects cannot move forward. If the burden is small then they do not need a subsidy. If the subsidy is needed then the project is not profitable using real prices.
Third, the correct policy is evenhanded and boring. It amounts to “don’t just do something, sit there”. All that needs to happen is to let the cooperative market function. Offer no tax abatements, no exemptions, no special electric contracts, no NYPA politically driven allocations, and no eminent domain. On the other hand, do not impose moratoriums, bans, or punishing taxes. Data centers should face the same set of calculations as any other business project. Have an organization buy the land, pay a market rate for power (privately produced power would help here), internalize all business expenses, and then take your chances in the marketplace. This stance will offend both the right and left, which is a good barometer that it is the correct one.
The question, in the end, was never whether the country could afford the water and the electricity. The question is whether any industry that needs $11.7 million per job in taxpayer money to get started has in any way proven that consumers really want what they are producing. Profits and losses tell us that. All we know so far is that legislators want this product or are willing to use force to prevent anyone from having it.
What this all adds up to is a needless, and sinful waste of scarce resources. Wantonly using coercion to affect a politically driven outcome regarding the use of scarce resources is antithetical to God’s plan for His children. It is also a practical failure. Peaceable, cooperative relationships are always the principled and practical way to live out the Gospel.
Also, at odds with a faithful following of God’s Word is the needless fear of data centers specifically, and technology in general. As we shall see in Part III, technology is not coming for our future, but fear will if we let it.
Praise Be to God
