Same Federal Reserve. Same interest rate. Very different outcomes depending on who is borrowing. Run your own numbers, then see what is happening 30 miles northwest of Indianapolis.
When inflation runs hot, the Fed raises rates so people and businesses borrow and spend less. That still works. But it does not work evenly. A first-time buyer feels a higher mortgage rate immediately. A company with $100 billion a year in operating cash, racing rivals for AI, can keep pouring concrete.
The economy has to slow by some total amount. If the biggest spenders don't slow down, the rest of us may have to slow down more.
Plug in your own loan below.
Bigger, relative to the economy, than the railroads, if projections hold.
Not forever. The risk moves to developers, utilities, and lenders.
Compare any mortgage against the 3% rates many buyers locked in a few years ago. Principal and interest only; taxes and insurance not included.
At 3%, that same monthly payment would have carried a loan of about $575,000. Same income, very different house.
In February 2026, Meta broke ground on a data center campus in Lebanon, the Boone County seat, valued at more than $10 billion. It is planned at about 1 gigawatt of capacity across 4 million square feet, with operations expected at the end of 2027 or early 2028.
Mortgage rates have climbed since then. The building did not stop.
Meta has also pledged to pay the full cost of the site's energy and water service, and to give $1 million a year for 20 years to a local fund that helps residents with energy bills. That raises an interesting question: why would a company help pay its neighbors' electric bills? Data centers increasingly need local consent, and consent has a price.
$10 billion would buy roughly 28,500 homes at $350,000 each, in cash, with no mortgage at all.
Lebanon is one of dozens of large sites Meta is building, inside a 2026 capital budget the company has put at well over $100 billion.
A family buying its first home in Boone County and a hyperscaler building in Boone County face the same Federal Reserve. They do not face the same choices.
Billions of dollars. Fiscal years differ, and these totals include non-AI and overseas spending. 2026 plans are far higher.
Average yearly investment as a share of GDP. The AI bar is a Brookings projection for 2025 to 2032, not money already spent.
Monthly principal and interest on a 30-year loan at different rates.
Annual payments on a 5-year loan. At 8.1%, the owner pays about $11,400 more a year than at 4%. That is a part-time employee, or the owner's own paycheck.
The Fed itself has started describing AI investment as unusually resistant to interest rates.
| Official | Date | What they said | In plain English |
|---|---|---|---|
| Michelle Bowman, Vice Chair for Supervision | Jan. 16, 2026 | AI and data-center projects are relatively insensitive to interest rates. | Rate hikes don't slow this spending much. |
| Lisa Cook, Governor | Feb. 24, 2026 | AI investment is soaring even though rates are elevated. | The usual brakes aren't biting here. |
| Lisa Cook, Governor | Sept. 28, 2026 | AI appears to be adding inflation pressure in the short run, with possible relief later. | Building it pushes prices up; using it may push them down. |
| Michael Barr, Governor | Sept. 29, 2026 | The buildout is having a measurable effect on prices. | This is now big enough to show up in inflation. |
| Philip Jefferson, Vice Chair | Oct. 1, 2026 | AI investment supports growth while raising costs in related goods and services. | Good for GDP, pricier for some inputs. |
Oracle spent $55.7 billion on equipment and buildings last fiscal year while generating $32 billion in operating cash, and borrowed heavily to cover the gap. Amazon's free cash flow fell sharply. The risk didn't disappear. It moved, onto developers, utilities, and private lenders.
Company totals include warehouses, overseas sites, and ordinary equipment replacement.
If productivity gains arrive, the economy can produce more without prices rising as fast. Building is inflationary; using may not be.
Home prices, too little supply, insurance, taxes, and zoning all matter more directly than data centers.
The $9 to $10 trillion buildout estimates assume that announced projects actually get built. Many announced projects never are.
Rate hikes still work. They just land harder on some borrowers than others, and that was true long before AI.