On May 28, 2026, Google announced a $15 billion infrastructure investment in New Florence, Missouri, a town of roughly 1,200 residents 75 miles west of St. Louis. The hyperscaler will build a data center campus that, per the company, will create thousands of construction jobs and hundreds of permanent operational roles.
Google is not borrowing to build. It is not syndicating a construction loan. It is deploying $15 billion of its own balance sheet into a single rural Missouri site. That is a capital allocation decision, not a real estate development deal.
The project sits within a broader $20 million Energy Impact Fund Google established to subsidize household electricity bills in surrounding counties. The fund is a political concession, not a charitable gesture. Missouri's Senate Bill 4, signed into law in 2025 by Governor Kehoe, requires hyperscalers to pay 100% of the power they consume and any directly attributable infrastructure costs. Google agreed.
To date, Google has contracted over one gigawatt of new generation capacity in Missouri. Through its partnership with Ameren, the company is supporting an additional 500 megawatts. That is 1.5 GW of new capacity tied to a single corporate tenant in a state that, prior to this, had no hyperscale data center presence of this magnitude.
The Laborers and Contractors Training Center in Eastern Missouri will train more than 2,300 construction laborers, including 1,500 apprentices, over the next two years. Google is funding the training pipeline directly. The company is not waiting for the local labor market to adjust; it is manufacturing the workforce.
This is not a speculative development. Google is not building a shell and leasing it to a third-party operator. It is building its own infrastructure for its own cloud and AI workloads. The capital is patient. The timeline is measured in decades, not quarters. The return is not a cap rate; it is the marginal cost of compute.
For institutional investors and REIT analysts tracking data center REITs like Equinix, Digital Realty, and CyrusOne, the implication is clear: hyperscaler self-build is accelerating. When Google, Amazon, or Microsoft choose to build their own campuses rather than lease from a REIT, they compress the addressable market for third-party data center operators. The REIT thesis depends on hyperscalers preferring OpEx over CapEx. That thesis is under pressure.
The Missouri project is one of dozens. Google has announced over $50 billion in U.S. data center investments since 2024. Amazon has committed $150 billion over the next decade. Microsoft is on a similar trajectory. The aggregate capital flow into hyperscale infrastructure is now larger than the entire U.S. office market annual transaction volume.
This is a structural shift in how capital is allocated in commercial real estate. Traditional CRE—office, retail, multifamily—is constrained by interest rates, construction costs, and lender caution. Hyperscale data center investment is not. It is funded by corporate balance sheets with AAA credit ratings and zero dependency on the CMBS market.
The $15 billion Missouri bet is a signal. The hyperscalers are not waiting for the CRE cycle to turn. They are building their own cycle. For lenders, developers, and investors still focused on the office recovery, the question is not whether demand exists. It is whether they are positioned to serve the only tenant class that is growing at double digits.