Part 3 of 13

AI, China, and America’s Future

By Randy Salars · Source-review edition: September 7, 2026

I want America to help shape artificial intelligence and remain able to supply and use it. Depending heavily on other countries for essential computing, chips, or software could limit our choices. Leadership also means making the technology useful, reliable, affordable, and secure. Counting buildings alone cannot tell us whether we are succeeding.

“Whoever wins AI wins everything” captures the stakes people feel. It goes further than the evidence supports. This competition has no agreed finish line, and today’s strongest model does not confer permanent control of the future. Our practical goal should be to preserve the ability of the United States and its partners to develop the technology, obtain essential supplies, and put it to work on terms we can accept.

One claim we can put aside is that China competes without large data centers. In August 2024, China’s government reported that its eight national computing hubs contained more than 1.95 million data-center racks and had received over 43.5 billion yuan in direct investment by the end of June. Those are official Chinese figures, covering a broad computing program rather than AI alone. They nevertheless make the existence of major infrastructure investment clear. China’s announcement about its national computing hubs.

China’s effort to connect demand in its eastern regions with computing in inland regions also offers a useful comparison for our conversation. The physical location of computing and the location of its users can differ considerably. A rural host community needs to understand what it supplies and what it receives. The mere fact that customers live somewhere else does not make a project undesirable, but it makes the local agreement especially important.

The competition is substantial. Stanford’s 2026 AI Index reports that leading American and Chinese models were close on prominent public performance rankings through March 2026. The same chapter describes shortcomings in the tests used to measure progress. These rankings offer evidence about particular tasks and user preferences; they are not a universal measurement of intelligence, national strength, or performance in every workplace. Stanford’s technical-performance findings.

A model may excel at one task and disappoint on another. Dependable results are a better guide to progress: better products, productive workers, useful discoveries, secure systems, and businesses able to compete. Additional computing capacity matters when it helps deliver those outcomes.

There are concrete benefits worth considering. Research involving 5,172 customer-support workers found that access to an AI assistant increased issues resolved per hour by 15 percent on average. Gains varied, with less experienced workers benefiting more. This demonstrates a useful result in one setting; it does not establish that every worker, industry, or national economy receives a 15 percent improvement. The researchers’ summary, published by Stanford’s Digital Economy Lab.

Scientific computing offers another example. The AlphaFold database makes more than 200 million predicted protein structures available to researchers. A prediction is not a proven treatment or a completed experiment, but it can provide researchers with something valuable to investigate. This illustrates why access to advanced computing can matter beyond consumer entertainment. EMBL-EBI’s description of the AlphaFold database.

Those are benefits worth pursuing. American researchers need tools for discovery, businesses need affordable access, and workers need opportunities to learn. A local business using AI well can benefit without owning a server. A community hosting thousands of servers might receive less benefit than expected if hiring, training, and public revenue commitments are weak. Hosting the equipment and using its capabilities are distinct opportunities.

Grant County could ask whether a project agreement would support instruction for local students, workers, and small businesses. That is an idea to negotiate, not a benefit established by the proposal. Useful training should lead to identifiable skills and opportunities, with a named provider and funding attached.

The hardware supply chain adds another layer. Stanford’s 2026 report identifies Taiwan’s TSMC as the manufacturer behind almost every leading AI chip. That dependence makes it difficult to discuss American computing as though every essential piece originates within our borders. Reliable partners and a broader supply base deserve a place in the definition of national strength. Stanford’s research and development findings.

Electricity, networking, skilled people, and good software must work together. Buildings without customers or usable equipment cannot deliver the benefits of an operating service. Hospitals, factories, and county offices also need dependable tools that fit their work. Those practical demands belong in any discussion of American leadership.

Security belongs in this discussion because useful capabilities can be misused. The 2026 International AI Safety Report describes growing evidence of AI assistance in cyber operations, alongside continuing problems with reliability and evaluation. Its findings support taking both capability and safeguards seriously. They do not establish that simply building a domestic facility makes its systems secure. The report’s extended summary.

Leadership should include the ability to build responsibly and maintain public trust. Incomplete information and unexpected public costs can undermine support for worthwhile development. Clear agreements and credible local benefits give infrastructure a stronger foundation for operating over decades.

For Grant County, the national case creates a reason to consider the proposal carefully. It does not supply the missing details. We still need to know what Site Layer 1 would do, who would use it, and what obligations its owners would accept. A general claim about American competitiveness cannot identify a customer, guarantee a job, or establish an available water supply.

I want America to remain strong in AI, and I want Grant County to make a sound decision. A project that advances useful computing while protecting its neighbors and paying its way could serve both goals. Whether this proposal meets that description is something its documents and commitments must show.

What a useful answer would include

The developer and intended operator should explain what AI work the site would support and provide evidence of customer commitments, with appropriate protection for confidential contract details. A proposed local training program should identify its provider, budget, participants, and measurable outcomes. Schools and employers should help assess whether it meets local needs.

Questions worth asking

• What evidence establishes this facility’s intended contribution to AI computing, beyond its proposed generating capacity?
• Which customers or uses are committed, and which remain business-development goals?
• Would the developer fund a local training program, and who would deliver and evaluate it?
• Who would operate the facility and be responsible for its security and local commitments?
• What concrete benefits for Grant County accompany the claims about American competitiveness?

AI assisted most of the research and initial drafting under my direction. Sources are linked; corrections are welcome.