The AI Data Centres Situation is Crazy

Summary

The video argues that the AI data-center boom rests on an uncertain assumption: that demand for AI will grow quickly enough to justify enormous spending on computing infrastructure. Data centers take years to build and require grid connections, land, transformers, cooling systems, water, and electricity, making these projects harder to scale back than software investments.

Infrastructure race

The video describes OpenAI, Oracle, and SoftBank’s Stargate plan, announced in 2025, to invest up to $500 billion in U.S. AI infrastructure and reach 10 gigawatts of computing capacity. It says Stargate and other large-scale plans have faced delays, changing requirements, or cancellations, including plans for additional capacity at the Abilene, Texas, site.

Demand and financial risk

The video questions whether AI revenue can support the planned investment. It cites OpenAI’s reported revenue and losses, ambitious future revenue targets, and projected computing costs. Even with optimistic growth, the company may remain reliant on substantial outside funding.

Nvidia and circular financing

Nvidia is presented as a major beneficiary of the buildout because of its dominant position in GPUs. The video examines investment and purchasing arrangements involving Nvidia, OpenAI, Oracle, and cloud provider CoreWeave. It argues that some demand may be reinforced by investments and purchase guarantees from companies that sell or finance the infrastructure, raising questions about whether that demand is independent and sustainable.

Cancellations and scaling back

The video cites Microsoft data-center lease cancellations and postponed projects, along with changes to OpenAI and Oracle plans. It also reports that community opposition has blocked or halted numerous proposed projects in the United States.

Local costs and promised benefits

Data centers can create construction jobs, but the video says far fewer permanent jobs may remain once facilities open. It also describes complaints about noise, water use for cooling, and increased electricity demand.

The video cites rising grid-capacity prices and higher power costs for some manufacturers. This raises the question of whether ordinary consumers and businesses may end up sharing the costs of infrastructure built for major technology firms.

AI’s limits in the workplace

The video points to the Commonwealth Bank of Australia’s reversal of planned customer-service job cuts as an example of the limits of automation. It also cites Gartner’s prediction that some companies that reduced support staff because of AI may later rehire for similar roles, particularly where customers need judgment or empathy.

Public and political reaction

The video describes protests across U.S. states, low reported public support for data centers in local communities, and New York’s temporary pause on permits for some large facilities. It concludes that the central risk may not be AI itself, but the assumption that demand for it is guaranteed.

Product segment

A sponsored segment describes Apex Guard VPN. The narrator says it replaces a user’s regular public IP address with the VPN server’s IP address, but does not make someone anonymous: websites may still identify logged-in users, and browser fingerprinting and accounts remain.

The segment distinguishes this from private browsing, which mainly controls what is stored locally after a session. It explains the product’s stated feature but is not a full review or tutorial.

Main speakers and sources

The primary voice is the Internet Education narrator. The subtitles also include clips or claims attributed to technology executives, data-center developers, energy companies, analysts, residents, protesters, and politicians.

Named organizations and sources include OpenAI, Nvidia, Microsoft, Oracle, CoreWeave, TD Cowen, Bloomberg, Reuters, Carbon Direct, Mordor Intelligence, and Gartner.