Until recently, the companies leading the artificial intelligence revolution seemed almost unstoppable. In pursuit of increasingly powerful AI systems, they have raised and spent extraordinary amounts of money—sums exceeding the annual economic output of many countries. But the path towards an AI-driven future is becoming considerably more difficult.
The industry now faces two powerful political challenges, either of which could slow its expansion. The first comes from the AI safety movement, whose warnings about the potentially catastrophic consequences of advanced artificial intelligence have gained considerable attention in Washington. The second is a rapidly growing public backlash against the construction of data centres across the United States.
The technology industry is far better equipped to confront the first challenge. The debate over AI safety largely emerged from within Silicon Valley itself, and technology executives have spent years thinking, arguing and writing about the risks posed by increasingly capable systems. Anthropic chief executive Dario Amodei’s recent manifesto of nearly 4,000 words illustrates how seriously industry leaders have engaged with these concerns.
The opposition to data centres presents a very different problem. It is less organised, less financially powerful and often less intellectually developed than the AI safety movement. Yet the public anger behind it is genuine and potentially much more difficult for Silicon Valley to manage.
There is a fundamental reason why this opposition matters. Artificial intelligence cannot advance on investment and computing innovation alone. It also requires enormous amounts of physical infrastructure. Data centres need land for the vast computing systems that train and operate AI models, while the electricity required to power those facilities demands still more land and energy infrastructure.
This exposes an important weakness in the industry’s ambitious plans. The extraordinary financial expectations surrounding artificial intelligence ultimately depend on physical resources located in communities where ordinary people possess political rights and considerable influence over how land is used.
And many of those communities increasingly do not want data centres in their neighbourhoods.
That resistance cannot simply be overcome by raising another round of investment or developing a more powerful AI model. Silicon Valley is accustomed to solving technological problems with capital, engineering and innovation. Community opposition is fundamentally different because it involves land, local politics and public consent.
The long-term challenge for the AI industry, therefore, is not merely to demonstrate that artificial intelligence can become more intelligent or economically valuable. It must convince the communities bearing the physical costs of its expansion that they will also share in its benefits.
Until data centres and the technologies they support produce tangible improvements in the lives of ordinary people, public resistance is likely to persist. The future of artificial intelligence may ultimately depend not only on what Silicon Valley can build, but also on whether the people living around that infrastructure believe it is worth building.
