It’s easier than ever to build software due to reduced cost and fewer barriers to entry. As a result, there’s a renewed emphasis on companies establishing ‘moats’ early on. I suspect the real ‘winners’ in the AI revolution will be those whose moats centre on infrastructure rather than applied use cases—regardless of business model. This doesn’t strictly mean building the foundation models themselves; this is highly impractical due to resource constraints. Rather, companies like ElevenLabs and Hugging Face are building strong moats despite relying on third-party frontier technology.
The thinking surrounding moats can be applied on a macro level as well. The “AI arms race”—equally artificial itself—suggests countries are increasingly seeking technical advantages to separate themselves from their counterparts. This signals a shift in policy and funding where countries’ successes are becoming closely tied to their progress regarding AI. For some countries, this manifests as increased investments in AI through innovation subsidies, R&D funding, education and so on. For others, it involves building new data centres, accelerating hardware manufacturing or forging strategic partnerships with other countries.
However, not all countries need to—or even should—develop AI infrastructure in the same way, but should be considering their respective moats based on their unique strengths. In an ideal world, countries would cooperate under Ricardian principles and specialise based on their natural advantages and existing capabilities. Such an approach could narrow the development gap, boost global adoption and shared prosperity, whilst paving the way for crucial technology shaped by universally agreed upon principles. Essentially: global governance enabled by technological diplomacy.
This is unlikely to occur through a formal treaty built on goodwill and shared ambitions, but practical cooperation is already happening to some extent. In other words, international trade activity centred on different infrastructural components is creating natural specialisations without formal coordination.
The foundations are already visible in current global patterns. Taiwan's semiconductor expertise drives AI hardware worldwide, whilst Iceland harnesses geothermal energy for computational infrastructure. Morocco develops massive solar installations that could power global data centres, and the United States leads in AI research and talent development. None of this is grand strategy but economic logic playing out naturally; the real potential lies in accelerating these trends deliberately. Instead of duplicating efforts across every nation, cooperative specialisation could maximise global AI capacity whilst ensuring benefits reach everyone. This transforms AI from a zero-sum competition into a collaborative endeavour that leverages each region's natural advantages whilst embedding shared values and governance principles from the outset.
An ‘allied intelligence’ framework ensures true collective, rather than individual, prosperity. This approach would be neither negligent in favour of rapid innovation, nor bureaucratic under the guise of regulation. The ongoing ‘AI boom’ has brought the Fourth Industrial Revolution to an inflection point, and the next decade promises even more changes. Perhaps now is the time for a cooperative approach to the technologies which may continue to transform our lives.