Elasticity:

Uruguay’s software industry is strange.

The country has a population of around 3.5 million people, yet its technology sector generated US$3.681 billion in revenue in 2024. Around 63% of that revenue came from exports, with the United States as its main destination. Uruguay is now the second-largest software exporter per capita in Latin America.

That means Uruguay’s software advantage has never been about scale. It has been about productive capacity.

The interesting question now is whether artificial intelligence can increase that capacity without requiring Uruguay to grow its software developer population at the same rate. If it can, AI could do something particularly important for a small economy:

increase the amount of economic value it can produce with the capabilities it already has.

But that outcome is not automatic.

For a long time, people often discussed economic development in terms of accumulation.

• More capital.
• More workers.
• More infrastructure.
• More education.
• More technology.

But modern development economics introduced a more complex idea: what matters is not only how many resources an economy possesses, but what it can do with them.

This is particularly visible in the work of Ricardo Hausmann and colleagues on productive capabilities and the Product Space. Countries do not become more sophisticated simply by accumulating more inputs. They develop capabilities that allow them to move into increasingly complex products and services.

That distinction matters enormously for AI. AI gives almost every company access to additional productive capacity. It does not give every company the same capability to turn that capacity into value.

Uruguay is not starting from zero.

Its software industry has spent decades accumulating capabilities around engineering, export services, product development, international collaboration, and technology infrastructure.

The results show up in the numbers.

In 2024, Uruguay’s ICT sector generated US$2.323 billion in exports, representing 63% of total sector sales. Seventy-six percent of companies reported having international clients, and the United States was the sector’s principal destination.

This changes the AI question. Uruguay does not necessarily need to become a country with more developers than its competitors. It needs to become a country where each unit of software talent can generate more sophisticated and valuable output.

That is an elasticity question.

The question becomes:

How much additional economic value can an economy generate from each additional unit of AI-enabled capacity?

This is where economic history gives us an important warning. Technology does not automatically translate into economic development.

The Solow growth tradition showed that technological progress changes the productivity of capital and labor, while endogenous growth theory, associated with economists such as Paul Romer, emphasized the importance of knowledge, ideas, and human capital in generating sustained growth.

But technology needs complementary capabilities:

• People need to know how to use it.
• Organizations need to redesign processes around it.
• Markets need to reward new forms of production.
• Institutions need to adapt.
• And businesses need access to customers willing to pay for the resulting value.

AI is no different. Giving a software company access to an AI model is not equivalent to giving that company a competitive advantage.

The advantage comes from the interaction between AI and everything else the company already knows how to do.

Uruguay’s National AI Strategy 2024-2030 is interesting precisely because it does not frame AI simply as a technology-adoption program.

Its objectives include governance, national capabilities and sustainable development. The capabilities axis explicitly includes talent and digital skills, infrastructure, data protection and cybersecurity. The development axis positions AI as a potential driver of economic growth, private-sector competitiveness, digital transformation, research and innovation.

That is important. A country trying to benefit from AI doesn’t just need companies using AI. It needs an ecosystem that can absorb AI. This brings us back to the first two articles in this series. At the organizational level, we asked:

How much AI-enabled capacity can a company absorb?

• At the individual level, we asked:

How much AI-generated work can a developer cognitively absorb?

• Now we can ask the economic version:

How much AI-enabled capacity can an economy convert into international value?

Suppose Uruguay suddenly gives every software developer access to powerful AI tools. The country’s theoretical software capacity increases. But that does not mean exports automatically increase by the same proportion. The additional capacity might instead create new bottlenecks.

Companies may struggle to find international customers. Product management may become the constraint. Senior technical leadership may become scarce. Quality assurance may lag behind development speed. Companies may produce more software but not more valuable software. Or they may use AI simply to deliver existing services more cheaply, without moving into more sophisticated activities. In that scenario, AI increases capacity without sufficiently increasing economic value.

Elasticity falls. The alternative is more interesting.

• What if AI allows Uruguayan companies to move from staff augmentation toward higher-value product development?
• What if small teams can compete for projects that previously required much larger engineering organizations?
• What if existing export relationships become channels for selling AI-enabled products and services?
• What if AI helps local companies enter more complex parts of the global software value chain?

Now the country is not merely producing software faster. It is increasing the complexity of what it can produce and export. That is a development story.

This distinction matters for Uruguay’s relationship with the United States.

Nearshore software has traditionally offered US companies advantages around geography, time zones, talent, and cost.

AI changes the competitive equation. If AI makes software production cheaper everywhere, cost becomes a weaker differentiator. The strategic advantage has to move somewhere else:

• Capabilities.
• Domain knowledge.
• Product thinking.
• Engineering quality.
• Trust.
• Communication.
• Speed of experimentation.
• The ability to integrate AI into complex organizations.

In other words, AI can make basic software capacity less scarce while making sophisticated capabilities more valuable.

For Uruguay, that could be an opportunity.

The country already has an export-oriented technology sector with deep exposure to international markets. Its next advantage may come from increasing the value generated by that existing capability rather than simply producing more of the same.

This is perhaps the most important implication. Public policy cannot manufacture a successful software company. It cannot guarantee a US contract. But it can influence the conditions under which additional technological capacity becomes productive. Uruguay’s AI strategy explicitly seeks to build national capabilities, strengthen infrastructure and skills, promote innovation, and improve private-sector competitiveness.

That means policy can potentially affect the slope of the elasticity curve. 

• Better talent increases absorption capacity.
• Better infrastructure increases technical capacity.
• Better data increases AI capability.
• Clearer governance reduces uncertainty.
• Research and innovation increase the possibility of moving into new products and services.
• International connections increase the market available for the resulting capabilities.

None of these guarantees growth. Together, however, they can change the conditions under which AI generates economic value.

There is an interesting paradox here. Large economies have an obvious advantage in AI: more capital, more companies, more researchers, and larger domestic markets. Small economies cannot compete with that scale directly. But they may move faster.

A small ecosystem can coordinate policy, academia, companies and talent more tightly.

Uruguay’s own AI strategy was developed through a participatory process involving public institutions, private organizations, academia, civil society and other stakeholders.

That kind of coordination becomes more valuable when technological change is fast. The opportunity for a small economy is therefore not necessarily to have more AI. It is to have less friction between AI and the capabilities already present in the economy.

Across this series, we have looked at three different constraints.

First, the organization. Adding AI-enabled capacity eventually reaches a point where more output creates more complexity than value.

Second, the developer. AI can accelerate production faster than human attention, judgment, and understanding can absorb that acceleration.

Third, the economy. A country can increase technological capacity without necessarily increasing the economic value generated from that capacity.

The same principle appears at all three levels.

Capacity is not value.

What matters is the relationship between them. And that relationship can be measured.

This changes the question companies should be asking.

Not:

How much AI can we deploy?

Not even:

How much faster can our developers work?

But:

Where is our AI elasticity highest?

• That question leads to better decisions.
• Where should AI replace repetitive work?
• Where should it augment expert judgment?
• Where should humans remain firmly in control?
• Where should we add more AI capacity?
• Where should we stop?

And, perhaps most importantly:

What capabilities do we need to build to move the elasticity point?

That is the difference between adopting AI and developing an AI strategy. For a developer, it means learning how to direct and evaluate AI. For a company, it means redesigning workflows around the new capacity. For an economy, it means building the capabilities, institutions, and international connections that allow AI to translate into higher-value production. 

The countries, companies, and teams that understand this will not necessarily be the ones with the most AI. They will be the ones that can extract the most value from the next unit of AI capacity.

That is AI elasticity.

What is happening in Uruguay is a useful real-world example of the opportunity created by AI-driven elasticity.

Praxis, a technology community focused on a post-AI future, recently selected Uruguay for a planned city-scale development within Colonia, announcing plans to mobilize up to $1 billion in investment over the next three years.

Beyond the project itself, the signal is important: as AI changes where knowledge, capital and highly skilled talent can create value, countries with a strong technology base, institutional stability and access to global markets can become magnets for new forms of technological activity. Uruguay is increasingly positioning itself not only as a place where software can be built, but as a place where the people, companies and ideas shaping the next phase of the AI economy can converge.