In a single week in May 2026, Taiwan formed a national AI strategy committee with binding government deadlines. Korea signed a cooperation agreement with six UN agencies to launch a Global AI Hub. Hong Kong announced “AI for All” — a commitment to reach over 500 schools — backed by a $3 billion subsidy scheme already in motion. Uttar Pradesh approved construction of a dedicated AI City in Lucknow, a 12-acre public-private project backed by state budget commitments already passed into law.
No press conference declared a race. There was no starting gun. It just started.
While all of this was happening, Europe was finalising the timeline for its AI compliance framework.
Asia is asking one question. Europe is asking a different one. America stopped asking.
The question Asia is asking is: how do we win?
Singapore’s answer is the clearest. The government committed more than $1 billion to public AI research and talent development from 2025 to 2030, stood up a National AI Council chaired by the Prime Minister, and launched national AI missions targeting sectors that together represent 40% of the country’s GDP. This is not a pilot programme. It is an economic strategy with named sectors, named timelines, and a named person at the top accountable for results.
Taiwan passed an AI Basic Act and immediately attached deadlines to it. Government agencies must complete AI risk assessments by July 2026. Relevant ministries must have industry regulations in place by January 2028. Civil servant AI competency guidelines arrive in June. The education ministry has been instructed to strengthen AI literacy in schools now — not as part of a future review, but as part of the current implementation plan.
Hong Kong is running the same playbook in parallel. The $3 billion AI Subsidy Scheme is already distributing. The “AI for All” initiative is a signed agreement with schools across 500 institutions. Government officials have stated the goal plainly: move AI from visibility to adoption, from pilot to deployment at scale.
The question driving each of these decisions is the same: what do we need to build to not be left behind?
Europe’s primary answer to the same moment is the AI Act — a risk classification and compliance framework that will be fully applicable in August 2026, with rules for high-risk AI systems extended until 2028. The regulation was already revised once because it was generating costs faster than it was enabling innovation. The EU acknowledged this directly when it moved to simplify the framework in late 2025, citing the need to reduce business costs and create more favourable operating conditions.
The question driving European policy is not how to win. It is how to make AI safe enough to allow.
America’s answer is different from both. The Trump administration’s AI Action Plan, released in July 2025, is built around a single directive: get out of the way. Remove regulatory barriers, accelerate infrastructure permitting, and let private capital move at the speed it wants to move. The federal government’s role is not to coordinate a national strategy but to clear the runway for the companies already building. With $285 billion in private AI investment in 2025 alone — more than 23 times China’s tracked figure — the bet is that the market doesn’t need a plan. It needs room.
Three questions. Three speeds. Three different theories of who should be in charge.
Asia believes governments should lead. Europe believes governments should govern. America believes governments should step aside.
These are not the same question. And they do not produce the same speed.
Europe has the talent. Someone else owns the companies.
This is not a simple story of failure. The numbers are more complicated than that.
Europe’s AI funding hit a record $21.8 billion in 2025. There are nearly 11,000 AI startups on the continent — almost twice as many as five years ago. Three of the world’s ten most-cited AI scholars are European. Monthly active LLM users in Europe nearly double those of the United States.
And yet almost every AI model those users run was built in America or China. Almost three-quarters of the investors backing Europe’s breakout AI companies are American. Europe grows the talent and the early-stage companies. Someone else owns them at scale.
European startups match their US counterparts at the early stage — but fall behind sharply at growth. As the Prosus and Dealroom State of AI in Europe report put it in March 2026: the gap isn’t in capability. It is in deployment at scale.
Australia’s financial regulator made the stakes explicit. The country risks producing a “lost generation” of citizens stuck with a lower standard of living if it fails to keep pace with global AI adoption. The warning was not about talent or research capacity. It was about system speed — the pace at which institutions move from awareness to implementation.
The cost argument that changes everything
Here is where the conversation usually goes wrong. The standard framing puts the race in terms of money, infrastructure, and research capacity. Under that framing, smaller countries and individual builders are structurally disadvantaged. They cannot compete with Singapore’s billion-dollar commitments or America’s $285 billion in private AI investment in 2025 alone.
But that framing is already out of date.
According to Stanford’s AI Index, the cost of running AI at GPT-3.5 performance levels dropped 280-fold between November 2022 and October 2024. Hardware costs are declining 30% annually. What cost millions to deploy three years ago can now be implemented by a small team with a modest budget.
This changes what the race is actually about. When the cost of capability collapses this fast, the scarce resource is no longer compute or capital. It is the willingness to implement — now, with what exists, before the conditions are perfect.
Finland illustrates the tension precisely. It has genuine technical depth: fourth in global government AI readiness rankings, a national AI research centre backed by €250 million in funding, world-class universities, and one of the most capable public digital infrastructures in Europe. Government spending runs at 57.5% of GDP — one of the heaviest public sectors on the continent. The research ecosystem is real.
And yet only 4% of Nordic companies report strong returns from AI investment, according to BCG research published in March 2026. Strategic preparedness across the region has dropped from 61% to 43% in a single year, and talent preparedness sits at just 14%, according to Deloitte’s State of AI in the Nordics 2026 report. The foundation is there. The implementation is not.
The bottleneck is not capability. It is the speed of the decision to act.
The individuals who don’t wait
Governments set the conditions. They do not determine who builds first.
Students from the American University of Phnom Penh did not wait for a national AI strategy before competing at the Huawei ICT Asia-Pacific competition in May 2026 — they placed second in the AI Innovation category, presenting AI-powered quality inspection systems for Cambodia’s garment sector. They identified a problem, built a solution, and entered a competition while their government was still drafting a framework.
This is the pattern that matters. While institutions debate whether conditions are right, the builders who move first are building on top of infrastructure that already exists — open models, falling API costs, global distribution — regardless of which country they are in.
The permission to act has never come from a government committee. The AI Act does not determine whether a founder in Helsinki or Tallinn or Lisbon ships something this month. The $1 billion Singapore committed to AI research does not flow to individual builders who are waiting for the right moment. America clearing regulatory runway for OpenAI does not clear it for you.
What determines whether someone builds is not the policy environment. It is the decision to stop waiting for it to be resolved.
When the cost of intelligence falls 280-fold in two years, the question is not whether your country has the right strategy. The question is what you are building while they figure it out.
If Taiwan can set a national AI deadline for July — what is stopping you from setting your own?
