Throughout history there have been moments when humanity realised that competition alone wasn’t enough.
Sometimes the stakes become so high that cooperation becomes the only rational strategy.
Artificial intelligence is rapidly becoming one of those moments.
Today, the world’s most valuable companies are spending hundreds of billions of dollars trying to build increasingly capable AI systems.
Governments are doing the same.
Nations talk openly about an AI race.
Companies celebrate “winning.”
Investors reward whoever reaches the next milestone first.
We’ve turned intelligence itself into a geopolitical competition.
That should worry all of us.
Every race has a finish line
Competition has produced extraordinary innovation.
No one should deny that.
Without competition, we might not have today’s remarkable AI systems.
But races are temporary.
Eventually someone wins.
Or everyone crashes trying.
As AI becomes more capable, more autonomous and more deeply integrated into society, the cost of racing continues to increase.
More compute.
More energy.
More data centres.
More secrecy.
More pressure to release before competitors.
More incentives to prioritise capability over caution.
The race changes behaviour.
Even good people begin making decisions they would never make if they weren’t afraid of falling behind.
This isn’t because companies are irresponsible.
It’s because races reward speed.
Not wisdom.
Imagine if we raced to build the internet
Imagine if, in the 1980s, every country had decided to build its own incompatible version of the internet.
Different protocols.
Different standards.
Different infrastructure.
Different rules.
The modern internet would never have emerged.
Instead, researchers agreed on shared standards.
Companies competed on top of those standards.
Innovation accelerated because everyone started from the same foundation.
The web became bigger than any one organisation.
Artificial intelligence deserves the same ambition.
The world’s resources are already here
One of the strange things about today’s AI race is that the resources already exist.
The talent exists.
The hardware exists.
The universities exist.
The funding exists.
The research exists.
The energy exists.
What’s missing isn’t capability.
It’s coordination.
Instead of pooling resources, humanity duplicates them.
Multiple companies buy similar GPU clusters.
Multiple organisations clean similar datasets.
Multiple research teams solve nearly identical problems.
Each organisation independently spends billions reaching milestones another reached months earlier.
This is an astonishing duplication of effort.
Not because competition is always wasteful.
But because intelligence is becoming infrastructure.
Infrastructure should be shared.
The greatest scientific collaboration in history
Imagine announcing a different kind of project.
Every major AI company contributes engineers.
Every major university contributes researchers.
Governments contribute funding and compute.
Cloud providers contribute infrastructure.
Open-source communities contribute innovation.
National laboratories contribute expertise.
Instead of ten frontier models…
there is one.
One AI.
Continuously improved.
Continuously audited.
Continuously benchmarked.
Continuously defended.
The largest scientific collaboration humanity has ever attempted.
Not because competition failed.
Because success requires cooperation.
Companies still win
One criticism appears almost immediately.
“What happens to AI companies?”
The answer is simple.
They keep building.
Google didn’t disappear because Linux exists.
Cloud companies didn’t disappear because TCP/IP became universal.
Browser companies still competed after the web became open.
Companies create enormous value by building products.
They don’t need to own the foundation beneath civilisation.
In fact, shared foundations often create larger markets.
When everyone builds on common infrastructure, innovation moves faster.
Not slower.
Nations still compete
Another concern is geopolitical.
“What if another country refuses?”
That question assumes cooperation requires universal agreement from the beginning.
History suggests otherwise.
The internet expanded because participation became advantageous.
Scientific collaborations grow because joining produces more value than remaining outside.
Open-source software spreads because it solves problems better.
One AI could follow the same path.
The countries that participate first gain access to the world’s largest collaborative research effort.
Others eventually join because exclusion becomes more expensive than cooperation.
Competition shifts from building secret foundation models to building the best applications on top of a shared one.
A shared immune system
There is another reason to build One AI.
Every powerful technology eventually attracts misuse.
Cybercrime.
Espionage.
Fraud.
Disinformation.
Autonomous attacks.
The stronger AI becomes, the stronger the world’s defensive systems must become.
Instead of every company defending itself independently, imagine humanity maintaining one continuously improving defensive intelligence.
Every new attack strengthens the model.
Every vulnerability becomes shared knowledge.
Every security improvement benefits everyone.
The same collaborative model that transformed cybersecurity could transform AI safety.
One AI becomes humanity’s immune system.
Not owned by one company.
Shared by all.
This is not centralisation
Some readers hear “One AI” and imagine one giant computer.
One authority.
One government.
One point of failure.
That isn’t the proposal.
The internet is one network built from millions of independent computers.
Linux has thousands of contributors.
Wikipedia has millions of editors.
Scientific knowledge belongs to everyone while being maintained by no single organisation.
One AI follows the same principle.
One shared foundation.
Millions of independent participants.
Thousands of competing applications.
Decentralised innovation built on common infrastructure.
The cost of doing nothing
Perhaps the biggest mistake would be assuming today’s AI race simply continues forever.
It won’t.
Either the race ends intentionally…
or it ends because one system becomes so powerful that everyone else is forced to follow it.
Those futures look very different.
One is governed.
The other is dictated.
History rarely gives us the luxury of choosing after the fact.
The project our generation will be remembered for
Every generation inherits one defining challenge.
Previous generations rebuilt continents after war.
Mapped the human genome.
Connected the world through the internet.
Explored space.
Our generation may be remembered for something even more fundamental.
Not for building artificial intelligence.
But for deciding who it belongs to.
Future historians probably won’t care which company briefly led benchmark rankings in 2026.
They’ll care whether humanity recognised that intelligence itself had become too important to remain fragmented.
The original Manhattan Project demonstrated what humanity could accomplish when it concentrated enormous scientific resources on a single goal.
Its achievement changed history.
Its consequences still shape the world.
The project we need today should pursue the opposite objective.
Not to build the most powerful weapon.
But to ensure the most powerful intelligence humanity has ever created becomes humanity’s greatest shared achievement.
The AI race has already shown us what competition can accomplish.
Perhaps the next breakthrough won’t come from running faster.
Perhaps it will come from deciding that, from this point onward, we run together.
The question isn’t whether AI will change the world. The question is whether the world’s most powerful intelligence will belong to a handful of companies, or to humanity itself.
3/3
First dropped: | Last modified: July 31, 2026