The borderless internet is dead. The global technology economy is splitting into US-led and China-led sovereign stacks.

I write from London. That matters. Silicon Valley sees one world — an American-led tech order that set the rules and won the day. Shenzhen sees another. London sits between them. It is home to both DeepSeek’s lab and DeepMind’s main offices. It doesn’t fool itself about being a tech power, which is why I think it can see the rift more clearly than most.

Here’s a number that should reshape your strategic plan: the United States holds 75% of global AI compute. China holds 15%. Everyone else is working with what’s left.

The borderless internet is dead. For thirty years, code and data crossed borders freely. That world is over. What’s replacing it are two closed systems — chips, models, cloud networks, and rules aligned with either Washington or Beijing. Any company still planning around one global tech setup is among the most exposed businesses on the planet.

The common view was that US export controls on chips would cap Chinese AI. The evidence says the opposite. DeepSeek built frontier-level models despite the restrictions. Zhipu trained a system comparable to the best Western models entirely on Huawei hardware. Export controls may not have slowed Chinese AI at all — they might have sped it up, by removing the option of buying from abroad and forcing home-grown capability instead. I’ve watched this play out and still find it striking.

Now consider which country might determine which of these two worlds wins. It isn’t the US or China. It could be India.

India is the world’s largest democracy, a big technology maker, and a country with deep ties to both Washington and Beijing. If India builds its AI networks around American systems, Western reach extends into South and Southeast Asia. If it builds its own stack — or tilts toward Chinese tools — the Global South may follow. Most Western strategists aren’t watching this closely enough. India is the swing state of the AI age. We’re not thinking hard enough about it.

For your business, the question is no longer whether the ecosystem will split. It already has. The question is: how do you run a company across two AI worlds with different data rules, different models, and different political risks?

Think through what that means in practice. A European pharmaceutical firm might need US compute and Chinese genomic data — under two different views of what patient data rights mean. A financial services firm building risk models must pick a stack, knowing that each choice carries a political bet. That’s a new kind of decision — one most boards haven’t faced before, and most strategy teams aren’t set up to make. We’ve been asked to act before we’ve worked out how.

And then there’s the wildcard. In Pakistan, households put up 22 gigawatts of rooftop solar in a single year — bypassing the state grid. In Iran, smuggled Starlink terminals kept communications open through shutdowns. When technology routes around the gatekeeper, the gatekeeper loses. The two-world model may get messier, not cleaner, as hardware gets cheaper and more people find their own way through.

Plan for two AI worlds. Build options. Understand that your technology choices are now political choices, whether you meant them to be or not.

Azeem explores this topic with leadership teams. Begin a conversation.