The race for AI dominance isn’t just about innovation—it’s about control. Whoever holds the keys to the most powerful computing systems will shape the future. Right now, the U.S. has a lead, but that lead is fragile. Without smart, aggressive policies, we risk losing our edge to nations that don’t share our values. The stakes couldn’t be higher, and the time to act is running out.
The Department of Commerce’s new Framework for Artificial Intelligence Diffusion is a critical step in the right direction. This rule, introduced in early 2025, sets strict export controls on advanced AI chips and model weights, ensuring that cutting-edge technology stays where it belongs—in secure, allied hands.
The Three-Tier Approach
The framework sorts countries into three tiers based on risk:
- Close allies with minimal restrictions
- Most nations with moderate limits
- Adversarial states facing tight controls
This structure is smart, but it could be stronger. The U.S. still leads in AI development, but competitors are catching up fast. Chinese labs, like DeepSeek, have made impressive strides by using chips stockpiled before export rules took effect. Their progress proves why strict controls are non-negotiable. If we don’t lock down our tech, we’ll lose our advantage.
Why Computing Power Matters
Here’s why this matters. Training the most advanced AI systems demands enormous computing power. America dominates in semiconductor tech, and export controls ensure we stay ahead. Every two years, computing power doubles, meaning our lead grows while others struggle to keep up.
By 2027, countries stuck with outdated chips could face AI training costs ten times higher than those with the latest U.S. hardware. That’s not just an economic edge—it’s a strategic one.
DeepSeek’s struggles show these controls work. Their engineers admit chip shortages are their biggest hurdle, forcing them to use two to four times more power for the same results. Without access to top-tier U.S. chips, they’re stuck with inferior alternatives. This inefficiency slows their progress, giving us more time to innovate.
Learning from Past Mistakes
But we can’t get complacent. If we loosen restrictions, AI infrastructure could shift overseas, just like what happened with solar panels and batteries. Decades ago, the U.S. made up 40% of global semiconductor production. Today, it’s just 12%, with nearly all advanced chips made abroad. We can’t repeat that mistake with AI.
The Diffusion Framework helps by requiring domestic compute investments and capping foreign access, ensuring the next wave of AI infrastructure is built here, by American companies. Supply shortages and global export rules create a powerful incentive: if other nations want top-tier computing power, they’ll have to work with us.
Closing the Smuggling Loopholes
Smuggling remains a serious threat. Chinese firms have already set up elaborate schemes to bypass controls, hiding chips in prosthetic baby bumps and shipping GPUs alongside seafood. They’re also creating shell companies in third countries to dodge restrictions. We need tougher measures to shut these loopholes.
Key Recommendations
Our recommendations focus on three key upgrades to the framework:
- Adjust the tier system. Some Tier 2 nations have strong data security—let them access more chips through government agreements that prevent smuggling and ensure compliance.
- Lower the no-license threshold for Tier 2. Right now, they can buy $40 million worth of advanced chips without oversight, making it easy for smugglers to exploit. Reducing this limit would force more transactions under review.
- Boost enforcement funding. Rules mean nothing without the resources to enforce them. More support for the Bureau of Industry and Security would make these controls far more effective.
The Time to Act Is Now
Delaying implementation is dangerous. Chinese firms are stockpiling chips before the May 2025 deadline. Any pause would just give them more time to hoard, weakening the rule’s impact. The moment to strengthen our export controls is now.
By refining the Diffusion Framework, we can ensure transformative AI is developed here, under our values. Our leadership in AI depends on maintaining our computing advantage—and that starts with decisive action today.