Â鶹´å

America Lost the 5G Race. It’s About to Lose AI the Same Way.

Imagine this: An American-made technological wonder inspires economic excitement — while attracting fast-following Chinese competitors. US companies and policymakers are confident in the technical superiority of their wares, overlooking downmarket competition. "Good enough, cheap enough" Chinese options sweep emerging markets and the developing world, giving Beijing commercial and geopolitical power to write the global rules of the road for a critical strategic technology. Washington and its allies scramble to mount a diplomatic counter-offensive, but are too late, too bellicose, and still too expensive.

Is this a history of the 5G "race," or a prophecy of our AI future? The causes of Washington's 5G failures are clear — and looming over this next race today.

As 5G deployment heated up in the 2010s, Chinese companies like Huawei and ZTE challenged leading Western firms at a fraction of the price. US and partner governments raised security concerns, bolstered by in Chinese infrastructure projects. But while these warnings elicited sympathy, they did little for developing economies eager for progress to overcome such a pricing premium.

Soon even that sympathy ran out as Washington's diplomatic campaign began accumulating political baggage. Its reactive bans arrived too late — “stop using it” is a hard sell when “good enough” is already installed. US diplomacy started sounding like an overeager sales call for American vendors. And worse, the United States seemed to be sliding toward equating telecom equipment procurement with alignment to either an American or Chinese bloc, a weighty and unwelcome “choice” it had long avoided pushing countries to make.

As a result, much of Asia, Latin America, Africa, and the Middle East adopted the "good enough, cheap enough" Chinese option, and Beijing amassed vast influence over the global telecom ecosystem. 

Now, as artificial intelligence explodes as the next contest to shape the global technology order, we appear poised to repeat those mistakes. 

American frontier labs pioneered the modern AI marketplace and continue to lead the world in performance. But Chinese models like Qwen and Deepseek are close behind — and openly distributed for companies and users to download, customize, and deploy for free. They're, even prompting some Silicon Valley. As with the 5G race, "good enough, cheap enough" is hard to beat, whether you’re a budget-conscious emerging market or a buzzy tech startup. 

Of course, AI isn't 5G. The critical inputs are more diverse and diffuse across its tech stack, the market is more consumer-driven, and the competition more visible. But the key variables of the last tech competition remain: an economy-shaping technology, a price-sensitive world, an affordable Chinese alternative — and an American ecosystem prone to mistaking technical superiority for customer appeal and market inevitability.

Washington isn't blind to these parallels but seems unaware of how close it is to repeating them.

President Donald Trump's AI Action Plan calls for supporting free and open-source model development, but details of implementation have been scarce. American AI companies have plenty of incentive to dominate the premium market; what's missing is serious investment in affordable, open models that can compete on accessibility, not just capability.

Just as critical, however, will be updating our diplomatic playbook from the 5G era to pair strategic urgency with a lighter touch. Foggy Bottom has a strong security story to tell; even at this formative stage, Chinese models are already found to for users it believes come from politically marginalized groups. But, fairly or not, the State Department must avoid being seen as the marketing arm of the American tech industry, or yet again cajoling countries into a "final decision" between US and Chinese blocs.

That means helping France’s Mistral model and other open and non-US alternatives genuinely succeed in the AI marketplace, not as token additions, but as real options. And it means helping democratically-aligned companies supply the base models, cloud infrastructure, and fine-tuning tools that regional startups, universities, and systems integrators need to adapt AI to their own languages, cultures, and markets. Countries will be far more likely to adopt AI systems they helped shape — and Silicon Valley stands a better chance fending off Chinese models as a regional ecosystem partner than as simply a more expensive competitor.

The foundation for such a strategy is starting to emerge. Initiatives like the recently-announced could pair American compute resources with allied research institutions. Recent AI gesture toward building trusted supply chains. But these remain tentative steps when the moment demands a sprint. 

If Washington wants a different outcome than 5G, it needs to stop selling and start building.

Harry Krejsa was President Biden’s Assistant National Cyber Director for Strategy. John Costello was President Trump’s Deputy Assistant Secretary of Commerce for Intelligence and Security. Phoebe Benich was a Senior Program Manager for In-Q-Tel. They are currently researchers for the Carnegie Mellon Institute for Strategy & Technology (CMIST).