Microsoft's Brad Smith sounds alarm on China's state-backed AI push

Microsoft’s Brand Smith is well-known for his diplomatic voice as the executive has spent years navigating Washington’s security instincts and Silicon Valley’s global business.

That’s why his latest warning landed with unusual weight: US tech companies, he said, should “worry a little bit” about China’s state-backed AI subsidies.

In a market that often treats subsidies as background noise, Smith is arguing that they are the strategy, and that the telecom playbook could be repeating itself in AI.

AI’s Huawei moment

Smith’s core analogy was blunt: China has run this movie before.

In an interview with CNBC at the AI Impact Summit in New Delhi, he said subsidies were “the core strategy” China used to disrupt telecommunications, helping firms like Huawei and ZTE expand with state support.

The result, he warned, was brutal for competitors: “Some American companies vanished,” and European firms like Ericsson and Nokia were “thrown on the defensive.”

He tied that lesson directly to AI infrastructure.

Smith pointed out that data centers from Chinese companies such as Huawei and Alibaba already exist around the world, and argued “it will not be difficult for China to subsidize those.”

His broader message was not that the US has lost the technology lead, as he explicitly acknowledged the US still has an advantage in access to the world’s most powerful chips and other innovations, but that subsidized competition can change global market structure over time.

Also Read: Fed minutes expose a blind spot at the heart of the AI boom

The real battlefield

Smith’s warning is as geopolitical as it is corporate.

He argued that cheaper, subsidized Chinese AI offerings could be especially attractive to developing countries, where budgets are tighter and “good enough” systems can scale fast.

That matters because AI adoption is not just about models; it’s about which ecosystems, standards, and vendor relationships become embedded in government services, telecom networks, education systems, and local businesses.

The timing is not accidental.

Reuters recently described how DeepSeek’s meteoric rise in early 2025 pushed low-cost, open-source models to the forefront of China’s AI industry, setting the stage for a broader wave of low-cost model releases.

If those models spread quickly through emerging markets, helped by state support on compute, energy, and deployment, US firms could face a pricing and distribution challenge that looks less like Silicon Valley competition and more like industrial policy competition.

What Washington needs to do

Smith’s comments carried an implicit policy ask as private capital and private pricing discipline are not designed to beat a state subsidy program.

In the same CNBC discussion, he said the rest of the industry must “compete with that” and do so “with the support of our governments.”

That support can take different forms, export controls, incentives for domestic buildout, and allied coordination on standards and supply chains, but the thrust is that the US can’t assume a purely market-led path will hold its advantage.​