Chinese AI rivalry threatens US model margins, not its ecosystem

AT FIRST glance, some Chinese artificial intelligence models seemingly deliver almost state-of-the-art capabilities with few resources and at low cost. Some are also “open weight”, allowing users to download and run them on local networks, easing data privacy concerns.

So why would corporate users pay a premium for frontier US models?

As US technology companies continue to commit capital to AI’s infrastructure, investors question whether the sector can generate returns in line with its spending.

This misses the bigger picture. While Chinese open-weight models are likely to pressure the pricing power and profit margins of US model providers, they do not fundamentally threaten a broader US-led AI ecosystem.

Cheaper and better AI models should accelerate adoption, expanding demand across the value chain, where much of the long-term value creation is likely.

Are China’s cost advantages as stark as they seem?

The perception that China is building the same or better AI for a fraction of the price is misleading.

Many cost-advantage claims of Chinese models in capital expenditure terms underestimate the impact of “distillation”, or partly training models on more advanced systems. And a low-price-per-token Chinese model does not necessarily mean low cost per result once the AI model is used.

A less-capable model may need more tokens, computing power, energy and time to finish the same job.

Cost per completed task and time per task are more useful benchmarks. By these metrics, many US models compare favourably with Chinese rivals. Nor are US developers standing still. They have slashed prices, and some provide open-weight models, with more planned.

If cheap or locally deployed models become standard, would that undermine valuations across the US AI ecosystem? We do not think so, for three reasons.

First, lower prices should expand AI demand. This is a version of the Jevons paradox: When a technology becomes cheaper, people use it more. Lower cost per AI task should encourage more tasks, longer reasoning and more autonomous agents.

The US-led ecosystem could therefore continue to thrive because of its cloud infrastructure, global networks and proprietary data, none of which depend on which AI model is used.

Second, frontier models may still attract a premium and have an overall capability lead. Chinese models match them on some benchmarks, but there are gaps on the most demanding cyber, scientific and long-horizon agentic tests (complex, multi-step tasks that AI systems must complete autonomously over time).

For some applications, such as scientific research, the best models still generate an advantage, and hence some pricing power for the most advanced models.

Third, security and sovereignty concerns point to a hybrid market with a role for both US and Chinese models. US restrictions on foreign access to some models have highlighted the risks of relying on a single country’s technology.

Yet, “sovereign AI” initiatives outside the US are slowed by the cost of building competitive AI infrastructure from scratch. At the same time, large foreign AI and cloud service users may still prefer US to Chinese models because of governance and security concerns.

It is therefore plausible that AI demand evolves towards widespread, cheaper models, including open-weight systems deployed locally.

That would make it harder for frontier providers to sustain premium pricing, margins and valuations, but does not mean that users will switch exclusively to Chinese alternatives.

We expect both systems to coexist.

What the next leg of the race could look like

The US-led AI ecosystem has a significant advantage in specialised advanced accelerators, high-bandwidth memory, and the networking that allows chips to work as a single system. US export controls should help preserve these advantages in the near term.

China may remain constrained by a shortage of advanced computing capacity in the short term. It will continue building domestic chipmaking capabilities, while encouraging closer cooperation among its technology companies.

Still, China has a critical advantage in electricity production. In 2025, the country generated more than double the electricity produced by the US, and is expanding faster.

A Chinese data centre may take between one and three years to build and be supplied with power, compared with five to 10 years for a major US grid connection and transmission upgrade.

For now, however, the US has a sizeable lead in installed AI compute capacity, reflecting its hardware advantage. This currently more than offsets China’s greater electricity supply.

China’s power advantage therefore remains a source of strategic opportunity that could become decisive if the country closes the gap in chips and talent.

The AI cycle will continue

The US retains the lead in frontier-AI capabilities, while China is well placed to accelerate adoption. But neither country will accept the lead of the other.

We expect the two ecosystems to evolve in parallel, driving AI innovation and investment globally, with some volatility as one temporarily gains at the expense of the other. This will continue to generate economic momentum, with positive growth impacts in both countries.

For investors, value creation is likely to accrue to the owners of AI infrastructure, chips, data and distribution networks, rather than just to frontier-model developers.

Both writers are from Lombard Odier. Michael Strobaek is global chief investment officer and Dr Filippo Pallotti is macro strategist.

This article has been abridged from a longer original analysis by Lombard Odier. The commentary and information here does not constitute investment advice, nor is it a recommendation to buy or sell.