How Nvidia’s B30A Chip Impacts US-China Trade Tensions

Nvidia is reportedly working on a new AI chip tailored specifically for the Chinese market, as tensions between Washington and Beijing over access to AI technology continue to dominate trade relations between the world’s two largest economies.

The company is developing the B30A, a processor based on its Blackwell architecture that would significantly outperform the H20 model currently approved for sale in China, according to two sources speaking to Reuters.

This latest development follows US President Donald Trump’s recent suggestion that he might allow Nvidia to sell more advanced semiconductors to Chinese customers. 

Yet industry insiders caution that winning regulatory approval remains far from certain, given the mistrust in Washington over sharing sensitive AI technology with Beijing.

Inside B30A’s design, purpose and impact 

The B30A is a careful balancing act for Nvidia. 

The chip employs a single-die design – essentially cramming all the processor’s components onto one piece of silicon rather than spreading them across multiple chips. 

While this approach will deliver roughly half the raw computing power of Nvidia’s top-tier B300 accelerator, it should still pack considerably more punch than the current H20.

Nvidia hopes to get samples of the new chip into the hands of Chinese clients for testing as soon as next month, though final specifications are still being ironed out.

“We evaluate a variety of products for our roadmap, so that we can be prepared to compete to the extent that governments allow,” the company says in a statement.

“Everything we offer is with the full approval of the applicable authorities and designed solely for beneficial commercial use.”

The new processor will include high-bandwidth memory alongside Nvidia’s NVLink technology – the company’s proprietary system that allows multiple chips to work together seamlessly during intensive AI computations. 

These same features appear in the H20, which runs on Nvidia’s older Hopper architecture rather than the newer Blackwell design.

Despite its single-die limitation, the B30A should deliver a substantial performance boost over existing options available to Chinese firms. 

The chip will handle both AI training – the process of teaching models to recognise patterns in data – and inference tasks, where those trained models make real-world predictions and decisions.

This versatility matters in a market where Chinese tech companies are racing to develop AI applications spanning everything from autonomous vehicles to language translation systems.

Nvidia’s processors also benefit from years of software development. 

The company’s CUDA programming platform has become the de facto standard for AI development, making it costly and time-consuming for engineers to switch to rival hardware platforms.

What does B30A and Nvidia’s development mean for the US? 

There have been many negotiations between Nvidia and the Whitehouse regarding regulation between the two AI powerhouses.

When asked about Nvidia’s China ambitions during a recent CNBC interview, Commerce Secretary Howard Lutnick says: “Of course Jensen would like to sell a new chip to China,” referring to Nvidia’s CEO Jensen Huang. 

“I’m sure he’s pitching the president all the time.”

He adds that he has witnessed the CEO’s presentations to Trump firsthand: “I’ve listened to him pitch the president – and the president listens to our great technology companies – and he’ll decide how he wants to play it.

“But the fact Jensen is pitching a new chip shouldn’t surprise anybody.”

China accounted for 13% of Nvidia’s revenue last year, making it too important a market to abandon entirely. 

Yet lawmakers from both parties worry that even scaled-back chip sales could help China close the AI gap with the US.

Trump has floated the idea of allowing a “30% to 50% off” version of Nvidia’s latest chips into China – a reference to reduced computing power. 

Yet he dismisses the current H20 as “obsolete.”

The mounting global regulatory and competitive challenges 

The political market remains treacherous for chip companies trying to serve Chinese customers. 

Congressional leaders have voiced concerns that any access to advanced semiconductors, even neutered versions, undermines America’s technological advantage.

Nvidia counters that staying engaged with Chinese developers prevents them from abandoning its ecosystem entirely in favour of domestic alternatives like Huawei Technologies. 

The Chinese telecoms equipment maker has made impressive strides in processor development, with its latest chips reportedly matching Nvidia’s raw computing power in some areas, though they still lag in software support.

Yet the situation has grown even more complex – as the Chinese state media have suggested Nvidia’s processors could pose security risks, while government officials have quietly discouraged local tech firms from buying the H20. 

Nvidia says its chips carry no backdoor risks.

Meanwhile, Nvidia is preparing a second China-focused chip called the RTX6000D. 

This lower-specification processor, designed primarily for less demanding inference work, will cost less than the H20 and uses conventional memory to stay within regulatory limits.

Its memory bandwidth of 1,398 gigabytes per second sits just under the threshold that triggers export restrictions.

“Nvidia plans to deliver small batches of RTX6000D to Chinese clients in September,” says one person to Reuters who is familiar with the timeline.