Nvidia plans to launch a lower-cost AI chip for China in June, aiming to protect market share under the US export controls and signal a broader shift toward affordable, segmented products that could impact global enterprise AI spending.
The new GPU, part of Nvidia’s Blackwell architecture, is expected to be priced between $6,500 and $8,000, significantly lower than the restricted H20 chip, which sold for $10,000 to $12,000, according to a Reuters report.
The price cut reflects reduced specifications and a more straightforward manufacturing process.
The chip is built on the RTX Pro 6000D server-class platform and will use standard GDDR7 memory instead of the more advanced high-bandwidth memory. It will also reportedly omit Taiwan Semiconductor Manufacturing Company’s (TSMC) Chip-on-Wafer-on-Substrate (CoWoS) packaging, a feature found in Nvidia’s high-end GPUs.
This comes as Nvidia CEO Jensen Huang recently said that the company’s market share in China has dropped to about 50%, down from 95% prior to the introduction of US export controls in 2022.
How enterprises stand to gain — or lose
Nvidia’s plan to release a lower-cost, region-specific AI chip for China may mark a broader strategic shift, not just a response to US export curbs. Importantly, the move presents both risks and opportunities for enterprises managing global AI infrastructure.
“Nvidia’s region-specific, compliance-driven chip strategy introduces manageable fragmentation risks, but also unlocks significant opportunities for global enterprises,” said Prabhu Ram, VP of the industry research group at Cybermedia Research. “While hardware and software inconsistencies may complicate unified AI deployments, these variants enable legal market access, cost optimization, and hybrid architecture flexibility across regions.”
Others see it as a potential catalyst for rethinking enterprise AI deployment strategies, especially in cost-sensitive markets.
“While this ‘relatively lower-cost but regionally compliant chip’ is designed with the Chinese market in mind, this sets a precedent and demand from enterprises for Nvidia to offer similar solutions that have tighter budgets,” said Neil Shah, partner & co-founder at Counterpoint Research. “This could be a blessing in disguise for some regions and enterprises where compute workloads could be very diverse or different and would welcome such customized, simpler solutions.”
According to Shah, Nvidia’s move could mirror moments seen in the Chinese AI market, where local players like DeepSeek have thrived by adapting under constraints.
“This could be sort of a ‘DeepSeek’ moment for Nvidia, driving innovative solutions under constraints. This will potentially expand Nvidia’s SAM (Serviceable Addressable Market) further and put pressure on Huawei and AMD,” Shah added.
To make the most of this evolving chip landscape, enterprises will need to adapt their strategies.
“By strategically segmenting workloads, leveraging software-agnostic frameworks, and building modular infrastructures, enterprises can mitigate fragmentation downsides and capitalize on Nvidia’s global-but-localized AI roadmap,” Ram added.
Nvidia’s move could also signal a broader move to serve budget-conscious global markets, particularly emerging economies and smaller AI startups that prioritize affordability over top-tier performance.
“This may drive adoption of modular AI strategies, where enterprises combine powerful GPUs for training with cost-effective chips for inference and edge tasks,” said Manish Rawat, semiconductor analyst at Techinsights. “A GPU priced under $8,000 could intensify pricing pressure across the AI hardware market, impacting Nvidia’s premium offerings and prompting rivals to adjust pricing, as businesses prioritize ROI on AI investments.”
However, the chip’s limitations, such as the use of GDDR7 and the absence of CoWoS, make it unsuitable for cutting-edge AI training, restricting its appeal to light enterprise AI and inference-focused applications.
Pressure mounts on AMD and Intel
Nvidia’s lower-cost Blackwell chip signals a broader push to serve diverse enterprise AI needs, potentially putting pricing pressure on rivals like AMD and Intel.
“This ecosystem advantage offers lower total cost of ownership (TCO), making even lower-spec Nvidia chips more attractive than switching to AMD or Intel platforms,” Rawat said. “If Nvidia expands this strategy beyond China, AMD and Intel may face pricing pressure and margin challenges, especially in cost-sensitive AI segments.”
The move adds to the competitive heat in the lower-power GPU segment, where both AMD and Intel have been gaining ground with enterprise buyers focused on budget, power efficiency, and regulatory compliance.
Finally, while these new China-specific Blackwell chips are unlikely to appeal to customers seeking high-end AI training performance outside the country, they could carve out a space in the mid-tier market.
The more accessible price point could attract enterprise buyers looking to enter the Nvidia ecosystem without the premium cost. “The most interesting aspect of this chip is seeing how Nvidia is being forced to create custom AI chips based on regulatory restrictions and working around these constraints to provide cost-efficient and marketable options,” said Hyoun Park, CEO and chief analyst at Amalgam Insights. “Nvidia does not want to give up the lucrative Chinese data center market in its role as the global leader in AI hardware, so it is trying to commit to building hardware that meets the letter of the law. Of course, this is a challenge when the letter of the law seems to change on a week-to-week basis.”
Source:: Network World