Nvidia is preparing to introduce a cut-price artificial-intelligence chip for mainland China, marking its third redesign since U.S. export curbs began in 2022.
The new processor, part of the company’s Blackwell generation, is being engineered to comply with Washington’s latest bandwidth caps while preserving enough horsepower for commercial data centers.
People familiar with the plan told Reuters the device should enter mass production in June, underscoring how quickly Nvidia is adapting its product roadmap.
Export Controls Force Another Pivot
A series of U.S. rules issued last year blocked sales of Nvidia’s flagship H20 GPU by limiting memory bandwidth to 1.7 terabytes per second.
That restriction removed a product that once generated tens of billions of dollars in annual revenue from a market representing 13% of the firm’s sales.
Nvidia has now stripped down the Blackwell chip’s memory subsystem and ditched chip-on-wafer packaging so that it sits just below the regulatory threshold.
An internal memo described the approach as the company’s “only viable path” to keep serving China without violating export licences.
Leaner Design Targets Mid-Range Buyers
The upcoming unit, expected to carry the B40 or 6000D label, will retail for between $6,500 and $8,000, far cheaper than the $10,000–$12,000 H20.
Using standard GDDR7 memory and simpler substrates reduces production costs and should allow Chinese cloud providers to fill racks without waiting months for supply.
“Until we settle on a new product design and receive approval from the US government, we are effectively foreclosed from China’s $50 billion data center market,” an Nvidia spokesperson told Reuters.
The company is also quietly designing a second compliant chip for September, signaling a rolling update cadence tailored to the evolving sanctions regime.
Huawei Competition Grows
Nvidia’s market share in Chinese AI accelerators has fallen from roughly 95% before 2022 to about 50% as customers test domestic alternatives.
Huawei’s Ascend 910B and upcoming 910D parts are gaining traction among state-backed research institutes and cloud vendors hungry for independence.
“Domestic Chinese technologies like Huawei are expected to catch up with the computing performance of downgraded versions within one to two years,” said Nori Chiou, investment director at White Oak Capital Partners.
Analysts warn that a fully competitive Huawei software stack could erode Nvidia’s long-held advantage in machine-learning frameworks.
Price and Production Timeline
Production partners in Taiwan say the Blackwell variant will roll off assembly lines in June with volume shipments penciled for early July.
Initial orders reportedly cover tens of thousands of units, enough to seed major Chinese data-center operators ahead of a wider launch.
Earnings in the Spotlight
The chip designer reports first-quarter numbers on 28 May, and Wall Street projects revenue will surge 66% year-on-year to $43.4 billion.
Even so, Nvidia shares slipped 3% last week as investors assessed the risk of fresh U.S. sanctions blocking the new chip before it reaches customers.
“We see upside … despite the loss of H20 sales to China,” Oppenheimer analysts wrote in a preview note, citing runaway demand from Western hyperscalers.
Strategic Significance
Founder and chief executive Jensen Huang conceded that China sales have halved in two years but insisted the market remains strategically indispensable.
“We will continue to make tremendous efforts to optimize compliant products and continue serving the Chinese market,” he said during a recent television interview.
Nvidia believes its CUDA software ecosystem and years of developer lock-in will help it defend share even with throttled hardware.
Rivals such as AMD and Huawei still lack broad support from machine-learning researchers who have spent the past decade targeting Nvidia architectures.
Outlook
If Washington approves the pared-down Blackwell chip, Nvidia stands to preserve billions in annual revenue and buy time to negotiate lighter regulations.
Failure to secure a licence would accelerate Chinese substitution efforts and could mark a permanent fragmentation of the global AI supply chain.