This offers a viable alternative for enterprises operating in or sourcing from China, especially amid tightening US export controls.
Huawei is reportedly preparing to ramp up shipments of its new 910C AI chip to Chinese customers as early as next month, a move that could accelerate the fragmentation of global AI infrastructure and challenge US chip dominance in enterprise workloads.
Some of the chips have already been shipped, according to a Reuters report, as Chinese AI companies scramble for domestic alternatives to Nvidia’s H20—a chip that had, until recently, been freely available in the region.
The shift comes after the US government tightened export controls, requiring Nvidia to obtain a license before selling the H20 to Chinese firms.
The move follows the launch of several advanced generative AI models in China, including the low-power DeepSeek, which has raised questions about the need for the massive computing resources required by rivals such as OpenAI.
For global enterprises, especially those with AI development centers or supply chains in China, the rise of a parallel AI hardware ecosystem could raise new concerns around interoperability, compliance, and long-term strategic planning.
Growing competition for Nvidia
As China ramps up its domestic AI chip development, Huawei’s 910C enters the market as a strategic alternative, though analysts point out that it still trails far behind Nvidia’s latest processors in raw performance and efficiency.
“From a performance standpoint, Nvidia’s new-generation chips — such as the B200 and the upcoming B300 Ultra, based on TSMC’s 4nm process and equipped with advanced HBM3/3E memory — have significantly widened the gap compared to Huawei’s 910C, which is likely built on SMIC’s N+2 7nm process (effectively 14nm) and lacks advanced HBM memory,” said Neil Shah, partner and co-founder at Counterpoint Research.
Shah noted that while Huawei’s chip may theoretically match the older Nvidia A100 or H100 in some tasks, it would require more power and heavy software optimization to handle diverse AI workloads.
“Global adoption of Huawei’s 910C is also hindered by limited developer support, ecosystem maturity, and integration challenges,” said Manish Rawat, semiconductor analyst at Techinsights. “However, it presents a viable alternative to Nvidia’s chips for Chinese enterprises or those affected by geopolitical constraints, especially as US export controls limit access to advanced Nvidia GPUs.”
Implications on enterprise AI adoption
Despite not being the best in the market, the 910C could still prove viable for many enterprise- and hyperscale-AI use cases.
It may take longer to train models compared to US-designed chips, but for many, it’s an acceptable trade-off given current geopolitical and supply chain risks.
“For enterprises – particularly those operating in or sourcing from China – it presents a credible alternative in the face of tightening US export controls, while supporting domestic innovation ecosystems like DeepSeek and reducing dependence on foreign technologies,” said Prabhu Ram, VP of the industry research group at Cybermedia Research. “Although Nvidia maintains an edge in software maturity and energy efficiency, Huawei’s progress reflects China’s growing strength in competing at the forefront of AI hardware in the emerging AI era.”
If companies operating in or sourcing from China adopt Huawei’s ecosystem, it could significantly influence their procurement strategies, vendor selection, and technology evaluations, driving greater alignment with Chinese technologies.
“To adapt, firms with Chinese operations may use dual technology stacks, incorporating both Western tools and Chinese platforms like Huawei’s Ascend chips,” Rawat said. “While this adds complexity, it provides a buffer against geopolitical disruptions and may offer quicker deployment compared to limited Nvidia supplies. For Chinese companies, Huawei chips enhance resilience, supporting local innovation and reducing dependency on Western suppliers.”
If Huawei chips gain traction, they could disrupt global supply chains, especially in telecom, AI, and cloud sectors. However, geopolitical tensions, security concerns, and regulatory challenges in Western markets could slow adoption.
How will China proceed?
As Huawei ramps up shipments of its 910C AI chip, the spotlight is on whether China’s domestic ecosystem can support scalable AI infrastructure without access to leading-edge global technologies.
“Scalability is something we are monitoring as SMIC N+2 7nm process has relatively lower yield, but things have been improving and should be self-sufficient for China domestic enterprises and hyperscalers for the next generation,” Shah said.
The bigger challenge lies ahead: advancing to cutting-edge nodes.
“The kicker is, how will Huawei evolve to more advanced nodes such as 5nm or below?” Shah said. “The only fab in China, SMIC, doesn’t have access to the latest EUV machines from ASML needed to advance to these nodes. That’s going to be a major challenge — one that Huawei and its partners are trying to solve, but not anytime soon and not without considerable sophistication.”
Still, China appears to be playing the long game. Developments like DeepSeek suggest that a fully localized AI stack may eventually emerge, with contributions from players like Huawei for compute, CXMT for memory, and a growing domestic base of software expertise. For enterprises, this raises important questions about long-term interoperability, performance trade-offs, and vendor diversification.
