Key Takeaways
- DeepSeek’s forthcoming V4 model is engineered to operate on semiconductor technology developed by Huawei Technologies, per The Information’s reporting
- Major Chinese technology corporations—Alibaba, ByteDance, and Tencent—have secured bulk purchases of Huawei’s latest chips, with orders reaching several hundred thousand units
- The Hangzhou-based startup collaborated closely with Huawei and Cambricon Technologies to modify V4’s core architecture
- Industry sources anticipate V4’s public release within weeks, alongside two additional model variations currently under development
- In a departure from typical industry protocols, DeepSeek denied American semiconductor manufacturers advance access to V4
DeepSeek’s next-generation artificial intelligence model, designated V4, has been optimized to function on processors manufactured by Huawei Technologies, according to reporting from The Information released Friday.
The publication’s findings draw from five sources with intimate knowledge of the development process.
Anticipating V4’s imminent debut, China’s technology powerhouses—Alibaba, ByteDance, and Tencent—have secured substantial procurement contracts for Huawei’s newest semiconductor offerings. Combined orders exceed hundreds of thousands of individual processing units.
The artificial intelligence model is projected to become publicly available in the coming weeks.
DeepSeek, headquartered in Hangzhou, dedicated several months to intensive collaboration with both Huawei and domestic chip architect Cambricon Technologies. The partnership’s primary objective involved restructuring and validating critical segments of V4’s foundational code to ensure compatibility with domestically produced semiconductor technology.
Two individuals with proximity to the organization confirmed these collaborative efforts.
American Semiconductor Manufacturers Excluded from V4 Preview
Domestic chip suppliers, prominently including Huawei, received privileged early access to DeepSeek’s V4 architecture before its commercial launch. Conversely, American chipmakers were denied equivalent early testing opportunities, representing a significant deviation from established industry conventions preceding major model deployments.
Reuters previously documented this exclusionary approach earlier this year.
The artificial intelligence startup is simultaneously developing two supplementary V4 iterations. Each variant targets distinct functional capabilities while maintaining architecture designed specifically for Chinese-manufactured processing hardware.
Both DeepSeek and Huawei declined to provide statements to Reuters’ inquiries, which were submitted beyond standard business hours.
Market Impact of DeepSeek’s Prior Releases
DeepSeek’s previous model iterations, specifically V3 and R1, precipitated substantial declines in global technology equity valuations upon their respective launches. Investment professionals began reassessing whether American artificial intelligence corporations required multi-billion-dollar expenditures on computational infrastructure.
This market response positioned DeepSeek’s upcoming V4 model under heightened international observation.
The V4 architecture remains unreleased to the public. DeepSeek has not issued official confirmation regarding a specific launch timeline.
Cambricon Technologies’ stock valuation increased 2.67% following the announcement. Alibaba’s shares declined 1.36% during U.S. market trading and dropped 1.49% on the Hong Kong exchange.
