DeepSeek is training a 2T-parameter model and plans to build an 8T-parameter one

DeepSeek CEO Liang Wenfeng told investors that using more domestic chips for AI training is now a major priority, with Huawei expected to begin deliveries as early as Q4. DeepSeek is currently training a 2-trillion-parameter model and plans to eventually build an 8-trillion-parameter model.

Wall St Engine on X: "DeepSeek CEO Liang Wenfeng told investors that using more domestic chips for AI training is now a major priority, with Huawei expected to begin deliveries as early as Q4. Note: DeepSeek is training a 2T-parameter model and plans to eventually build an 8T-parameter model." / X Wall St Engine on X: "DeepSeek CEO Liang Wenfeng told investors that using more domestic chips for AI training is now a major priority, with Huawei expected to begin deliveries as early as Q4. Note: DeepSeek is training a 2T-parameter model and plans to eventually build an 8T-parameter model." DeepSeek CEO Liang Wenfeng told investors that using more domestic chips for AI training is now a major priority, with Huawei expected to begin deliveries as early as Q4. Note: DeepSeek is training a 2T-parameter model and plans to eventually build an 8T-parameter model. DeepSeek CEO Liang Wenfeng told investors that using more domestic chips for AI training is now a major priority, with Huawei expected to begin deliveries as early as Q4. Note: DeepSeek is training a 2T-parameter model and plans to eventually build an 8T-parameter model. DeepSeek prioritizing domestic chips with Huawei deliveries as early as Q4, while training toward 2T then 8T parameters, is a supply-chain CapEx decision first. Export controls show up as wafer and interconnect lead times, not just model cards.