DeepSeek's 'Price Butcher' Terrifies Zhipu and MiniMax

On April 24, DeepSeek released a preview version of V4, leading to a decline in the domestic AI large model sector—was this a concentrated release of the market's response to a "potential threat," or a genuine reshaping of the competitive landscape?

According to the news from the Chasing Wind Trading Desk, J.P. Morgan Securities (China) released a research report on April 26, offering a starkly different judgment on the market reaction: The release of V4 is an industry boon, not a zero-sum shock.

What did the market misunderstand behind the 9% plunge in stock prices?

On the day of the launch of the DeepSeek V4 preview version, the stock prices of Smart Spectrum and MiniMax both fell by 9%, while the Hang Seng Index rose by 0.2% on the same day. On April 27, after the market opened, MiniMax dropped another 12%, and Smart Spectrum fell by nearly 3%.

The market's logic was straightforward: DeepSeek has released a new model, intensifying competition and compressing the survival space for pure large language model (LLM) companies.

The analysts, however, believe this to be "an overreaction".

Their core judgment is that: The release of V4 actually bolsters the three key pillars supporting China's LLM industry—computing power supply release, pricing discipline, and structural cost curve compression. The only dimension where competition has intensified is the relative competitive positioning, but this is "fierce competition" rather than a "reshaping of the landscape".

More importantly, V4 had been the "biggest single competitive catalyst" on the event calendar followed by analysts for April- May. With V4 settling in the market and being digested, this uncertainty has been officially cleared.

Computing Power Bottleneck Loosens: This is V4's Greatest Industry Significance

Analysts pointed out that where the market diverges the most from their research framework in terms of "price and interpretation" is precisely in computing power.

DeepSeek V4 is compatible with Huawei's Ascend chips, successfully supporting the reasoning of cutting-edge models with 1.6 trillion parameters in scale. This is a key validation: domestic chips are technologically feasible and may even have cost competitiveness compared to international chips.

In the past, computing power constraints have always been the invisible ceiling for Chinese LLM companies— there is token demand, yet it is difficult to convert into recognizable revenue (ARR). DeepSeek has clearly stated that with the mass market launch of Ascend 950 supernodes in the second half of 2026, the reasoning cost of V4 Pro will further decrease.

The analysts believe that this is a direct benefit to Smart Spectrum and MiniMax: **the expansion of computing power supply means these companies will be able to offer services at a lower cost, thereby clearing the path from token demand to recognized revenue conversion**.

**We believe that this is a key benefit for the entire Chinese LLM industry, including Smart Spectrum and MiniMax, as it alleviates the constraints of computing power supply—one of the main barriers preventing these companies from fully expanding their operations. DeepSeek has stated that with the wider application and optimization of Ascend chips, the cost curve of reasoning will improve, opening up new paths for Smart Spectrum and MiniMax to convert token demand into recognized revenue**.