Moonshot's Kimi K3 spooks chip stocks, drags Bitcoin below $64K

Moonshot AI's Kimi K3, the world's largest open-weight model with 2.8 trillion parameters, triggered a selloff in semiconductor stocks and dragged Bitcoin below $64,000 on Thursday. The model outperformed several Western counterparts in third-party benchmarks, according to published evaluations, narrowing the performance gap with U.S. frontier systems such as Anthropic's Claude series and OpenAI's GPT family. The launch comes just weeks after U.S. authorities withdrew Anthropic's Fable and Mythos models over security concerns, creating a vacuum that Chinese AI labs have moved to fill.

"The scale and open-weight nature of Kimi K3 challenges the assumption that only closed-source U.S. models can deliver frontier performance," said a Hong Kong-based technology analyst at a global investment bank. "If Chinese labs can match Western capability with optimized GPU kernels, the premium chip demand narrative faces a structural challenge."

Kimi K3's 2.8 trillion parameters compare with OpenAI's GPT-4 at roughly 175 billion parameters, though parameter count alone does not determine capability. The model demonstrated advanced reasoning and coding performance in third-party tests, matching or exceeding leading U.S. models on multiple benchmarks. Moonshot also released API services and comprehensive developer documentation alongside the model, lowering adoption barriers for enterprises. The launch follows similar releases from Chinese AI labs Z.ai and MiniMax, which have accelerated release cycles to challenge Western dominance. These companies, backed by major stakeholders including Alibaba and Tencent, have pushed the frontier of AI capability with faster iteration speeds than their U.S. counterparts.

The open-weight nature of Kimi K3 marks a shift in the AI industry's competitive dynamics. Unlike closed-source models that restrict access to their underlying architecture, Kimi K3 allows developers to use, modify, and distribute the software without licensing fees. This approach could accelerate adoption across small and medium-sized enterprises that cannot afford the high costs of proprietary systems, potentially compressing margins for companies that rely on closed-source revenue models.

The selloff spread beyond AI stocks into crypto markets, with Bitcoin falling below $64,000 as risk assets correlated with tech weakness. The Federal Reserve's upcoming policy meeting added to the cautious tone, with traders pricing uncertainty around rate decisions that could further pressure growth-sensitive positions. Bitcoin's 24-hour trading volume surged as the decline triggered liquidations across leveraged positions, with the broader crypto market following the downturn. The correlation between AI-linked equities and digital assets has strengthened this year as institutional investors treat both as risk-on exposures.

For investors, the implications cut both ways. Semiconductor stocks face a potential repricing if Chinese AI labs can achieve frontier-level performance with fewer or less advanced chips, challenging the demand thesis that has driven the sector's rally. Nvidia, trading at roughly 35 times forward earnings, and AMD came under pressure as the market weighed whether Moonshot's GPU kernel optimizations could reduce reliance on premium hardware. Bitcoin's correlation to risk assets pulled it lower, with open interest data showing a sharp decline in long positions ahead of the Fed meeting. The episode shows how quickly AI model releases can ripple through correlated markets, from chip stocks to crypto, as the boundaries between technology sectors continue to blur.

This article is for informational purposes only and does not constitute investment advice.