While these announcements were exciting for AI developers, their market impact was far more dramatic. Investors interpreted the arrival of cheap, high-performing open-weight models as a bearish signal for the semiconductor industry and a bullish one for the platforms that deploy AI. The core thesis: if frontier AI becomes accessible and inexpensive, the value shifts from the hardware companies that train the models to the software companies that deploy them.
Morningstar analyst Ivan Su captured the sentiment directly: "I see today's move in Hong Kong internet stocks less as a reaction to any fundamental change and more as a result of funds rotating out of AI hardware names" . The Straits Times reported that Moonshot's Kimi K3 unveiling was "fuelling a tech rout" in semiconductor stocks
.
The flip side of the Chinese tech rally was a historic sell-off in chip stocks. The Philadelphia Semiconductor Index (SOX) officially entered a technical bear market on Friday, July 17, falling 20% from its June 22 peak . The index dropped 4.8% on July 17 alone and fell 10% for the week
. The sell-off wiped out roughly $3.3 trillion in chip sector value
.
Memory and storage stocks were hit hardest. As of July 20, Sandisk had fallen 42.4% from its all-time high, Western Digital was down 40.3%, Micron had dropped 32.4%, and Seagate had lost 31.2% . The sell-off was so severe that it even knocked Nvidia from the top market cap spot, ceding it to Apple
.
Importantly, this bear market followed an extraordinary run—the SOX had risen 105% in just three months before the crash and was still up more than 60% year-to-date as of July 17 .
As capital rotated out of chips, it flowed into Chinese internet heavyweights. On Monday, July 20, Alibaba jumped 3.7% at the close , and its shares had already been rising since July 16, when it gained 3.1%
. Tencent rose 3.5% on the same day
. Earlier in the week, Tencent had already surged 10.5% (to HK$481.6) on June 2 following reports that it was testing an AI agent prototype for WeChat
.
The broader trend was clear: investors were rotating into less-crowded parts of the market and betting on Chinese tech leaders as AI model makers began developing their own chips, boosting confidence in China's tech ecosystem .
Kimi K3 (released July 16, 2026) is a 2.8-trillion-parameter sparse mixture-of-experts model with a 1-million-token context window and native vision capabilities . Its architecture uses only about 16 of its 896 experts per token, keeping inference costs far below what the raw parameter count suggests
. Full open weights are scheduled for release by July 27
. Kimi K3 has topped LMArena's Frontend Code Arena, overtaking Claude Fable 5
.
Qwen3.8-Max-Preview (announced July 19, 2026) is Alibaba's 2.4-trillion-parameter multimodal model . The company describes it as "one of the most powerful models available today," second only to Anthropic's Claude Fable 5
. A preview version is live on Alibaba's Token Plan, Qoder, and QoderWork platforms, with open weights promised "soon"
. Bloomberg reported that Alibaba shares rose after this unveiling
.
While sources do not explicitly draw a comparison to the January 2025 DeepSeek moment, the core narrative is consistent: the arrival of cheaper, open-weight frontier models expands the total addressable market for AI applications, benefiting platforms and cloud providers that deploy them over chipmakers that train them . The rotation out of semiconductor stocks into Chinese internet names is the market's way of expressing this thesis.