Moonshot launches China's largest AI Model Kimi K3
Moonshot AI released Kimi K3, billed as China's largest AI model, intensifying open-weight competition with US frontier labs.
5 articles · 2 categories
The finishable daily brief
Monday, Jul 20, 2026
5 articles · 2 categories
read top to bottom · then stop
In 30 seconds
Moonshot's Kimi K3 launch kept reshaping markets a day later: Samsung and SK Hynix shares rebounded on bets that cheaper, more capable open models raise compute demand rather than cut into it, and Moonshot is now moving toward a Hong Kong IPO with Kimi K3's US market impact cited as the catalyst.
A newly surfaced 2022 email shows Sam Altman once proposed the same playbook Moonshot is now running: release a capable open model specifically to discourage funding for smaller open-source rivals. The document emerged from the Musk v. Altman litigation, not from OpenAI.
Moonshot's Kimi K3 launch is reshaping downstream markets: semiconductor stocks are rebounding on rising compute-demand bets, and Moonshot is moving toward a Hong Kong listing.
Moonshot AI released Kimi K3, billed as China's largest AI model, intensifying open-weight competition with US frontier labs.
US labs are reportedly rattled by China's strategy of releasing frontier-capable models as open weights, with Kimi K3 as the latest example.
Samsung and SK Hynix shares rebounded on the thesis that cheaper, more capable open models like Kimi K3 increase overall compute demand rather than cannibalize it.
Moonshot is moving ahead with a Hong Kong IPO, with Kimi K3's impact on US tech stocks cited as a catalyst.
A Sam Altman email disclosed in the Musk v. Altman litigation shows OpenAI once planned to pre-empt open-weight competitors with its own weaker open release — the same dynamic now playing out with Kimi K3, in reverse.
An October 2022 email to OpenAI's board, surfaced in the Musk v. Altman case, has Altman proposing a GPT-3-class open model to run locally, explicitly to 'discourage others from releasing similarly-powerful models' and starve smaller open-source efforts of funding.
You are caught up for this edition