TickerTain
$SPCX····$NVDA····$MSFT····$GOOGL····$ANTHROPIC·$TSLA····$CRWV····$NBIS····$HIMS····$TSM····$ASML····$FIGURE·$LYFT····$UBER····$SNOW····$META····$ELEVENLABS·$CURSOR·$COGNITION·$OPENAI·$AAPL····$RKLB····$XAI·$AMZN····$AMD····$000660.KS····$INTC····
← newsroom
$MOONSHOT-AI · [Emergency Episode] Moonshot’s Kimi K3 has Arrived! China has a Frontier Model
now playing · $MOONSHOT-AI
$MOONSHOT-AIbullish· highmax
Moonshot's Kimi K3 ranks third globally behind GPT-4.5 and Sonnet 3.6
Kimi K3 achieves frontier-level performance at 2.8T parameters, making it the third-best model globally behind only GPT-4.5 and Sonnet 3.6, though the delayed open-weight release…
$GOOGL···bearish· highmax
Google should be embarrassed as Chinese model surpasses its AI standing
Google has fallen from the perceived 'AI big three' as Moonshot's Kimi K3 surpasses it, yet market perception lags with many still believing Google remains a top-three frontier la…
$CRWV···bullish· lowmax
CoreWeave and neoclouds in licensing talks to serve Kimi K3 on GB300s
Neoclouds like CoreWeave are likely negotiating licensing deals with Moonshot to serve Kimi K3 on latest GB300 infrastructure, creating new revenue streams for GPU cloud providers…
$NBIS···bullish· lowmax
Nebius positioned to serve Kimi K3 via licensing deal on latest hardware
Nebius (referred to as 'nebuses') is among neoclouds negotiating to serve Kimi K3 on GB300s, leveraging their GPU infrastructure for open-source frontier model hosting as a differ…
$NVDA···bullish· mediumjordan
Kimi K3's 2.8T parameters require latest Nvidia B300/GB300 chips to serve efficiently
Frontier models like Kimi K3 at 2.8T parameters cannot fit on H200/B200 GPUs, driving demand for Nvidia's latest Blackwell generation (B300/GB300) for single-server inference and…
$AMD···bullish· lowjordan
AMD MI355X positioned as alternative for serving 2.8T parameter models
AMD's MI355X is mentioned alongside Nvidia's latest chips as capable of serving frontier-scale models like Kimi K3 on a single server, indicating competitive positioning for high-…
$OPENAIneutral· mediumjordan
OpenAI likely holds back larger models due to US government restrictions
OpenAI has larger, more capable models than publicly released (GPT-4.5), but US government restrictions prevent their deployment, artificially narrowing the gap with open-source c…
$ANTHROPICneutral· mediumjordan
Anthropic similarly constrained by US restrictions on frontier model releases
Anthropic faces the same US government restrictions as OpenAI, preventing release of their most capable models and allowing open-source models to catch up artificially to the depl…
$HUAWEIbullish· mediumjordan
Huawei Ascend leads Chinese accelerator push for frontier model inference
As China prioritizes domestic AI compute, Huawei's Ascend chips are positioned as primary infrastructure for running frontier models like Kimi K3, benefiting from national policy…