newsroom
$MU · AI Found the Trades We Missed
now playing · $MU
$MU···bullish· highjosh kale
Micron up 180% as AI memory demand drives 80% margins for DRAM makers
Memory is the largest cost component in AI compute; surging HBM/DRAM demand for GPU clusters has pushed top memory manufacturers to 80% margins and 153% average returns.
$INTC···bullish· mediumunknown
Intel seen as 35x government-backed play on domestic chip manufacturing
Federal investment via CHIPS Act and national security compute mandates could drive massive rerating for Intel's foundry and manufacturing assets.
$000660.KS···bullish· highunknown
SK Hynix eclipses Samsung as Korea's most valuable company on HBM dominance
SK Hynix's leadership in high-bandwidth memory (HBM) for AI accelerators has made it Korea's most valuable company, surpassing decades-long leader Samsung.
$SSNLF···bullish· highunknown
Samsung becomes world's most profitable company surpassing Nvidia on memory boom
AI-driven memory demand has made Samsung's semiconductor division the world's most profitable business, exceeding even Nvidia's earnings.
$VALOR-ATOMICSbullish· highunknown
Valor Atomics 13x valuation on SMR thesis for AI data center power
Small modular nuclear reactors (SMRs) are the solution to the AI data center energy bottleneck; Valor Atomics went from seed to $2B valuation (13x) by targeting this exact need.
$OPENROUTERbullish· mediumunknown
Open Router 2.6x valuation as multi-model routing layer wins developer adoption
Companies won't commit to a single model; Open Router's aggregation layer with stealth model launches and intent data capture positions it as the universal API for AI model access.
$BOOM-SUPERSONICbullish· mediumunknown
Boom Supersonic pivots to gas turbines for AI data center power generation
Supersonic jet developer repurposes turbine expertise for high-power density generators needed by AI data centers, turning a transport play into an energy infrastructure bet.
$PERPLEXITYbullish· highunknown
Perplexity valuation jumps to $21B on Samsung default AI agent deal
Samsung making Perplexity the primary AI agent on all phones provides massive distribution moat, driving valuation from $18B to $21B and validating the answer-engine model over ch…
$XAIbullish· mediumunknown
xAI (Grok) positioned as comeback contender with SpaceX launch infrastructure
Integration with SpaceX's satellite launch capability gives xAI unique access to space-based compute for training, potentially leapfrogging terrestrial power constraints.
$STARCLOUDbullish· mediumunknown
YC-backed StarCloud pioneers orbital GPU clusters for space-based model training
Launching H100 clusters into orbit bypasses terrestrial power/cooling limits; first mover in space-based compute with direct path to SpaceX-scale deployment.
$SPCX···bullish· highunknown
SpaceX to lead space-based AI training via satellite GPU clusters
SpaceX's Starlink satellite constellation enables orbital data centers that solve power and cooling constraints; the rebranded xAI team is actively building Grok training in space…
$ADAMSbullish· highunknown
Travis Kalanick's Adams builds narrow-use robots for precious metal logistics
Specialized robotics for critical material extraction and distribution solves real-world supply chain constraints, leveraging Kalanick's operational scaling expertise from Uber.
$TSLA···bullish· highunknown
Tesla Optimus humanoid robot scaling to match Model 3/Y production velocity
Tesla's manufacturing expertise will enable humanoid robot production at automotive scale, making Optimus the first mass-market general-purpose robot platform.
$META···bearish· mediumunknown
Meta cuts frontier model spending as API costs drive shift to open-source alternatives
High frontier model inference costs ($10/$50 per million tokens) are forcing even hyperscalers to adopt cheaper open-source models on owned infrastructure.
$AAPL···bullish· mediumunknown
Apple's on-device model strategy positions it for local AI agent dominance
Apple's vertical integration allows custom silicon-optimized models running locally, avoiding frontier model API costs and enabling privacy-preserving personal agents.
$ETCHEDbullish· highunknown
Etched targets 10x inference cost reduction with transformer-specific ASICs
Specialized transformer ASICs can deliver equivalent token throughput at one-tenth the cost of GPUs, fundamentally altering AI economics and enabling mass deployment.