$MU - Micron has sold out entire 2026 HBM supply, revenue up 4x year-over-year, and DRAM is identified as the most important bottleneck in AI infrastructure with only three companies globally able to produce high-bandwidth memory for AI servers.
$SPACEX - SpaceX orbital compute economics becoming increasingly favorable as terrestrial data center costs inflate (now $60B per gigawatt vs $40B for orbital), with launch costs potentially dropping to $5B per gigawatt as Starship becomes reusable, creating structural cost advantage.
$ANTHROPIC - Anthropic valued at $3 trillion by Gavin Baker with revenue expected to exceed $100B this year and potentially $200-300B by 2028, with 85% gross margins on inference at scale.
$TSLA - Tesla developing Megapod modular data centers that could enable 90-day build cycles versus traditional multi-year timelines, with potential to place compute at Supercharger stations and integrate with Powerwall/Starlink for distributed compute network.
$AIMODELS - Composable AI model architectures identified as the future, with enterprises using routers to direct queries between open-source models (85%+ of tokens) and frontier models for hardest problems, shifting economic value to infrastructure while maintaining role for frontier models.
$DATACENTER - Existing terrestrial data centers becoming increasingly valuable as new builds face rising costs ($60B+ per gigawatt), regulatory challenges, and component shortages, creating scarcity premium for operational facilities with power and cooling infrastructure.
Bearish:
$AAPL - Apple forced to pass on significant DRAM cost increases to consumers with 14-25% price hikes across product lines due to AI-driven memory shortage, representing 'AI inflation' hitting consumer electronics and potentially causing demand destruction.
$CVRS - Cerebras broke IPO deal price triggering price-insensitive selling from portfolio managers who automatically sell stocks that break deal price, with shorts piling on to push it lower, and company struggling to communicate power scaling story effectively.