Daily Alpha · Substack
· Post-Market Alpha · by Buzzberg Research
A technically driven newsletter added US equity and leveraged metals exposure and levered up copper, while SemiAnalysis previewed the robot inference architecture fight.
Themes on this desk
Metals positioning
Leveraged gold-miner and copper-beta adds dominate the desk's trade posts.
Robot inference
On-device versus datacenter inference framed as the driver of robot silicon and DRAM demand.
Rooster adds to US share portfolio
A Rooster post titled 'ADDING TO US SHARE PFOLIO' discloses the author adding to a US equity portfolio, consistent with the same session's 'NASDAQ HAS NOT CRASHED' framing.
Confirms the newsletter is deploying capital into US equities rather than de-risking, a concrete bullish positioning signal beyond the technical commentary.
Watch Continued Nasdaq strength and follow-through in the added positions confirms; a sharp US equity drawdown would invalidate the add.
Source →Rooster reiterates JNUG as a buy
A separate Rooster post states plainly that JNUG is a buy.
Reinforces leveraged gold-miner exposure as an active long within the same session as other metals buys.
Watch Sustained upside in JNUG after the call confirms; a break of the referenced stop level would invalidate.
Source →Rooster rotates COPJ into leveraged COPZ
Rooster discloses selling COPJ and buying COPZ, described as a leveraged wire, in a single rotation.
Shows the author increasing copper exposure beta rather than exiting the theme, implying continued conviction in copper upside.
Watch Leveraged copper vehicles should outperform COPJ if the copper thesis holds; underperformance would flag a failed rotation.
Source →SemiAnalysis frames on-device vs datacenter inference as the robot-brain question
SemiAnalysis previews a piece comparing on-device versus datacenter inference for robots, covering robot models, silicon and DRAM efficiency, Jetson Thor versus B300 TCO, deployments, and 'the network wall'.
Positions the robot inference architecture choice as a driver of silicon and DRAM demand, a cross-source signal that memory and edge-AI silicon are contested battlegrounds.
Watch Evidence that robot inference workloads concentrate in datacenters (favoring B300-class TCO) versus on-device (favoring Jetson Thor-class silicon) would resolve the debate.
Source →