Daily Alpha · YouTube
· Premarket Alpha · by Buzzberg Research
Who was talking 50 videos · 11 channels
AI funding quality
1 video · 1h 34mThe buildout increasingly requires leverage, making refinancing and customer solvency as important as chip shipments.
OpenAI's custom inference chip performance claims
OpenAI claims its new in-house inference chip, 'Jalapeno,' outperforms Nvidia's flagship processors in throughput and latency, offering 1.8x to 4x better performance-per-watt in internal tests.
Inference competition
1 video · 1h 34mOpenAI’s internal chip claims put production benchmarks and workload economics at the center of the Nvidia debate.
OpenAI's custom inference chip performance claims
OpenAI claims its new in-house inference chip, 'Jalapeno,' outperforms Nvidia's flagship processors in throughput and latency, offering 1.8x to 4x better performance-per-watt in internal tests.
Duration intervention
1 video · 5mTreasury buybacks can move positioning and spreads, but speakers disagree that intervention can dominate fiscal supply.
Treasury intervention in bond markets seen as futile
Peter Boock Boockvar argues that Treasury Secretary Bessent's attempts to cap long-end Treasury yields (specifically near 4.75%) are likely to fail because the bond market is significantly larger than the Treasury's intervention capacity.
Potential reduction in HBM capacity for next-gen chips
Reports suggest Nvidia's 'Rubin Ultra' chips may significantly reduce HBM capacity requirements compared to initial projections, potentially due to supply constraints or design shifts.
Northrop Grumman capacity expansion for munitions and solid rocket motors
Northrop Grumman has added 2 million square feet of manufacturing space over the last two years and is building additional plants to reach 3x original capacity for solid rocket motors, specifically to serve as a second source for tactical missiles like PAC-3.