Daily Alpha · X
· Premarket Alpha · by Buzzberg Research
X converged on Treasury-led duration management, Korean capital returns and biotechnology, while specialists supplied the sharper disagreements in memory, custom silicon and refining.
Themes on this desk
Treasury versus duration
Independent macro voices agreed that bill-funded buybacks change the maturity mix, but disagreed on whether that creates a durable long-bond floor or merely moves pressure into the dollar and curve.
Korean capital returns
SK Hynix's confirmed repurchase supported the won and memory sentiment, while Samsung's much larger reported return plan remains an announcement catalyst rather than a completed action.
Custom silicon competition
Marvell's Google warrant confirms supplier diversification, but specialist estimates still leave Broadcom with the largest TPU value share.
Ticker heat
?
Top voices by smart followers and alpha score
Market Radar →Winbond capacity pre-booking
Customers are reportedly negotiating for 2029 and 2030 capacity with Winbond, indicating extreme forward demand visibility for legacy memory.
Taiwanese media reports that customers have already begun preliminary discussions with Winbond—a niche/legacy memory maker—over its 2029 and 2030 capacity.
Suggests persistent supply constraints in niche memory segments, potentially benefiting margins for legacy manufacturers.
Watch Monitor future capacity announcements and supply chain reports for Winbond.
Source →BofA analysis of TPU market share
BofA estimates Broadcom will retain 55-60% of TPU value share, with Marvell at 10-20% and MediaTek at 20-30%, despite Marvell's recent deal with Google.
BofA: Marvell TPU Deal > Broadcom to retain 55-60% of TPU TAM value share (~$250-350bn) > Marvell at 10-20% share
The analysis suggests Broadcom remains the dominant player in the TPU market despite competitive inroads by Marvell.
Watch Assess whether Marvell can capture higher-than-projected share in future TPU cycles.
Source →Treasury buyback policy and AI capex
The Treasury announced a doubling of long-end buybacks; analysts suggest this liquidity support may facilitate continued hyperscaler debt issuance for AI capex by capping long-end yields.
Today, the Treasury announced that they'd "at least double" long-end buybacks with the 30Y at pre-GFC highs.
Suggests government policy is indirectly subsidizing AI infrastructure buildouts by managing duration demand.
Watch Impact on long-term Treasury yields and hyperscaler debt issuance volumes.
Source →Refining margins expected to remain elevated through 2027
The author argues that refining margins will not compress as the Street expects because geopolitical conflicts and refinery damage will persist, with normalization taking 12-18 months post-resolution.
Analysts across the board believe refining margins will compress and I simply don't believe that will be the case.
Contrarian view on refiner EPS projections; suggests current Street estimates are too bearish.
Watch Monitor Hormuz transit volumes, US distillate inventory builds, and Russian diesel export ban status.
Source →Biotech short squeeze and degrossing dynamics
The biotech sector experienced an unprecedented 1000bp spread between shorted baskets and crowded longs, which the author characterizes as a 'degrossing' event rather than a simple short squeeze.
Almost a 1000bp spread (insane) for context a bad squeeze day is usually 250-300bp.
Suggests that generalist capital is cutting exposure to high-quality longs to fund short covering, creating a potential buying opportunity once the liquidation phase stalls.
Watch Look for a 'hard bid' in crowded long names once gross exposure rebuilding begins.
Source →