Daily Alpha · X
· Post-Market Alpha · by Buzzberg Research
X's strongest verified signals were the power constraint on AI buildout, Micron timing into the September 30 print, a disclosed Credo buy at $152.92, hyperscaler credit widening before equities, and the AAON-versus-Vertiv cooling comparison.
Themes on this desk
Memory tightness versus cycle risk
BofA pushed back on de-speculation after meeting memory makers and still sees large 12-hi HBM4 shipments, with Samsung expected to more than double HBM4 output next year and total HBM up about 40% to 250,000 wafers a month. The bear side is
Power as the AI bottleneck
bubbleboi argues most GPU sales are sitting in warehouses, that power is slower to fix than compute, and that whoever can bring power online or cut consumption holds the advantage; bitcoin miners already hold power and grid interconnections
Rates repricing versus risk assets
Treasury volatility posted its biggest weekly jump in over a year with the MOVE index up nearly 30% and October hike odds near 70% from virtually zero a month earlier. Barclays warns the equity risk premium is near multi-decade lows; Citi s
Ticker heat
?
Top voices by smart followers and alpha score
Market Radar →Power, not compute, is the binding constraint on AI buildout
@bubbleboi argues most GPU sales are sitting in warehouses, that power is the slower-to-fix constraint versus compute, and that anyone who can bring power online or cut power consumption holds the advantage.
Goes without saying we are power constrained and anyone who can bring power online (or reduce power consumption) is king.
If power availability rather than chip supply gates AI revenue, value accrues to grid-connected power owners, transformer and electrical equipment suppliers, and efficiency plays rather than to GPU volume alone.
Watch Confirmation is neoclouds publicly disclosing unplugged B300 inventory or delayed energization dates; invalidation is hyperscalers reporting compute (not power) as the binding limit on cloud revenue.
Source →AAON pitched as cheaper, faster-growing data center cooling alternative to Vertiv
The author argues AAON's data center brand BASX grew 216% last quarter and is now half the company, with AAON at 30.2x forward earnings, 17.3x EV/EBITDA and 3x sales growing 55-60%, versus Vertiv at 31.8x forward earnings, 23.6x EV/EBITDA and 6x sales growing 23%; both are well below recent highs for reasons the author says don't hold up.
Growing 55 to 60%. Backlog $2B, up 98%. Raised full year growth guidance. Same forward P/E.
If AAON's BASX growth is durable, the market is paying a similar forward P/E for more than twice the growth rate, a potential mispricing in data center thermal management; the comparison also frames Vertiv's selloff as timing rather than demand.
Watch AAON's next quarterly BASX revenue growth and margin, plus whether the Memphis plant ramp converts to positive operating leverage; Vertiv order/backlog trends for demand confirmation.
Source →Micron FQ4 modeled near $52B with Q1 guide step to $58-59B
The author models Micron's fiscal Q4 near $52.1bn, aligning with UBS's reported ~$52bn, and argues part of the step UBS pushes into Q1 guidance ($58-59bn) is already in the current quarter, giving a $56.2bn base case for the July-Sept quarter. The cited UBS work also flags DRAM fulfillment around 60%, server DDR bit demand up ~80% y/y, server plus storage SSD bits up more than 100% in 2027, and a buyback ramp toward $40-50B a quarter after CHIPS Act limits expire December 9, 2026.
UBS reportedly has Micron's fiscal Q4 near $52bn, with the big step pushed into Q1 guidance ($58-59bn).
The disagreement is about timing, not direction: if the revenue step lands a quarter earlier than UBS models, the Sept 30 print and guide could surprise relative to a Street that still sits under the $58-59B guide.
Watch Whether Micron's actual FQ4 revenue lands near $52B and whether FQ1 guidance reaches $58-59B, plus whether the 60% DRAM fill rate holds.
Source →Caesar Capital discloses buying CRDO dip at $152.92 average
Caesar Capital disclosed starting a new position in Credo Technology at an average price of $152.92 per share, framing the buy as purchasing a dip.
of $152.92 per share 🏛️ https://t.co/o73gNVcZGQ
A disclosed fresh long in an AI networking name at a specific cost basis gives a concrete reference level for sentiment and positioning in CRDO.
Watch Whether CRDO holds above the ~$153 average cost and whether other disclosed buyers add on weakness.
Source →Hyperscaler CDS widening 2-4bps while equities hold
A Goldman Sachs mid-day wrap cited by ZeroHedge says equities have been resilient amid higher rates but cracks are appearing in credit, with hyperscaler CDS most impacted, trading 2-4bps wider and underperforming unchanged QQQ; heavy issuance is expected to continue and drive wider spreads.
credit space. Hyperscaler CDS is most impacted and trading 2-4bps wider, underperforming unch’d QQQs.
Credit markets pricing hyperscaler risk ahead of equities is an early warning signal for AI-capex-funded balance sheets; sustained spread widening could pressure funding costs and eventually equity multiples.
Watch Confirmation: continued hyperscaler CDS widening and further large bond deals at wider concessions. Invalidation: spreads retrace as issuance is absorbed without concession.
Source →