Bullpen.News From Edwin Dorsey
Get Daily Recap Feedback

Sunday, November 8, 2026

Back to the front page

From X

@ACapitalLP 1 click

A Capital on X

$FLEX looks interesting into the Axiom spin in Q1 2027. I think everyone can agree RemainCo is worth somewhere in the $40s, so the only real disagreement is what value to assign Axiom. You’re getting a cloud and power infrastructure business growing at gangbuster rates (80%+ FY28 revenue growth) that just raised $2bn of convertible preferred from General Catalyst at a $37.5bn EV… Meanwhile, at ~$77, the market is implying Axiom is worth ~$33/share. Some SOTP estimates have it north of $100. Obviously there are legitimate concerns around capex, working capital and FCF conversion, but that’s a pretty enormous disconnect. November 10 Investor Day should be interesting…
@convequity 2 clicks

Convequity on X

The usual explanation for CoreWeave’s credit pressure is that its loans outlast its customer contracts. The filings say otherwise. All of $CRWV’s debt is due before its current contracts end, though only just: $14.6B of notes falls due in 2031-32, as the last contract cash comes in. That distinction matters. For a GPU cloud, debt that outlives its contracts is the dangerous kind. When a contract ends the chips are old, and there is no contract left to borrow against. CoreWeave’s real problem is speed. Each new build needs fresh capital before the last one has paid for itself, and almost all of the debt sits with the parent company. $IREN has the opposite shape. Its Microsoft project loan is matched to the contract. The risk is $6.3B of convertible bonds due 2031-33, mostly after its contracted cash has ended. If IREN’s stock is high by then, bondholders will likely swap the bonds for shares. If it’s low, IREN has to repay them in cash. Which leads to an odd conclusion: slower AI demand growth could help both. Fewer new commitments would give finished projects time to earn back their cost.
@convequity 1 click

Convequity on X

$UMC's nearer-term driver is its 12nm process with $INTC, and our longer-term view is that logic folding could make chips on its existing processes faster and more energy-efficient. UMC reports Q3 earnings tomorrow, while production is targeted for 2027 at Intel's Arizona factories. The partnership combines UMC's foundry experience with Intel's manufacturing capacity, giving UMC access to U.S. production without building an American factory. The immediate attraction is winning customers and expanding its business. Folding replaces long horizontal connections with shorter vertical ones, so signals need less time and energy to move. That could make chips built on established processes more capable without matching the newest transistor nodes. Older processes are, in our view, a more manageable starting point for this prospective thesis: UMC would still need to solve bonding, heat removal and reliable production, but it would not also be introducing a new transistor process.

Investor Letters

All letters

Axon 3Q26 Business Update

Axon cut full-year guidance after delayed federal Dedrone contracts slowed 3Q revenue growth to 28%, while software and services grew 43%.

AXON $31B
Speedwell Memos 3 clicks

Publications we follow

Every publication on this site, hand-picked.

129 publications

Newsletters 109

YouTube 4

Blogs 15

Seeking Alpha 75 of 279

Nominate a publication

X accounts we follow

Every account on this site, hand-picked.

91 accounts
Nominate an account

Investor letters we follow

359 firms

Sectors

Ordered by how often our editor reaches for them.