Bullpen.News From Edwin Dorsey
Get Daily Recap Feedback

Monday, November 9, 2026

Back to the front page

Roblox: Capture This Moment

Roblox’s Q3 FY26 DAUs fell year over year for the first time, while bookings declined 12% to $1.7 billion.

RBLX $35B
TSOH Investment Research 1 click

From X

@convequity 1 click

Convequity on X

The neocloud debt pullback is about how long delivery takes, and the useful response is a targeted renegotiation that pays lenders for more time, not a liquidation of the borrowers. Demand for AI compute is huge, and investors were eager to fund the buildout. That demand is still there. What is drying up is debt for neoclouds that cannot show revenue outside AI data centers. In our view the pullback is almost entirely about timing. Compute gets billed only once the site is built, powered and staffed, long after the cash has gone in. That wait is the squeeze. As a hypothetical, if the neoclouds could stand compute up and start billing ten times faster, the problem would ease. They would be generating revenue and cash flow to fund more of the next investment themselves, and outside investors could ask less interest on future money. On the latest reported figures, $CRWV's operating cash flow is positive but its free cash flow is negative, because the build is still being paid for. Lenders are now pulling back from neoclouds that have no revenue outside AI data centers. The race between neoclouds pushed CoreWeave too far. They took the debt probably knowing the delivery timetable was tight, because a firm that is not aggressive enough ends up with no compute at all. If lenders call a default and CoreWeave is then bankrupted and liquidated, the credit investors are left holding the GPUs. Taking the collateral does not make it earn. Someone still has to power, staff and bill the site, so the lenders must find another company to run it. In our view they lose money net-net, the handover is costly, and the buildout stalls to a degree. Amazon, Microsoft and Google can keep funding GPUs from businesses that already earn. A stall on the independent side leaves more AI compute with them, which in our view is bad for an AI-led economy. Who would buy a failed neocloud is still open. Blanket leniency would be a moral-hazard problem. If every late project gets time for free, covenants stop binding and the next borrower takes the same risk again. The exception here is that the demand behind the loans is still there. A lender can give more time to a project that will cover its debt once it is billing, and be paid for the wait with a higher margin, a fee or upside. When the cash flow recovers, that is the lender's reward. Musk predicting that 200-plus neoclouds will fail or consolidate makes the pullback worse. More credit investors step back, and the wait gets longer for everyone still building.

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 4 clicks

Publications we follow

Every publication on this site, hand-picked.

129 publications

Newsletters 109

YouTube 4

Blogs 15

Seeking Alpha 76 of 280

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.