Bullpen.News From Edwin Dorsey
Get Daily Recap Feedback

Tuesday, October 20, 2026

Back to today

Vicor Update, Part Three

Vicor’s $66.6 million in AI power-delivery royalties generated about $50 million of operating profit, eclipsing its module business as revenue reached $197.1 million.

VICR $10B
CV's Substack 15 clicks

From X

@ParadisLabs 2 clicks

Paradis on X

Once again, the GPU depreciation bear thesis is dead. Elon on the $SPCX call: "The value of compute per watt will increase over time, not decrease... that's why you see things like even old computers like H100s renting for more than they rented when they were new." When supply is this constrained, GPU rental prices are set by the value of the marginal token, not the replacement cost of the chip. Models keep getting better, and better models raise the value of every installed watt faster than new GPUs deflate the old ones. Various datapoints all stack up: -> $SPCX says compute contracts are renewing "at higher dollars per watt, not lower dollars per watt" -> 1-year H100 contract pricing went from ~$1.70/hr to ~$2.35/hr between Oct 2025 and Mar 2026 (a 2022 chip!) -> $CRWV signed an A100 contract running into 2029 (a 2020 chip) and said older-gen pricing is at / above a year ago, and raised list prices ~25% in July. -> $NBIS's first Blackwell auction cleared 15% above its previous high and as a reminder....they told customers H100 on-demand goes up ~17% (from Oct 1). -> Zuckerberg: $META is getting offers for its compute "at a significant premium over what we paid for it". -> Lenders are now lending against GPUs out to 2030 e.g. $NBIS borrowed against its GPUs at SOFR+250 in July. Feels like the 2-3 year GPU useful life bear thesis keeps getting setback. Ofc, the one thing that ruins the party is the S/D gap closing -- Rental prices fall back toward replacement cost + older GPUs get repriced on perf/W. Which won't happen anytime soon.
@aleabitoreddit 2 clicks

Serenity on X

Yeah this statement was heavily bullish on neoclouds like $NBIS along with many others "Our belief is that the value of compute per watt will increase over time, not decrease. The reason for that, I think it should be obvious to people really, is that as you improve the AI software, as the models get better and better, you get more useful intelligence per watt And the value of that compute is proportionate to the useful intelligence per watt. And that’s why you see things like even old computers like H100s renting for more than they rented when they were new. And I think that’s likely to continue. I think we’re going to face a shortage of superintelligence compute and that shortage is going to remain for a long time"

Investor Letters

All letters

Heico Corporation

HEICO sells FAA-approved aircraft replacement parts at 30% to 50% discounts to OEMs, gaining customers as restrictive aftermarket contracts…

HEI $43B
D Invests 6 clicks

Copper Fox Metals: Research Flash

Copper Fox’s C$9.1 million Schaft Creek program advances metallurgical, tailings, water and road studies toward a pre-feasibility study,…

TECK $31B
Atrium Research - Equity Research 5 clicks

AppLovin Deep Dive

AppLovin rebuilt its ad model with AXON 2.0, lifting LTM revenue to about $7 billion and margins to 77% as Unity competition threatens ad…

APP $92BU $19B
Speedwell Memos 63 clicks

Is The Market Turning The Page?

CoreWeave’s debt and equity price action is framed as a wake-up call for investors amid a broader market discussion.

CRWV $34B
WHAT ARE WE DOING? “Contrarians At The Gate” 54 clicks

Constellation: Stars Aligning?

Constellation Brands trades near $127 at roughly 11x next-twelve-month EV/EBIT, with recent business results offsetting concerns over…

STZ $22B
TSOH Investment Research 28 clicks

Publications we follow

Every publication on this site, hand-picked.

128 publications

Newsletters 108

YouTube 4

Blogs 15

Seeking Alpha 52 of 244

Nominate a publication

X accounts we follow

Every account on this site, hand-picked.

90 accounts
Nominate an account

Investor letters we follow

189 firms

Sectors

Ordered by how often our editor reaches for them.