New Ask Claude or ChatGPT about any stock, using Bullpen. Bullpen in Claude or ChatGPT. 7 days free

Wednesday, November 18, 2026

Back to the front page

From X

Serenity on X

Now that we're close to 2027 it's a good idea to stress test downside scenarios where you have: 1. Capex cuts past 2028 2. Stranded Compute / DC Slowdowns 3. Finance deterioration (base case) 4. 2x+ rate hikes Working in parallel and see what happens with valuations of AI longs. If you see acceleration despite all 4, I'd argue that's an opportunity. > Next-gen of photonics is one of the rare segments with an extreme rise in optical content and ASPs that can offset aggregate capex cuts. Everyone’s moving into that space, including Elon Musk doing M&A, Marvell, Samsung and others with M&A… and we haven’t seen economic impact on emerging architectures even hit players from Celestial to Ayar + their suppliers yet. If we look at a scenario where there's excess GPUs but power constraints: We might see slowdown in aggregate accelerator/GPU/ASIC procurement, but your optical engines + light sources + switches players see acceleration. Especially as incentive to replace power inefficient networking... since different (cheap BOM) optical architectures improves GPU utilization or even HBM requirements. So value migrates to players more high-beta to architectural shifts. If Celestial projections are $500m -> $1B, and we expect -30% aggregate capex cuts, I'd argue we'd still see upward revisions from Celestial over the next 2 years. But financing with #3 matters a lot as seen with $AAOI dropping off of a $550M, vs. $COHR getting prepayments for capacity expansion from their customers. Companies that rely on external foundries might be less sensitive to #3 since they're not as capex intensive. On the other hand, companies in sectors have a lot of floating interest like certain Neoclouds, and require further financing might be the most impacted. > Legacy and specialty memory is a different cycle that markets conflate with AI capex. If consumer demand weakens but at the same time, players like Windbond exits DDR2, pricing durability is longer than markets expect. > Wafer/substrate procurement such as ABF, SOI, InP are low cost but large downstream destruction if not secured, so we’ll likely see acceleration here as seen with $NVDA earnings call. > qualification cycle players within bottlenecks remain very compelling, since first revenue in a large TAM, relative to MC is material to that company. If a company is valued at $400m then they get a $200m of orders, it's more asymmetrical to a certain company than a rounding error to a $50B incumbent revenue. _ There’s period of times where prices are dislocated from fundamentals from leveraged players getting wiped. I’d argue that time is now. I disagree with a lot of the stupid "it's over takes" but I agree it's a good idea to haircut future capex expectations and forward valuations due to financing tightening + macro. It's also good idea to see the nuances and stress test downside cases. But TLDR: I'd argue next phase of AI infrastructure is going to be increasingly disconnected from aggregate AI capex growth. I just personally wrote 4 things I'm using to stress test jointly some of the names I'm looking at, so I'm more comfortable waiting things out... despite seeing hedge funds getting liquidated left and right and overwhelming short positioning. I personally see acceleration in more narrow parts of the supply chain, while markets treat the entire AI trade as one. We'll see how things play out on longer horizons.

S&P Global

S&P Global’s valuation swung from 7x earnings in 2009 to 50x in 2024 as ratings, indices and proprietary data expanded through acquisitions…

Monte Investments Market Brief Vol. CXXXVI

Monte Investments rotated Nvidia into Alphabet before earnings and reviews Knife River’s margin squeeze, AMD’s World Labs acquisition, and…

Envisioning Entravision

Entravision’s AdTech revenue rose 133% and EBIT 261% in Q3 as management weighs TV-station sales before its TelevisaUnivision agreement…

BLND · long

Blend Labs INC

Blend Labs (BLND) is presented as a long investment idea, with prior VIC writeups cited for business background.

Why I Agree With Rob Vinall on Wix

Wix’s paying-subscriber conversion fell from 2.72% to 2.01% since 2019, while Business Solutions drove growth and activist-led cost cuts…

Publications we follow

Every publication on this site, hand-picked.

130 publications

Newsletters 110

X accounts we follow

Every account on this site, hand-picked.

91 accounts

Investor letters we follow

383 firms

Sectors

Ordered by how often our editor reaches for them.