Bullpen.News From Edwin Dorsey
Get Daily Recap Feedback

Tuesday, October 27, 2026

Back to today

From X

@convequity 1 click

Convequity on X

We prefer $UMC to $GFS, and we are increasing our UMC position. Both companies are foundries that make chips on mature processes, well behind the leading edge. We hold GlobalFoundries as a recovery investment. United Microelectronics offers two things on top of a recovery: a manufacturing partnership with production targeted for 2027, and a longer-term way to make its existing processes more competitive. 1. The nearer-term driver is the 12nm process UMC is developing with $INTC. 12nm is a mature process generation, not a leading-edge one. Production is targeted for 2027 at Intel's factories in Arizona. UMC contributes its foundry experience and its experience working with chip designers. Intel contributes the manufacturing capacity. UMC gains a way to offer customers chips made in the United States without building an American factory itself. UMC could also earn income from intellectual property and revenue sharing, but the commercial terms are not yet clear, so we do not yet count on royalty income. The attraction today is the chance to win customers and expand the business through the partnership. 2. The longer-term driver is logic folding. In a conventional chip, the circuits sit on one layer and signals cross it along long horizontal wires. Folding arranges the circuits across two or more layers connected vertically, so many long horizontal wires become short vertical ones. Signals travel a shorter distance, which takes less time and less energy. The chip gets faster and more energy-efficient without smaller transistors. For UMC, folding could make chips built on its established processes more capable and easier to sell, which extends the value of its factories without matching the most advanced processes at $TSM. We described the mechanism in our Huawei series on the Kirin 9050. 3. We think a mature process is the more manageable place to start folding. Whoever folds chips has to solve bonding, heat removal and reliable high-volume production. UMC would solve those on a transistor process it already understands. Applying folding to TSMC's newest processes would combine two difficult manufacturing changes at once: a new generation of smaller, more complex transistors, and active circuits connected across several layers. A problem with either one can reduce the share of usable chips, raise costs or delay customer launches. This is a difference in the risks being combined. TSMC already offers advanced stacking technologies and has the expertise. 4. The commercial incentives differ as well. For UMC, folding extends the useful life of processes it already owns and lets it compete for business that would otherwise need a newer process. For TSMC, folding on older processes competes with the returns it expects from its leading-edge investments and could disrupt customer roadmaps built around those investments. So UMC has the stronger reason to become an advocate for folding on older processes. This is our prospective thesis about where folding is most likely to be adopted. UMC has not announced a logic-folding program. We have two reservations. UMC's shares have already recovered from their low, so the entry point is less attractive than it was. Logic folding could take years to contribute to revenue. The Intel partnership is our reason to add now. The chance to make UMC's existing processes more competitive is why we hold a larger position in UMC than in GlobalFoundries.
@convequity 4 clicks

Convequity on X

I feel like $SSNLF's record quarter is closer to a peak than a new base. Samsung guided to about 107.4 trillion won (roughly $80bn) of Q3 operating profit. The shares fell 2.4% on the day. Here is why I read it as a peak: 1. The record came from price, not volume. Samsung's own guidance has conventional DRAM shipments flat year on year. HBM grew, but most of the jump is higher prices per bit. 2. Samsung's own phone unit is now a priced-out customer. Samsung is cutting Q4 phone output by 20–30% because memory is too expensive, and it has reportedly looked at buying DRAM from China's CXMT for Galaxy phones. When your own phone division shops around, the price is too high. 3. Buyers are working to use less HBM. AI chip designers are cutting HBM specs and finding ways to need less of it per accelerator. That attacks the mix gain directly. 4. Revenue missed even as profit beat. 195 trillion won against about 207 trillion expected. Margin did the work, and margin built on price is the first thing to reverse. My view: a commodity product earning these margins is unhealthy for everyone downstream, from phones and PCs to data centers. Buyers respond by cutting units, despeccing and adding suppliers. That is how ASP-driven cycles turn. The divisional split comes on October 29. I will be watching whether memory profit grew much faster than group profit, which would show the phone unit paying for it.
@convequity 1 click

Convequity on X

The $70bn revenue figure reported for OpenAI in late September was a gross-basis comparison number, not OpenAI's own. OpenAI's own run rate, the one it told investors, was approaching $50bn at the end of September. The number that now needs the same scrutiny is Anthropic's, before its IPO. The two labs count partner sales differently. When a customer pays $AMZN or $GOOGL for Claude, Anthropic records the full price as revenue and books the cloud's cut as a cost. OpenAI records only its own share of such sales. The $70bn was OpenAI's number grossed up to Anthropic's method so the two could be compared. Three things follow. 1. OpenAI's revenue did not fall. Its net run rate was approaching $50bn at the end of September. Yet the stocks that fell on Thursday were its suppliers: $ORCL, $CRWV and $NVDA. 2. Anthropic's $65bn run rate at the end of July is a gross figure. On OpenAI's basis it is smaller by whatever the cloud partners keep, and that share has not been disclosed. The claim that Anthropic has overtaken OpenAI rests on two numbers measured differently. 3. Which method applies is an accounting judgement about which company controls the sale to the customer, so Anthropic's prospectus for its planned November IPO may keep the gross figure, with the partners' share inside cost of revenue. That filing will show the cost line. OpenAI, still private, does not have to. In our view Thursday's selling hit the suppliers to the lab whose number did not change. The comparison investors should redo before November is Anthropic's $65bn against OpenAI's $50bn after the cloud partners' share is taken out of both.

Investor Letters

All letters

Right-Sizing Memory

Micron’s AI memory demand depends on workload-specific accelerator designs that shift bits among HBM, GDDR and LPDDR, changing wafer…

MU $946BNVDA $5T
Chipstrat 4 clicks

Are Neoclouds Dead?

Neoclouds face tighter debt financing as AI infrastructure capex outruns cash flow, leaving cash-burning GPU providers more exposed than…

NBIS $42B
Daniel Romero 9 clicks

Reddit: Deep Dive Analysis

Reddit generated $805 million in Q2 2026 revenue, 95% from ads, but slowing Google search referrals expose AI-driven traffic and…

RDDT $18B
Hated Moats 6 clicks

Qualcomm Inc. (QCOM)

Qualcomm supplies chips for on-device AI agents and data center inference, with Meta and AWS agreements expanding its role across both…

QCOM $195B
SixSigmaCapital: Navigating the Financial Markets 5 clicks
WULF · long

Terawulf INC

TeraWulf is shifting from bitcoin mining toward data centers, seeking a higher-quality and more defensive business model.

WULF $5.4B
Value Investors Club

The Data Center Bridge

Small- and mid-cap data center stocks trade as one despite sharply different funding bridges: 12 of 15 cover fixed costs on signed…

IREN $9.9B
CV's Substack 16 clicks

Amkor Q3 FY12/26 Earnings Review

Amkor's Q3 sales were flat at $1.991 billion as computing and automotive/industrial growth offset a 27% communications decline; Q4 sales…

AMKR $10B
Cestrian Capital Research 4 clicks

Publications we follow

Every publication on this site, hand-picked.

129 publications

Newsletters 109

YouTube 4

Blogs 15

Seeking Alpha 69 of 272

Nominate a publication

X accounts we follow

Every account on this site, hand-picked.

91 accounts
Nominate an account

Investor letters we follow

190 firms

Sectors

Ordered by how often our editor reaches for them.