Bullpen.News From Edwin Dorsey
Get Daily Recap Feedback

Wednesday, October 21, 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 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.
@convequity 1 click

Convequity on X

Software multiples have fallen further since our January report. As of 7 Oct 2026, 238 companies in our all-software watchlist have EV/Sales (NTM) data on Koyfin. The median is 3.1x, down from 3.7x in January. The average is 5.4x, down from 6.2x. The 25th percentile is 1.8x, down from 1.9x. The 75th percentile is 5.5x, down from 6.1x. 191 of the 238 (80%) trade below our 6.7x private equity floor. 113 trade below 3x. The Bessemer Cloud Index shows the same pattern. Its median is 4.7x, down from 5.1x in January. 46 of its 68 companies (68%) trade below 6.7x. The cheapest are $CXM at 0.9x, $SPT at 1.1x, $MNDY at 1.7x and $WIX at 1.8x. In our watchlist, only 21 companies trade above 10x. The most expensive are $PLTR at 44.5x, $CRWD at 42.4x, $NET at 38.6x, $PANW at 24.2x, $DDOG at 19.5x and $SNOW at 16.3x. Growth still earns a higher multiple. The 44 names expected to grow revenue 20%+ a year over 3 years have a median of 5.2x. The 85 names expected to grow under 10% have a median of 2.6x. Even for the faster growers, the 5.2x median is below the floor. The market is paying up for very few names. We think that is right. Many software companies have slowing growth, high costs and no clear AI strategy. We would not buy the sector, or a cloud index, on valuation alone. In software, we focus on the few companies rebuilding their products around AI, and our SaaS Survival Scorecard tests key SaaS names against the pressure from AI agents.

Investor Letters

All letters

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…

VICR $9.7B
CV's Substack 28 clicks

Heico Corporation

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

HEI $42B
D Invests 13 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 6 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 $18B
Speedwell Memos 68 clicks

Publications we follow

Every publication on this site, hand-picked.

129 publications

Newsletters 109

YouTube 4

Blogs 15

Seeking Alpha 54 of 247

Nominate a publication

X accounts we follow

Every account on this site, hand-picked.

90 accounts
Nominate an account

Investor letters we follow

190 firms

Sectors

Ordered by how often our editor reaches for them.