Bullpen.News From Edwin Dorsey
Get Daily Recap Feedback

Saturday, October 24, 2026

Back to today

Tower Semiconductor Deep Dive

Tower Semiconductor’s low-loss photonics chips halve lasers in 1.6T optical modules, supporting a $680 million annualized photonics run rate and customer wafer prepayments through 2027.

TSEM $19B
CV's Substack 7 clicks

From X

@convequity 6 clicks

Convequity on X

Nvidia's ($NVDA) own new programming language makes it easier to move AI code to rival chips. In December 2025 Nvidia launched CUDA Tile. In a tile language the programmer describes work as blocks of data and the compiler works out how to run it on whichever GPU it is given. Our read: Nvidia no longer writes code around the CUDA core. Code written for one GPU generation increasingly breaks on the next, so a higher-level language was the only practical way forward. Tile code moves easily to other AI chips. On 30 September DeepSeek released an Ascend version of TileLang, the open tile language it uses, so every TileLang operation in its training now runs efficiently on Huawei's chips, not just Nvidia's. Few investors have noticed. Our view: once the code that matters is written in tiles, the switching cost that supports Nvidia's pricing power falls.
@StableBread 2 clicks

Fajasy on X

Only part of the $1.0-1.2B Boost Run $BRUN expects to spend on GPUs in 2026 is paid in cash up front. $BRUN plans to pay for the GPUs with (1) customer prepayments and operating revenue and (2) equipment leases that spread the cost over the lease term and generally end with an option to buy the GPUs outright. During H1 2026, lessors (the companies that buy the GPUs and lease them to $BRUN) funded $268.3M of new GPU equipment through finance leases. $BRUN made $39.4M of advance payments to secure those leases and paid $29.6M in cash for equipment it bought outright, so lessors funded 79.5% of its $337.3M of new GPU equipment. The $268.3M came from 34 separate GPU leases with 30- to 36-month terms, and the June 30 lease schedule implies an interest rate of ~7%. Its finance leases as of June 30 call for $103.5M of payments in 2027 and $99.8M in 2028, then just $12.1M in 2029, about when its three-year contracts end. So the lease debt runs off with contracted revenue, not with renewals $BRUN still has to win. Customer deposits, the payments customers make up front on their contracts, were $128.4M as of June 30. New deposits made up $112.9M of H1 2026’s $114.2M of operating cash flow, while $BRUN paid $99.9M on its finance leases over the same six months. Effectively, customers’ prepayments are already paying for the GPUs they rent. Founder and CEO Andrew Karos said on the Q2 call that every deal $BRUN closes comes with a prepayment, averaging 22% of total contract value (TCV). The June and July orders alone carry $93.6M due by their start dates ($37.9M on November 30 and $55.6M in installments through March 2027). Karos also said $BRUN’s TCV-to-capex ratio, its contract value divided by what the GPUs cost, keeps rising to 1.4 and above as contracts get longer. So a three-year contract brings in ~$1.40 for every $1.00 of GPUs, which also has to cover the lease interest, the rent, and the staff. In other words, a three-year contract mostly covers the GPUs and the cost of running them, so how much $BRUN makes on those GPUs depends on renting them out again after the leases end in 2029, without lease payments, at whatever older B300s fetch by then.
@convequity 1 click

Convequity on X

SpaceX ($SPCX) targets 13GW, close to our own estimate. In August 2026, our calculation put SpaceX's visible compute capacity over the coming years at roughly 13GW. On the Q3 call, Elon Musk targeted 13GW online by the end of 2027. The numerical match is encouraging. Musk's target concerns terrestrial compute, while our work examined the longer-term orbital opportunity. It supports our view of the scale SpaceX can pursue, although the deployment path and timing differ. We see 13GW as an achievable first stage of the buildout. SpaceX's ambition to manufacture 100GW of solar panels annually never meant all of that would become operating AI capacity in the near term. Our calculation focused on how much capacity we could see a credible path to deploying. For orbital compute to reach 20–30GW and keep growing, solar manufacturing becomes much more consequential. SpaceX needs high-efficiency cells produced at scale in the US, or an alternative supply route that gets those panels integrated into AI satellites. More power from each kilogram of solar hardware means more compute capacity for the mass launched. Starship has now demonstrated orbital payload delivery. In our view, that clears the fundamental launch hurdle and gives us a high degree of confidence that the first 10–20GW can reach orbit relatively soon, with or without rapid reusability. Much larger deployments require rapid reuse. Flying the same vehicle repeatedly reduces the number of new rockets SpaceX must manufacture and allows launch capacity to grow with a smaller fleet. Solar production and launch cadence then have to expand together. We think SpaceX will get there. The pace of high-efficiency solar production, satellite deployment and Starship turnaround will determine how quickly it can move beyond the first tens of gigawatts. #SpaceX

Investor Letters

All letters

Deep Dive: $ASML

ASML’s €230 million EUV machines underpin advanced AI chip production, with demand booked through 2028 after nearly $9 billion and two…

ASML $616B
Rebound Capital 7 clicks

Margin Call

Petrobras’s Morpho-1 well revived hopes that Brazil’s Equatorial Margin can become a major new oil frontier after hydrocarbons were…

PBR $96B
Doomberg 7 clicks

Our Views on CoreWeave

CoreWeave's financial statements point to a business model familiar to the authors after a closer review of the AI infrastructure company.

CRWV $34B
WHAT ARE WE DOING? “Contrarians At The Gate” 11 clicks
TCEHY · long

Tencent Holdings LTD

Tencent, valued near $500 billion, is a long idea built on its Chinese gaming, social-network and advertising franchises.

TCEHY $449B
Value Investors Club

Capital Wars

CoreWeave’s $35 billion to $39 billion 2026 spending plan relies on debt as unsecured yields reach 15%, making its GPU cloud expansion a…

CRWV $34B
Net Interest 8 clicks

Putting bear cases to rest

Medpace authorized $1 billion of buybacks through 2028 after Q3 revenue rose 8.5% to $716 million; service revenue grew roughly 12% as…

MEDP $18B
Best Anchor Stocks 12 clicks

Publications we follow

Every publication on this site, hand-picked.

129 publications

Newsletters 109

YouTube 4

Blogs 15

Seeking Alpha 61 of 251

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.