New Ask Claude or ChatGPT about any stock, using Bullpen. Bullpen in Claude or ChatGPT. 7 days free
Bullpen.News From Edwin Dorsey
Get Daily Recap Feedback

Wednesday, September 30

Wednesday, September 30, 2026

Everything we published on this day.

76 stories — the front page that day →

@aleabitoreddit 1 click

Serenity on X

Bro, $AMZN semi ETF portfolio is probably going to end up more compelling than what $NVDA owns... Today Amazon signed a $49.8M private placement into PCB maker Gold Circuit Electronics (2368). So now Amazon has ownership/warrants in: - $AAOI - Alchip - $MRVL (Celestial) - $ALAB - $CRDO - $JBL - $FLEX - $STM - $CBRS - $GNRC - $FN - TD SYNNEX - $QCOM - Gold Circuit Electronics And others.. If we look at the private placement amount for GCE in speciifc, it's like .36% of the company... Kinda small financially, meaningful strategically. Esp. since there were claims that GCE supplies ASIC-server PCBs to the four major hyperscaler CSPs... Regardless it's pretty funny to see Amazon to Nvidia start to own all the worlds leading companies. Since Amazon has exposure to the entire supply chain like PCBs now, optical fabric with Celestial, lasers with $AAOI, assembly from $FN to $JBL, ASIC partners like $QCOM / ALchip.

Zentek Research Note

Zentek shipped its first ZenGUARD air-filter order to a Canadian federal facility as its Albany graphite PEA confirmed a US$3.85 billion…

ZTEKF $49M
Atrium Research - Equity Research 1 click

The Next Consumer is a Computer

Meta’s Muse illustrates how persistent AI agents shift compute from per-query to per-user, driving heavier GPU, CPU and memory demand.

META $1.6TNVDA $3.9T
SixSigmaCapital: Navigating the Financial Markets 6 clicks
@convequity

Convequity on X

$PANW and $DDOG both put something real inside the customer's systems that an AI lab cannot recreate by calling SaaS APIs. Palo Alto's install is still harder to replace with AI. 1. Datadog's agents mostly sit outside the traffic and watch production. A bad agent version rarely takes the business down the way a bad firewall rule can. 2. Palo Alto's firewalls and SASE cloud gateways sit inside the traffic and decide what to allow, deny, decrypt or inspect. A wrong allow or deny hits live traffic and live machines. 3. Many Datadog users start the day in a coding agent and open Datadog when something looks wrong. Many Palo Alto users start in its console and stay there through alerts, triage and response. For them, an outside AI agent is an add-on. Cybersecurity work splits two ways on AI risk. Work done before code ships ("shift-left") or from outside the traffic, such as scanning code, checking cloud settings and sorting collected alerts, gives answers that are easy to check. AI agents are already taking over more of it. Work in live traffic and on live machines ("shift-right") means judging whether an attacker is still inside and whether blocking them is safe. That is more ambiguous and mistakes cost more, so enterprises move slower and demand earned trust. Palo Alto is not risk-free. In our view its exposure is the easily checked detection and vulnerability work inside its platform, which AI agents can take over sooner. The threat is not an AI-native firewall startup.
@ActAccordingly

PAA Research on X

Following the action in $FICO, it seems this post from @pulte has drawn increased investor attention to $STC $FAF $ORI $FNF and their title insurance pricing policies. We have had a short call on $STC and $FAF for some time. There's no doubt there are several technology options that could lower title insurance costs (AI) if not eliminate the need for it altogether (blockchain). You already have many brokers offering bundled solutions for title, mortgage, and advisory. Lots of pressure on these busineses.
@convequity

Convequity on X

Anthropic's biggest near-term threat isn't the S-1 or AI-safety language. It's competition, especially an OpenAI price war. The S-1 is public, and the timing lines up with OpenAI DevDay. My read is that it isn't a coincidence. That's my read, not something either company has said. What surprised me more came out of DevDay: OpenAI is ending its price war. Here's what that means in practice. OpenAI removed the Pro 20x subscription at $200/month and plans to cut usable tokens on that $200 plan by about half. The allowance was also more generous than it looked. Over the past quarter, OpenAI did roughly 2.5 manual usage resets per week. So heavy users didn't only get a big allowance. It got refilled before they hit the wall, over and over, for about three months. That's a price war fought in tokens instead of list prices, and it is expensive to fund. Why it mattered for Anthropic: for heavy users, cheaper tokens beat a better model. Once a model is good enough, what limits you is how much work you can push through it per dollar. A better model you have to ration loses to a good-enough model you can run all day. Anthropic went the other way and raised margin. That's the right posture for an S-1. It's the wrong posture when a competitor is paying for volume. That left Anthropic in a trap going into the IPO: - Match the pricing, and margin thins just when margin is the story you're selling. - Hold the line, and growth slows as heavy users drift to wherever tokens are cheapest. Either way the numbers get worse at the worst possible time. Ending the price war takes that trap off the table for now. Anthropic gets a cleaner tape into the IPO. Why did OpenAI stop after one quarter? I see two reads, and both are mine, not confirmed: 1. The burn. My read is that OpenAI postponed its own IPO because of how much it was burning, and investors pushed it to stop. 2. OpenAI's own later IPO. If it wants to list, its numbers have to look like a business too. One quarter of subsidized tokens is a campaign. A year of it is a problem in your own filing. The second-order point is that in both reads OpenAI stopped because of its own finances. Anthropic didn't win the war. It ended because the side paying for it chose to stop. The playbook still exists and it ran for a full quarter. It can be turned back on. That's the risk that remains. OpenAI restarts it. Or a rival like $SPCX xAI ships a next-gen model and discounts tokens hard to buy share. It's a bit like Uber's cash burn: a well-funded player can keep subsidizing longer than a margin-focused company wants to wait. I don't think that's fully priced in.
@convequity

Convequity on X

$SPCX's Starship reached orbit and delivered cargo. The gigawatts still depend on how often it flies. We have been building the orbital-compute case with three key inputs: kilowatts per ton, tons per flight, and flights per year. 1) Ship 41 reached orbit, about 275 km, and released all 26 Starlink V3 satellites. They are in service. Until this flight Starship had only made short hops and come straight back. 2) The flight was shorter than the plan. One of six upper-stage engines shut down early, and the ship came down at the first health check after deployment, about three hours after launch, instead of staying up for about ten hours. Both stages splashed in the ocean. That was the plan. The ship can take a hard splash and be scrapped. It still delivered a commercial payload. Catching it is how later flights get cheaper. 3) A typical 1 GW site on the ground loses about 370 MW to power conversion, cooling and other overhead that is not the chips. Elon has said terrestrial solar produces about an eighth of the power of the same panel in space: no night, and no atmosphere in the way. In orbit most of that 370 MW overhead goes away as well. Sunlight is continuous, and the heat leaves through radiators. 4) On our Special View assumptions, year 4 is 200 tons at 100 kW per ton, which is 20 MW on one flight. Five hundred flights is 10 GW. Year 5 is still 20 MW a flight. Five thousand flights is 100 GW. The tons and the kilowatts stay the same. The flight count is what changes. Flight 14 showed Starship can reach orbit and deliver cargo. The 10 GW and the 100 GW still depend on how fast SpaceX can fly it again, and at what price.
@BrokenMoats 1 click

Broken Moats on X

Someone really should investigate the price action into some of these ancillary biotech stocks.....very reminiscent of a certain Situational Awareness with chip stocks earlier in the year. No one is filing ownership and yet these stocks push higher regardless of macro, or geopolitical news or the overall index moves. Someone with swaps? or options likely at play then using the momentum pf retail and algos to add to the bid. The trade is melting up perfectly to the end of the quarter (not surprised to see them even more aggressively up today than usual to mark performance). We saw this with the chips trade after an unprecedented April-June. But some of the moves with this trade exceed that euphoria (Look at $TXG, $TWST, $SDGR, $RGEN, $GRAL, and a host of other companies)........DO we see the similar move after a giant quarter with a reversal to kick of the next quarter, that was the fate with chips and caused the fund to blow up. Be interesting to see if they learned any lessons, Im guessing not based on recent performance, and if it is a different player the strategy looks the same. These companies need financing and should hit the market bid which is what set off Viacom and was the undoing of Bill Huang in 2021. Lets see, but a basket of these stocks looks interesting at current levels sometime today for gravity/secondaries to come early in Q4
@david_katunaric

Mikro Kap David on X

No one earned this drawdown more than the growth-at-any-price bros paying >100x trailing P/E for $FICO at the 2024 high...with revenue at record levels and net margins having expanded from ~LDD% to ~30% in just six years. If you're even slightly off on any of the ambitious assumptions you must have been making around market position, pricing-driven growth, or the exit multiple, you lose A LOT of money. That said, if there’s a definition of a business worth paying up for, FICO is probably it. Or was... interpreting what the recent changes to the mortgage-scoring landscape mean is no small feat for a tourist like me.

Microsoft: The Xbox Segment Revamp

Microsoft plans to rebuild Xbox after FY2026 through studio cuts, higher hardware prices, Game Pass growth, exclusive releases and the 2028…

MSFT $3.9T
Khaveen Investments 2 clicks
@DOMOCAPITAL

DOMO Capital Management, LLC on X

$PBI

💥Yield!💥

Beyond Meat’s Q2 2026 revenue fell 8.2% as U.S. retail and foodservice volumes declined, while a flatter, not inverted, Treasury curve…

BYND $101M
PETITION
@convequity

Convequity on X

Judging Tower Semiconductor ($TSEM) by how advanced its chip process is misses what its chips are for. Skeptics often set Tower's older 65 nm process against the more advanced processes at TSMC ($TSM). That yardstick suits logic chips, which are judged on transistor density. It does not suit the analog chips Tower makes for optics. 1. In an optical engine, the analog front-end is the circuitry that drives the laser and reads the returning light. It needs speed, linearity and power efficiency, not maximum transistor density. 2. Tower's strength is its silicon-germanium BiCMOS processes. They have long dominated the merchant market for these laser-driver and receiver chips because they deliver high performance without the most advanced CMOS logic. 3. A less noticed strength is hybrid bonding. Co-packaged optics needs an electronic chip bonded reliably onto a photonic one. Tower already runs mature, high-yield hybrid bonding for stacked image sensors in high-end phones, where a signal-processing chip is bonded onto a light-sensing chip. A close technical parallel, already at volume. The limit: if the electronic chip has to take on more complex signal processing, such as clock recovery or adaptive equalization, TSMC's advanced FinFET processes do offer real advantages in power and integration. Different foundries may end up making different parts of the system. So Tower's position in optics rests on silicon-germanium and photonic process expertise, not on advanced logic.

Research Roundup Vol. LIV

Research Roundup covers AAON's data-center cooling ramp, Knife River's margin recovery, ChronoScale's Microsoft-backed neocloud buildout…

AAON $6.6BKNF $3.2B
Monte Independent Investment Research