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

Friday, November 20, 2026

Back to the front page

From X

Serenity on X

$P earnings were actually interesting, not really in terms actionable investments. But more on AI capex broadening, where deployments are happening, and what Jensen at $NVDA might envisioned with open weight model + GPU demand. - Everpure CEO stated that enterprises are increasingly building on-premises AI factories, stating: "The ROI for companies to buy their own GPU environments and run their own models is starting to come to the forefront." - "I'm talking about standard enterprises, regular finance organizations. A wide variety of organizations are starting to buy these AI factories for the purpose of running their own open weight models From "regular insurance companies, et cetera" - "One is they're starting to see that these open weight models are actually being able to provide good results, and they're doing so at a fraction of the cost of using the cloud for these workloads" "that the ROI for companies to buy their own GPU environments and run their own models is starting to come to the forefront" This type of broad industry transition would be heavily bearish as it threatens $ANT, OpenAI and frontier labs over the long term... And probably what Jensen saw things heading with $NVDA + additional GPU demand regardless of what frontier lab wins. However, I expect more of the open weight effort to collapse or see heavier regulation with how Washington + frontier labs were lobbying. - In regards to NAND prices dropping: "If prices come down slowly, I expect to see demand come up along with that" (Jevon) - LOL this part was hilarious and the same argument I was using for why qualification cycle players like in optics were so interesting. On data center slowdown potential: "So on the first part of it, remember, I mean, really, the thing that insulates us is we have such a small share going into the hyperscale market" "even in a down market, they’re going to spend a lot more than our current share amounts to", and their market capture goes up over time relative to overall TAM.

Lee Roach on X

Geospace $GEOS reported tonight and the stock is $2.76 after hours. That is a $36 million market cap for a company that lost $44.3 million last year. I spent the evening in the filings, the export data, the federal contract records and the ship trackers, and I want to walk through what I found. Geospace is a Houston company that makes seismic sensors for oil exploration, plus connectors for smart water meters and some security gear for the government. Revenue fell from $110.8 million to $80.7 million this year. Gross profit was $2.8 million. The business burned $35.1 million of operating cash, which is about the entire market cap, and cash on hand went from $26.3 million to $6.1 million. The reason anyone still owns it is one contract. In June 2025 Petrobras hired Geospace to build a permanent reservoir monitoring system for the Mero field off Brazil. Picture hundreds of kilometers of fiber optic cable with sensors, laid on the seabed two kilometers down, listening to the oil reservoir for decades. It is worth about $90 million, more than the company's whole revenue last year. Management says it sets up a return to profitability in fiscal 2027. So the whole stock comes down to when that $90 million shows up. So far it mostly has not. On a contract like this, revenue is booked as the work gets done, and cash arrives at milestones. Geospace has collected $16.3 million. It booked $5 million of revenue in the first nine months of the fiscal year, and the balance sheet implies about $7 million more in the September quarter. Call it $12 million out of $90 million, 17 months after the award. The timeline keeps moving too. The filings first said the work would be done "over the next 18 months." Then "24 months." Then by mid-2027. On the May call it became early 2028. The next payment, $35 million, was expected by the end of this calendar year. In August that turned into sometime between January and June. That payment is the one that matters, because the company cannot wait long. I wanted a way to check progress without relying on management, so I went to the export data. Every shipment leaving the United States gets recorded by the Census Bureau by product type, origin state, destination, weight and how it traveled. Seismic gear from Texas to Brazil normally goes by air in small lots. From March through September of 2025, 7 tonnes went by ship. This year, from March through June, 209 tonnes of geophysical instruments went by ship. Then July, August and September: nothing. Here is why that matters. The company's own spec sheet says the cable weighs 700 kg per km in water, and Petrobras says this phase is 316 km. I estimate the full system at 350 to 550 tonnes. So something heavy left Houston in the spring, but it was declared at under $20 a kilo, far too cheap to be finished sensor cable, and then the shipments stopped. If the system were on its way to Brazil, I would expect to see it here. I do not see it yet. Even when the cable does ship, somebody has to lay it. That job belongs to a Brazilian contractor called Blue Marine, which has said it expects to finish in mid-2027. On its recent Petrobras projects Blue Marine worked from a cable ship called Bold Maverick. So I looked it up. In May that ship was renamed Eileen Navia and reflagged to Spain, and tonight it is doing cable work in the eastern Mediterranean. Maybe they have another vessel lined up. I could not find one. Now look at what the contract is supposed to carry while all of this plays out. The core seismic market is shrinking. Houston's exports of seismic instruments to the whole world were $220 million in the first nine months of 2024, $164 million in 2025 and $137 million this year. Geospace's biggest land customer, Dawson, bought $24.2 million of equipment and Geospace lent it $18.2 million of the purchase price at 8.75%. Dawson still owes $13.2 million, everything has been delivered, and its latest filing describes no new order. The water meter business was supposed to be the steady part. It fell 52%. The government business lost a $13.3 million border patrol contract that closed out in April and was replaced with one worth $1.9 million. There is a new $10.84 million Navy award, which helps, but it runs through December 2027. Management's answer has been to cut. Headcount went from 519 to 390 and quarterly operating expenses from $12.8 million to $10.6 million. That saves about $8.5 million a year. The loss was $44 million. The job boards tonight show two openings, both for factory maintenance. Put it together and this is my math. The company costs about $64 million a year to run. Outside the Brazil contract it is doing about $56 million of revenue and needs about $93 million to break even, so without the contract it loses roughly $24 million a year. With the contract, fiscal 2027 gets to around breakeven. One year, and only if the timeline holds. The cash is tighter than the earnings. Because the work comes before the payments, I have them borrowing about $24 million by March on a credit line that tops out at $25 million. That line is new. The old bank required $85 million of tangible net worth, and year-end equity was $85.7 million including $5.9 million of intangibles. The loan was replaced in October with one that costs prime plus 3.5%. After the contract ends, the losses are still there. Book value was $9.79 a share a year ago and is $6.63 today. The stock looks cheap at 0.42 times that number, but on my estimates book is near $4 by 2028 unless a second contract arrives. Petrobras is spending $450 million on monitoring at Mero, so the market exists. I looked for a new tender in English and Portuguese and found none. The stock was $29.89 in October of last year. Three things would change my read: a month where Houston's exports to Brazil jump past $10 million, a named ship, and the $35 million actually arriving. I am tracking all three.

Thursday Tailwinds #15: Nov 19, 2026

India’s 236 GWh battery target boosts local electrolyte-salt makers, while Mirion’s radiation monitors and Overwatch’s passive drone…

SBAC · long

Sba Communications CORP

SBA Communications trades at 14.9x forward AFFOPS, near a record low, despite being pitched as the market’s cleanest tower operator.

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,…

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…

LXS.DE: a 2028 sale it can force

Lanxess can force Advent to buy at least half its Envalior stake in April 2028, potentially raising about €745 million to reduce net debt.

The Agentic AI Roadmap for SaaS

SaaS companies including Salesforce, ServiceNow, CrowdStrike, Synopsys and Veeva are ranked by agentic AI progress, from pilots to…

Build-A-Bear: A Narrower Path

Build-A-Bear's push into prestuffed toys and broader retail distribution tests brand stretchability, but recent results warrant caution.

Publications we follow

Every publication on this site, hand-picked.

130 publications

Newsletters 110

Seeking Alpha 98 of 301

X accounts we follow

Every account on this site, hand-picked.

94 accounts

Investor letters we follow

383 firms

Sectors

Ordered by how often our editor reaches for them.