Monday, November 9, 2026

@valuedontlie Oct 15

Colin King on X

$GME up 50% since this post as insiders scoop up shares hand over fist

@AsifSuria Oct 15

Asif Suria on X

@RAMCOCAPITAL That is exactly where I landed! I can see how this would appeal to Buffett even if Greg couldn't get his head around it. Although like the $SNOW IPO, maybe it was someone else who made the decision.

@convequity Oct 15

Convequity on X

We think the market is pricing in too steep a growth slowdown for $AVGO. In Q3 FY26, Broadcom's AI semiconductor revenue reached $16.7 billion, roughly 56% of its $29.6 billion total revenue. In the same quarter, AI revenue grew 221% year over year, compared with 86% for total revenue. As AI accounts for more of the business, it has a greater influence on consolidated growth, so models that assume Broadcom's growth must slow each year risk missing this change. Our valuation work suggests that the market is pricing in revenue growth of approximately 66% in FY26 and 62% in FY27, followed by a 20% compound annual growth rate in years three to seven. Our view is that this assumes growth slows too quickly. Our base case gives more weight to the customer pipeline and management's outlook for the AI business. On that basis, we estimate Broadcom's total revenue could grow by 100% in FY27, compared with the 62% the market implied, with growth remaining well above the market's assumptions as we approach 2030. We increased Broadcom's portfolio weight from approximately 2.2% to 3.4% because we think the market underestimates both how fast Broadcom can grow and how long that growth can last.

@finphysnerd Oct 15 1 click

Myles on X

@stockgutter The only problem I see with this is that $SKHY is trading at such a premium to hynix that even if the Korean listing stabilises, you could get killed on a reversion to fair value here.

@convequity Oct 15 2 clicks

Convequity on X

Is $NVDA at 18x forward earnings a bargain or a value trap? Its GPU is turning into a TPU. Tile languages and AI-written code are weakening CUDA. On OpenAI’s new chip, AI-written code ran up to 1.8x faster than code written by experts. And as AI labs build their own chips, we think Broadcom is the bigger winner.

@convequity Oct 15 4 clicks

Convequity on X

We think hardware manufacturing and chip costs, more than launch, will decide how much $SPCX earns from orbital compute. Our SPCX model estimates how much orbital compute can be made and launched, what it costs and what it earns as it ages, then adds that revenue to our forecasts for SpaceX's launch, Starlink and Cursor businesses in our SPCX DCF. Each year, deployment equals the gigawatts of solar and AI computing hardware the supply chain can make for orbital data centers, capped by what Starship can carry. In our base case, launch is the tighter limit only in the first four years of the forecast, and manufacturing is the limit after that. Our base case puts chips and racks at about 80% of the cost to build each gigawatt and launch at about 11%. Solar arrays, satellite structure and laser communications make up the rest. Because chips make up most of the cost and launch only a small part, whether SpaceX uses merchant GPUs, custom chips or makes its own chips at Terrafab will matter more to returns than how cheap launch gets. Revenue per gigawatt depends on the business model, and we assume it is higher for selling AI services through an API than for renting out compute. Revenue falls as the hardware ages, and we assume each satellite is replaced after a 7-year life. The model's price-cut settings let us test price competition. We think Chinese companies could use their own launch and computing infrastructure to offer cheaper AI services internationally, pressuring the prices U.S. AI companies charge and what they can pay for capacity. The U.S. and China could also tie access to their AI services to defense and trade relationships. Countries with ties to both, including India and some Middle Eastern states, could keep access to U.S. and Chinese providers and negotiate lower prices.

@BrokenMoats Oct 15 3 clicks

Broken Moats on X

Have to at least acknowledge rate reversal and move into low quality and retail/software today...... But oil/diesel/soft commodities/transportation all staying bid, find this more positioning over extended om rates then durable change (especially with the data and the still unresolved issues causing the inflation) AI sold off I think is what is driving both moves in rates (more out needs a home) and retail/rate sensitive/software were the funding shorts (see sectors and components today) ......but geez at these prices an easier trade I think is to fade software/retail then guess next stop on AI and funding.....the are correlated (for now) anyways. Go through the sectors and components and sort by % gain today almost stacked ranked from worst business (biggest intraday gain) to quality at the bottom.....tells you whats going on. Some might say low quality always leads on a reversal.....this is not a start this is after a stealth sharp move higher for $ARHS $ETSY $EBAY $KSS $SHAK $WMT $CBRL $APPN $DOCU $IT $FRSH $BSP etc.

@StableBread Oct 15

Fajasy on X

If the AI model Nubeva Technologies $NBVAF (TSXV: $NBVA) is training to read malware works, it could also take over the one step in its ransomware recovery software that engineers still do by hand. Nubeva’s Ransomware Reversal software, or NuRR, copies the encryption keys during an attack, but someone still has to figure out how ransomware uses them before Nubeva can build a decryptor (a program that reverses the encryption). AI agents have run NuRR’s routine support (bug fixes, support tickets, and product maintenance) for some time, CEO Randy Chou said, but Nubeva’s engineers still handle each live ransomware case by hand. Nubeva’s first granted patent finds the encryption method by running the ransomware in a sandbox (an isolated test computer) and matching the result to a library of standard methods. But the sandbox method breaks down in two cases Chou described, since homemade encryption (encryption ransomware authors write themselves) won’t match any standard method, and malware that detects the sandbox does nothing. A model that reads the malware’s compiled code could get around both problems, since (1) homemade encryption still shows up in the code as steps a model can follow, and (2) the malware’s checks for a real computer only run when the malware runs, so reading the code never triggers the checks. With NuRR supplying the keys and the model working out the encryption method, Nubeva would have both inputs a decryptor needs without an engineer taking the ransomware apart by hand. Frontier models, the most advanced AI models from labs like OpenAI and Anthropic, can already work out what some binaries (compiled programs) do just by reading their code. Anthropic said in April 2026 that its Claude Mythos Preview model can take “a closed-source, stripped binary,” meaning one with its function and variable names removed, and reconstruct “(plausible) source code for what it does.” Obfuscation (scrambling code to hide its purpose) is harder for a model to get past, since compressed malware keeps its real code hidden until it runs. Chou blamed obfuscation for the ~70% of malware binaries frontier models can’t figure out, so Nubeva’s model still has to prove it can do better. Chou didn’t connect the model to NuRR in his October 5, 2026 interview, but working out the encryption method is the same step Nubeva’s FY2026 annual filings say its AI decryptor builder will automate. Therefore, if the model works on live ransomware, NuRR could take on more cases at once, easing the capacity problem that cost Nubeva its biggest customer in 2023 (risk advisory firm Kroll).