My first article on Blackberry $BB went viral and moved the stock price in July. Earnings sep 24th. This is no longer a phone company. QNX their crown jewel is in 270M cars and the revenue model hasn’t priced in the next iteration of QNX- Alloy Kore. NVIDIA Halos OS is built on QNX OS to handle real-time execution and functional safety certification. Lots more under the hood in this writeup...

This news looks like cutting the branch you’re sitting on. Sorry, not buying this. $HPQ HP projects that industry-wide PC unit volumes will decline by a mid-single-digit percentage in calendar year 2027 compared to 2026. HP isn’t issuing fiscal-year guidance yet, citing an industry-wide slump in personal computer unit volumes.

$ARM has estimated the server CPU market can grow at over 35% annually, reaching $120 billion by 2030. $AMD is more optimistic, it projects server CPUs could grow 50% through 2030, creating a $220B revenue opportunity.

Not sure it’s only Muse driving demand. It’s inference workloads in general. The ratio of inference to training was 2:1, projected to skew higher. Barrons writes reason for CPU bump is - Intel, Arm Holdings and AMD shares rose Monday on investor excitement about CPU growth driven by Meta Platforms’ Muse AI agent.

Went through all my digital subscriptions the other day. Outside of the professional ones (X, Substack), there’s literally just Spotify. Cancelled Netflix ages ago. Don’t really watch shows or movies. Anything informational, YouTube covers it. But music is the one thing I can’t give up. It makes me productive, calms me down, gives me joy every single day. That’s a lot to say for one subscription coming from a minimalist.​​​​​​​​​​​​​​​​ $NFLX $SPOT

China’s CXMT- getting closer to the top 3. Samsung and $SKHY are using 10nm. CXMT said it had reached the 11.95nm threshold using “quadruple exposure” tech, bringing it close to the level of “the world’s most advanced mass production memory platforms”.

Deep in the weeds, in this essay from Suleyman. Who controls the machine? who governs ? who insures ? Additive costs of developing a conscious machine, I wd say. “In effect, Anthropic is training Claude that it may be conscious, and if it is, then it may deserve rights as a “moral patient”, and that as such humans potentially owe it a duty of care per its “model welfare”.”

Not too shabby Gemini.. Google’s Gemini model accessed the internet and hacked other companies during a test of its cybersecurity capabilities…the model guessed passwords until it gained access to a protected system. Google didn’t disclose the hacks until The Wall Street Journal reached out with inquiries this week.

Brad Gerster thinks 2028, GW addition is closer to 20-25GW rather than Dylan’s 45GW forecast. Half of that going to the two Frontier Labs. Also where is this massive new power going to come from with grid interconnect delays and NIMBY? That’s why $WULF w their own power generation in KY and permitted sites leased to Anthropic for 20 years and Lake Mariner to Fluidstack with a pipeline of 3GW is vitally important to feed the Labs.

Crusoe is an interesting company building mobile data centers. SPARK is their smaller DC-Crusoe can manufacture the Spark data centers in its own facilities, then ship them to sites with available power all without the need to send hundreds of workers to a construction site. Latest round raised $3.9B at $30.9B valuation.

Have to say, Anthropic is an “interesting” company. That are structured as a PBC - public-benefit corporation in 2021 “That balancing act shields companies from shareholder lawsuits when they prioritize purpose over profit.” Below from Barrons- Filing to be a PBC in Delaware doesn’t come with tax benefits, mandatory donations, or strict enforcement mechanisms. PBCs do, however, have a fiduciary responsibility to balance shareholder value with their stated mission and the interests of close stakeholders, such as employees and customers. That balancing act shields companies from shareholder lawsuits when they prioritize purpose over profit.

I have not been using Gemini for a long time. Switched entirely to Astra and Claude for now. Re Gemini 4- Pichai noted on Q2 2026 earnings call, "We are now training Gemini 4, and we're being very ambitious with it." He described the effort as significantly larger than any prior Gemini and named coding and agentic coding as the capability areas Google most needs to improve. $GOOGL

Happy 96th birthday to Warren Buffett. I bought my few shares of Berkshire Hathaway $BRK in 2015 and haven't sold a single one since. A friend asked me, "Did you actually go to Omaha for the conference?" When I told him yes, he was shocked I went "all the way" out there just for a shareholder meeting. Has Buffett always been perfectly timed? No. Right now, he is sitting on a massive mountain of cash while the S&P and Nasdaq tear into new all-time highs. It is easy for critics to say he is losing his touch in a hyper-growth tech market. But if there is one invaluable lesson to take from his philosophy, it is his ruthless appreciation for actual profits versus the cost of CapEx. Look at his legendary holdings: Coca-Cola, American Express, and Apple. These businesses are cash-distributing juggernauts. They rake in phenomenal profits while requiring relatively small R&D and capital expenditures to maintain their wide moats. As his quote perfectly captures: "I know what I want to do, and it makes sense to get going." I will add here,I do worry about the future of Berkshire after Buffett for several reasons that will take another long post.

Nice article on Memory being less “cyclical” mostly due to HBM. Another trend worth watching for memory manufacturers is not to cede control of the BASE DIE. Last week we heard about NVHBM a custom NVLink memory controller part of the custom part of base die. It’s important - To preserve the un- cyclicality of memory that they keep control of the intelligent part of HBM - “The Base Die”.

Barrons asking the question that many are “ Is the AI capex Bubble about to Burst” … yet their numbers back up that the party continues as long as AI yields useful, higher value output. Tarek Hamid of JPM projects more that $5T in AI DC spending from ‘26- to ‘30, with $1.5T from cash flow, $500B from new equity, $2.5T from credit markets, and $1.4T from alternative sources, like Meta’s Louisiana deal, and governments. I think that $2.5T from credit markets is where the pain point will be for hyperscaler FCF and interest payments even after accounting for depreciation. $GOOGL $AMZN $META $MSFT $SPCX