You and I posted some topical market content today but I also posted quite a bit of dirt. I apologize for that. To be direct, I thought you were leaving or I would not have soiled your thread. I put out my shingle to see if anyone wants to discuss semis, AI, oil, energy, real estate, or financial. These are my primary areas. I will check every few days for a month, or so. As much as I constantly read and talk semis, they aren't my strongest area. Truth is, I don't believe the entire field is knowable by anyone. I've heard experts say things that were objectively wrong and I don't believe they were saying it for the purpose of deception. It's a gigantic dog pile of people going in similar, but slightly different, directions. There is also the conspiracy theory of AI which has gotten so big it now controls broad market swings. I believe this is almost certainly true. The Russians slowly gained control over the NYSE, starting in the 1990s and progressively more into the 2000s. I assume that means Russian mafia. It is not inconceivable that a small group of folks are oscillating the market while repeatedly harvesting the tick. It seems far more likely than not, to me. BTW, A few months ago I had SpaceX in 4th place to win the AI race to power. I now consider them 3rd place for likely dominance. Agentic and OpenAI, meanwhile, just pooped the bed yesterday.
In case someone is interested in semiconductors, I will comment on IBM's claim to have produced a 7 Angstrom chip. Analysis: Odds of it being true -> 0%. Odds of there being some truth to it -> 90%. Odds of it being entirely false -> 10% First, IBM doesn't have a production fab. They outsource. They do have a research fab. Next, it wouldn't surprise me if Intel has some samples from a 7A process node. Node advancements are a pipeline. They take years to evolve and stabilize. The first lithography will be a proof of concept of chemistry, optics, or whatever. They will iterate that slowly over a long period, trying to figure out what the hell just happened. They make wafers using masks GloFo used to call "Structures". Other fabs probably use a different term. They don't start printing working circuits. They make a big lattice work of semiconducting substrate and then examine the living shit out of it it under a scanning electron microscope or, more likely, transmission electron microscope. It's designed to highlight process accuracy and issues, not to semiconduct. Once they have something that passes their various test for conduction, dielectric, etc., they begin the next 99% of the process. I promise IBM is not making transistors for stereos that fit on a boron atom. The most they have is likely a wafer of structures that holds promise for a node that won't be ready for 6 years. This is a small flex on possible semiconductor chatter.
On the other hand, if someone wishes to discuss the type F-A personalities that drive the oil field service industry, that's cool too. There are a few areas in this zone of industry in which only a dominant mo-fo can succeed. When an MBA takes over an oil field company, I look for an exit like the building is on fire. To be honest, I look for the door when any company is taken over by an MBA. There have been successful CEOs with MBAs but they are rare. They generally run companies into the ground and then move to another company to do it again.
Tom, seeing you are into the oil thing, have you watched or looked into FET? I was in them for a while from about $10-$20 but got out when it lagged around and dropped off for about a year. Seems the market finally been rewarding them but sadly I’m not in it anymore. Still likely has a lot of room to run. I work for one of the major refinery companies, won’t name it but it’s about the biggest that don’t also drill oil which would help figure out which one lol. They’ve been on fire stock wise.
What a small world. I watched FET for many years but have never owned them. They're in my watch list as one of the companies that might end up in Ukraine. At the moment, I'm considering a blanket strategy to pick up a few service companies, likely including FET, once Putin slips and falls off a balcony. I wish I knew more about refining but I'm no chem engineer and I have no visibility into that industry. I owned Valero Energy for a while, starting in 2012. The books were amazing. They started boosting the dividend like crazy. I wish I had been smart enough to hold that company. I've never owned your company but they are in my watch list more as a data point than a company I am actively interested in. Anything you might share about how the industry works would be golden words.
Like every industry, it's extremely difficult to get honest information out of the Ukrainian war but I think Ukraine will control the Donbas in 2026. I speculate it will be earlier than later. Meanwhile, it looks like Alexey Dyumin is one of the front runners to take the thrown when Putin springs from this mortal coil. He is just one of many possibilities. There is no possibility of a pro-democracy leader taking power. Every possible replacement is an asshole. The most that can be hoped for is the next Russian president may withdraw from the Ukraine war. This is likely but not certain. That doesn't mean the Donbas will be a hot zone of oil industry in 2027 but it's a strong possibility. After years of watching, I believe we move closer to time to move on oil service companies.
I will mention, the issue I have with FET is they are a small cap company. It appears to be a well run company, although I know less about it than some other oil field service companies, but they will absolutely not 4x on Ukrainian service contracts. The real money will be made in the micro cap space. This is also where real money will be lost. lol!
I'm no expert in Russian politics. Alexey Dyumin seems the most likely to succeed Putin but I'm tracking 7 people who have been mentioned as Russian power brokers. It's both possible and likely, 6 of those seven people will die around the time Putin does. The reason I mention it is, I feel it's important to Russia's future. I see 0% chance of Russia thriving post Putin. They will probably do better, for a while, but it is not going to be a situation in which the population is free to thrive and prosper. It will continue to be an absolutist regime that is obnoxious and sews the seeds of evil globally. Europe will no doubt immediately restore full economic relations with Russia, when Putin dies. They are some of the dumbest leaders on the planet so it likely will take them some time to realize the chef may have changed but the same gruel is coming from the kitchen. On the other hand, this war seems to have put Ukraine through a metamorphosis that I believe is likely to lead to Ukrainian prosperity. It seems clear, Ukraine will become a substantially bigger player on the global stage. What they have done is admirable and astonishing.
I’m too far down the totem pole to have any really good refining info. But, I can say it’s a tough business that takes spending ALOT to earn a little in return percentage wise. It’s just a volume game. since Covid, the profit per barrel of oil has been increased by a lot. Double or more than any other time in history really. Several obvious reasons. Probably 1M barrels of supply per day was taken off the market by smaller weaker refineries shutting down, and still there’s some closing the doors in the hard to operate zones such as California. The wars also increased demand ALOT. So same or greater demand with less supply equals money. I don’t know about the others but I’m sure it’s the same, near all of upper management gets some sort of % of their pay through stock options and such so they are very in tune with pushing the stock prices up. I’m not so sure that’s a great plan “forever” but it so far seems to be working. outside of market swings, it’s hard to boost profits in the business since it takes tons of money to build units, or upsize them for more capacity. Everything gets built with a rate limit and to change that takes big investments. However, there’s 100s of little things that can and are being done to make money. There’s alot of little things that can be done relatively cheaply where the return on investment is made back in a short 1-2 years, and that is currently where the focuses are it seems to me. i like to look periodically at the change in outstanding shares, seems a lot of the stock price boost definitely relate to that.
Here's a thought on the SpaceX $17B terafab investment earmark. This amount of money could mean a number of things. It's too little for a fab less than 2nm plus advanced packaging. It's way too much for a 16nm or 64nm fab. - Many sensors are made on ancient fabrication gear. Old, 200mm wafer, gear is ideal for a lot of this work. There are quite a few sensors that don't work well at a smaller feature size. - It would make sense for SpaceX to start at a more course node pitch but that is not at all how Elon has been talking. Personally, I think Elon is spewing lunacy, except for the possibility he undoubtedly plans to use AI to accomplish the heavy lifting of chemistry and process. - Any AI processing is going to absolutely require 2nm and below. I recall Lip-Bu Tan mentioning SpaceX has contracted Intel's 14A node. That will be some time in the future, as it's debatable if even 18A is ready today. - Chip fabrication at 2nm and below requires special construction with ridiculous specifications for thermals, vibration, and size. These are not cheap buildings. This is where I begin to speculate like a woman's tea circle. It is pretty much certain, SpaceX will build a 20A->18A->14A capable fabrication building. I suspect things get crazy below 14A but nobody is talking about how to do it. The building will probably be in the range of 1.5B and that's just the first of many. They will also build some buildings nearby which are not certified for that specification, probably well before they break ground on the fancy 4 story dreadnought class fab(s). The low spend is probably attributed to Tesla doing some of their own engineering and construction. Tesla has an engineering division that does a lot of their own building. There will also be contractors, though. I'm confident Elon plans to build some of his own automation equipment at Tesla Automation (formerly Grohmann Engineering) in Germany. There is 100% chance they are working hard on it, right now. It seems certain, SpaceX will acquire high-NA lithography equipment from ASML. I'm pretty confident, this is a mistake. They could make a lot of great CPUs/micros/controllers/ASICs with used 6/7nm equipment similar to the gear Micron just sold. A more course node would get production online a whole lot earlier. Maybe it doesn't matter. It's possible Elon is worried about being shut out of leading edge nodes, hence the impetus for the entire project. Whatever the case, Elon is light years ahead of me on this so I will watch closely. I can't imagine SpaceX will start with advanced packaging so they may be able to afford a couple of high-NA EUV lithograhy machines. Advanced packaging is very difficult to do. They can buy the hardware but the toolchain is problematic and the dies are prone to fail due to warpage. Aligning two dies within 20~50nm is hard but the real problem happens when heat causes the dies to bend and disconnect from each other. TSMC recently announced they have 98% yield from their advanced packaging process. That means they are taking multiple dies which have 50~90% yield each and then losing another 2% when they attempt to bond the dies together. It's probably still a significant yield gain over monolithic dies but it's not a free lunch. A monolithic CPU or GPU complex may have 30% yield but consider: CPU core: 2nm CPU/GPU 50% yield Memory controller: 4nm memory controller 85% yield 3D cache: 7nm 3D cache 92% yield Adv. Package 1 bond core to MCL 98% yield Adv. Package 2 bond core to MCL 98% yield Total yield = 37.6% The need for leading edge power drastically reduces production. I speculate I may not see a node smaller than 14A in my lifetime. It simply isn't needed. We are absolutely busting our ass to create machines that will decide our fate.
In case it's not obvious, the reason for the mad rantings above is to figure out who's lunch SpaceX intends to eat. Of course, intends is one thing and eating is another. Still, who is at risk. 2nm = CPU/GPU 3/4nm = NPU/TPU/MCL/CPU/GPU 6/7nm = DRAM/MCL/NPU/TPU 12/16nm = NAND/NOR (SSD) but SSD requires a massively different process than traditional semiconductors 64nm = sensors All of these nodes are useful.
From the SpaceX earnings call, Elon said he expects to be operating Starship at the rate of one flight per day in one year. I doubt any of us believe that timeline but if they can accomplish this goal but it indicates Elon plans to hurl Starlink v3 and/or Starmind AI satellites into low earth orbit in one years time. Considering they haven't broken ground on Starfactory in Florida or the Gigasat factory in Texas, this sounds comically optimistic. On the other hand, I don't know the capacity of the factory and operations center in Redmond, Washington but if Redmond operations can handle the volume, why would they build two more factories to manufacture satellites? Whatever the case, they're going to need a lot of integrated circuits. I understand Tesla has contracts with Samsung, TSMC, Intel, and Micron. There are probably others but I'm not aware of them. If I were to guess, I would speculate they also have contracts with Siemens, NXP, and TI. This is wild speculation. I'm going to guess SpaceX either deals with the same companies as Tesla or has similar contracts with the same companies. Considering the Starship grid fins are actuated by a Tesla Model 3 motor, Tesla makes the battery packs for Starship, and Tesla buys stainless steel from the supplier that makes stainless steel for Starship, it seems a reasonable assumption there is fluid cooperation between Tesla and SpaceX.
AMD just acquired Taalas. The Taalas is a specifically designed around Meta's Llama AI model. I'm not familiar with Llama at all but I'm blown away by the work Meta is doing. They are clearly ahead of xAI, in several key ways. In fact, they may be leading the industry. It's either Meta or Google, with their narrow AI implementations, who currently dominate. Meta and Google will not be win the race to AGI with their narrow AI implementations. Taalas claims their HC1 is 50x faster than nVidia's just released B300 while being 68 times more energy efficient. This sounds ideal for sending to space. There can be no doubt, SpaceX is interested. It wouldn't surprise me if AMD bought Taalas specifically at the request of Meta and possibly Tesla. As much as this could wildly simplify satellite configuration by enabling a relatively tiny solar panels and radiator, the HC1 will become obsolete as Llama evolves. It's hard coded specifically to Llama. Perhaps AMD will create a custom version for Grok. That would be amazing. The downside is, they will have to hurl new hardware into space periodically when the Grok software changes.
Perhaps a shotgun valuation will hold some interest. SpaceX, $SPCX, Friday's range: 114.56 - 133.48 13.18B shares $115 Market Cap: 1.516T If we include long term debt, cash, and equivalents (IOW, Enterprise Value), SPCX EV = $1.43T At $82.15, SpaceX will be at roughly $1T of enterprise value.
