I’d like to start this episode with a thank you to the nearly 300 Intellionaires who have subscribed to this newsletter - I hope you are all enjoying the recent gains in the stock as much as I am. It’s been very rewarding to receive messages of thanks from people over the last couple of days who were inspired by myself & others in the subreddit to go against the consensus and to invest in the stock when it was $17-$22 - during peak FUD - who are now up well over 200%.
It takes serious guts to buy into a stock at peak fear. And even harder, it takes serious discipline and knowledge of the investment to ride the price up without selling along the way. Every price rise you are met with calls from strangers, friends & family to “sell now and take profits before it’s too late”. I heard this at $22, $25, $30, $35, $40, $45, $50 and now I’m hearing the same again at $60.
To make my position clear:
I had a thesis on Intel stock when we started the subreddit, and we are starting to see the early phases of this thesis play out. Intel is now what I would consider ‘fair value’, at approximately 2x book value. The number of comments I see online claiming that “it’s priced to perfection” and “all the upside is gone” are one of the primary reasons why I made the subreddit & the Intellionaire newsletter - to try and counter this gross misinformation.
Little by little, through our efforts, the social media narrative is starting to shift. If you haven’t already done so, please subscribe to the Intellionaire and share it!
Now, before we go on to talk about Terafab (P.S. - a very warm welcome if you are coming from the Elon/Tesla/xAI/SpaceX crowd), I’ll quickly re-cap my Intel thesis:
Intel will become the single most important semiconductor manufacturing company in the US & EU, more important than TSMC, doing advanced logic and advanced packaging (and may also get back into memory to compete against Micron/SK Hynix/Samsung).
Intel will eventually have a thriving custom design business, similar to Broadcom. Their custom design business has only just started and already has an ARR of $1Bn+.
Intel will once again produce leading CPUs, as well as a new business line of GPUs for both consumer and data centre. This competes against AMD, Qualcomm & Arm.
If they continue to execute in Foundry, Custom Design & CPU/GPU, I expect that Intel will attain a conservative valuation of somewhere in the region of $1.5 trillion in the early-to-mid 2030’s, or 5x higher than today’s price. The initial “explosive” growth we have seen over recent weeks and months has been a rapid catch up from ‘profound undervaluation’ to now ‘fair valuation’. I anticipate future gains will be more steady and incremental over the coming years.
Of course, these are my personal thoughts and not investment advice. I put my money where my mouth is, and my entire portfolio is >90% Intel.
It will come as no big surprise to long-term Intellionaire readers that Tesla, xAI & SpaceX are joining forces with Intel for the Terafab Project.
There have been plenty of signs pointing to this over the last 6+ months which we have tracked in detail, both here and on the subreddit. I’m going to recap the backstory, explain what Intel has to offer & where the Terafab might be located - and then give my final thoughts on how this new relationship will play out.
To summarise the backstory for non-INTC nerds:
Intel Xeon is already the head node of choice for xAI/Colossus, so xAI is already built around Intel x86 CPU.
The previous CEO, Pat Gelsinger, was invited on a tour of xAI Colossus - and likewise extended this offer to Elon to tour the Intel fabs last summer.
Elon & the new CEO, Lip Bu Tan, followed each other on X last year and exchanged pleasantries. “Intel new CEO is cool”.
Intel was reported to be doing all the advanced packaging for Tesla Dojo, which was then claimed to have fallen through when Dojo was canned in favour of a unified AI chip.
Elon reports that AI7, “will need a different fab, since compared to AI5/AI6 it’s a more adventurous design” - one could infer this meant they were moving towards Intel as the primary manufacturer for future chips with backside power delivery and/or High-NA EUV.
Random Andy Grove quotes about “only the paranoid survive”, as well as directly mentioning at the Tesla investor day that “we might even do something with Intel”, back when Terafab was first announced.
Now, with that out of the way, let’s continue…
Part 1: What does Elon hope to achieve with Terafab? Geopolitically secure chips, quickly.
“To be clear - we are very grateful for our current partners - TSMC, Samsung, Micron - but there is a maximum rate that they are comfortable expanding, and that rate is much less than we would like. We need to build the Terafab, or we don’t have the chips. And we need the chips, so we are going to build the Terafab”.
These chips will be going into terrestrial xAI data centres, future space xAI data centres, Tesla vehicles & Optimus. Elon predicts that he will need billions of chips over the coming years, and he simply cannot get the supply from his existing partners. The aim is to start with a capacity of 100,000 wafer starts per month, and build from there.
The total number of wafers that all of Elon’s companies consume per month is hard to calculate. What we do know is that he has an existing multi-year $16Bn dollar deal with Samsung for 16,000 wafers per month from the their Texas fab (which he reportedly requested to increase to up to 40,000 wafers per month).
So, he would be looking with the initial stages of Terafab to increase his existing US capacity somewhere in the region of 2-5x compared to what he’s getting from existing partners at the moment. Intel is the obvious only short-to-mid term option for getting this capacity.
Eventually, the aim would be to scale from 100,000 WSPM to around one million or more wafer starts per month - a crazy high figure that would far exceed any current fab output (for reference, TSMC designates fabs that are able to produce >100,000 wafers per month as ‘GigaFabs’).
Separate to the Terafab is the small ‘Advanced Technology Fab’, which they are already building at their Texas site - “we will have all of the equipment necessary to make a chip of any kind, logic or memory, and also the equipment to make the lithography mask. The aim is to have an incredibly fast recursive loop for chip design”.
I have seen online speculation that Intel fabs won’t be used for Terafab since they don’t contain mask, logic, memory & packaging under one roof.
However, I believe this is incorrect.
There is no statement I’m aware of that that suggests the main Terafab site will use this combined strategy. As far as I’m aware, Elon has only confirmed that their Advanced Technology Fab will use this approach in order to rapidly iterate, before then sending finalised design out to a high volume manufacturing site, AKA the Terafab - which would presumably have separate logic, memory & packaging modules.
Part 2: What Intel brings to the table
Firstly, Intel is the only US company with leading edge R&D, for both logic and advanced packaging, in the United States. They have decades of experience making chips and around 40,000 - 50,000 employees dedicated to this in their Foundry division.
They have the most leading edge fabs in the Western hemisphere, and they are currently the only company able to do advanced packaging in North America; TSMC & Samsung have to send all of their completed wafers abroad for packaging, which massively slows chip production & introduces uncessary supply chain risk.
Intel are also the only company with advanced EUV fabs in the EU, something that neither TSMC nor Samsung have - this could be a big bonus to have chip manufacturing sovereignty in both the US & the EU for selling cars/humanoids into different geographical markets.
For someone like Musk who wants a quick leg-up into the chip manufacturing world, Intel is the perfect partner. Not only do they have all of the necessary skills, fabs & employees, they are also open to co-operation - they have actively been separating their Foundry business from their Products business, both legally and financially, over the last couple of years.
In fact, they are openly looking for an outside investor to take joint ownership of the Foundry business. Intel has to retain 51% ownerships of the Fabs, but that means the other 49% is up for sale. This latest announcement has led me to believe that Tesla potentially may be the long term partner and co-investor that Intel have been looking for. Time will tell.
Intel also have a number of unique technological advantages that would be desirable to Musk. I won’t detail all of them here, but in particular their PowerVia 2.0 backside power delivery, which will be used on 14A, may prove to be very useful for space-focused AI chips where there is no thermal convection or conduction. I expect the direct contact backside power will be helpful for heat radiation, especially as it creates backside thermal hotspots which could actually improve heat transfer in space vs. front side power delivery.
Intel also have ongoing Rad-hard projects for space chips with BAE, as well as being leaders in certain niche technologies like Gallium nitride, or GaN based chips - Elon actually liked someone’s post on X which mentioned this!
P.S. for those who don’t know what backside power delivery is, or why it’s so important, please read my previous Intellionaire Ep. 8 that covers it for reference:
In addition to technological advantages, Intel has gained key contacts on its board of directors, including Eric Meurice from ASML - as well as a strong relationship with them as the lead development partners for High-NA EUV.
If Musk wants to use High-NA EUV for any future chip designs, Intel are a country mile ahead of any of the competition in terms of not only accumulated experience, but also the number of High-NA EUV machines available now & on order in the future.
Intel also own the majority of IMS Nano, who produce the Multi Beam Mask Writers that are essential to the production process.
Part 3: My Final Thoughts
I envision the deal will involve Tesla using the Intel 18A-P/14A process node for future AI chips (likely AI7 onwards) & manufacturing these initially at one of the existing Intel locations (probably Fab 52/62). This will deliver some immediate starting wafer capacity.
The Terafab will then be built in tandem at an alternate location, and will licence the Intel technology for a fee once it is ready for high volume manufacturing.
Another possibility is that the Terafab will be a true shared-site JV between Tesla and Intel (more on that below for Intellionaire members). Tesla may eventually become a 49% cash buy-in partner to Intel Foundry using the proceeds from the SpaceX IPO, with full access to all of their technology, experience & manufacturing capacity across the globe.
I expect there will be continued use of Intel x86 CPUs as head nodes at xAI, cementing this already strong relationship further. We may also see Intel chips used in Optimus in some capacity in a CPU role, especially as Intel is really pushing into edge AI & robotics.
I do not anticipate that Intel will get back into DRAM manufacturing any time soon for Tesla, but with Naga Chandrasekaran at the helm (ex-Micron), never say never! I also don’t expect SoftBank will be happy to tie their SaiMemory JV in with Tesla for Z-Angle Memory, but you never know.
Either way, I’m hoping that we hear more information at the upcoming earnings call later in April.
Finally, to the Intel employees - thank you for your hard work. I remain amazed by the technologies that you are creating and manufacturing. As an investor, it is incredibly interesting to read about all of the new advances being made at Intel, in both design and manufacturing. You should be proud of what you do, and the world will come to bear witness to your efforts!
For paid members below, I will cover my thoughts on the Terafab location and how this may intersect with Intel’s existing construction plans. I’m also thinking of designing some Intellionaire merch that will be available exclusively to paid members in the near future - with new releases at different price targets that are achieved - let me know in the comments if this is something you would like before I start on the designs!