Q3 2026 Portfolio Update

I’ve been a super lazy writer lately. I’ve been trying to write this bigger post on AI and Robotics and I’ve thrown away like five drafts already. So I’m giving myself an easy softball topic by just combing over all my pre-IPO buy/sell/pass decisions of the past couple months, as well as a couple key updates on major holdings. I don’t know if I’m making correct decisions or not, I’m just tracking my thoughts in and real time and it’ll be fun to revisit later, win or lose. Let’s just get it started, this blog needs some fresh ink.

Pete’s buy decisions

FigureAI — I’m already a FigureAI holder from last year, was just a small flierm but I recently just quadrupled my position at the same price (39.5 billion, the series C price). My viewpoint has changed–I went from thinking it’s a promising but overvalued company to now having some credible “intelligence” on it–I have good reason to believe it’s the leading humanoid name in the United States. They’ve created an amazing company culture where none of the top talents leave. They’ve created excellent hardware. They’re well ahead on intelligence. Maybe the only edge they don’t posess is Tesla’s manufacturing capability for Optimus, but they are ramping that up too. The company definitely has haters, which once made me skeptical but now I’m a believer.

Positron — So, if you read my Groq/Cerebras post, you’ll see I’m already bullish on the inference chip market. As AI adoption grows, inference demand compounds with usage, since every new user and every additional query requires compute to generate a response. There’s been a huge memory bottleneck trade driving the market in 2026, with stocks like Micron and SK Hynix going parabolic because they sell the high-bandwidth memory (HBM) that AI chips increasingly need. The problem is HBM is expensive, and inference at massive scale is going to need something cheaper and more efficient. That’s where Positron comes in. Instead of HBM, its chips use LPDDR, a much cheaper type of memory, with an architecture designed specifically around inference. The basic bet is that Positron can get the memory bandwidth AI inference needs without paying the HBM tax—and its chips are already being used at Oracle, which is a pretty meaningful early validation. I’m getting in at $5 billion post-monery valuation, I think 10x upside is achievable if Positron simply takes a 1% share of the inference market away from the big incumbents like NVDA. Bonanza upside if it can do better. In short: I like the market, I like the pricing, I like the differentiation, I like they are validated at a large hyperscaler.

Pete’s sell decisions

Deep Fission — I’ve sold a lot of my Deep Fission around $11-12. It’s unfortunate but the market was not just able to support the private valuations. First they wanted to have a large IPO with at $21-23 with credible underwriters like Canaccord, William Blair, and Stifel. This would’ve given them a lot more runway with their cash, as nuclear startups burn a massive amount of cash to build their tech and clear regulations. Due to lack of demand, they had to massively downsize to $16 IPO with Benchmark, a fourth-tier player. That’s red flag 1. Then the stock craters below the $15 private placement price, that was supported by Blue Owl. Red flag #2. The public nuclear SMR sector as a whole had been in a massive 50%+ decline since 2025 highs–red flag #3. As a trader, I saw the writing on the wall… broken stock, no matter what the company does. Validation of their deep borehole concept is still at least a year away and probably requires more dilution. So I sold. Still a nice 4x gain in less than a year. I remain a small holder and I wish the company well.

SpaceX — So this opened at around $2 trillion on IPO day and had your classic 3 day parabolic pop. I hedged 88% of my position at $185 average, locking in a 7.5x gain from my $22.4 cost avg. Now I’m just waiting for the shares to be delivered so I can slowly unwind both legs and get to cash again. I went into the IPO thinking it was overvalued and I haven’t really changed my mind. There’s too much of a technology risk here for what the IPO valuation was supposed to be underwriting–the whole space datacenters thing. We don’t know if it that can be a thing yet. Launch, Starlink, Cursor, and their AI infrastructure are still good solid businesses with rapid growth. xAI has fallen behind a bit, but it’s still valuable, even if it only ends up being used as the Elon Industries go-to intelligence rather than a leading frontier lab for consumers and enterprise. It would’ve been a genius move to unhedge at $110 when the squeeze happened in anticipation of a lockup sell-off that didn’t transpite (classic stock market shit right there)–but I didn’t do it. I’m not in the market to be a galaxy brained genius right now. Happy with the gain for now and tightly monitoring for inflection changes in the business or stock chart. Will revisit owning the company long-term closer to lower 1T range (about $80-100 range) if it can get there.

Pete’s pass decisions

Lots of interesting companies have flooded my deal flow network lately, we got the elite of the elite operators promising big high level concepts on massive TAMs.

Atoms — Atom’s is Travis Kalanick (founder of Uber)’s “new” venture–not that new because it’s three verticals that’s being rolled up, including his eight years old Cloud Kitchens concept. You have food, mining, and transport. It’s a weird mesh. In his interview with Atoms biggest backer a16, Travis says hes visualizes these 3 verticals forming a digital computer, but for in the physical world. “Gainfully employed robots” would cook food, deliver food, mine rocks, and transport stuff (presumably the food, the rocks, and the robots themselves). He’s basically backing specialized robots over generalized humanoids.

I mean… I wouldn’t bet against him but I also can’t 100% buy into the vision, at least not at the offering price of $16 billion. Eight years into the Cloud Kitchens concept and it never really delivered anything special–plus I am sort of biased againt the dystopian idea of factory-created slop bowls being the primary food medium of the world. I want special food cooked by a trained chef. Travis is back and that’s cool and all but I’m passing.

Blue Origin — another big name and big concept with Jeff Bezos’s space startup. $144 billion valuation for the offering. It’s an easy pass for me because all I can do is compare this bet to the one I made in SpaceX two years ago and it’s inferior in every possible way. No Starlink cash cow and far less revenue. History of slow execution and massive cash burn and talent leaving. Not really an innovator in reusability the way SpaceX was with their development of Falcon 9 when everyone said it wasn’t possible. Their rocket blew up just a couple months ago. Wish them 100% success though.

Helion Fusion — I took a detour in fusion for a day and came out believing this might just be the quantum computing of energy–perpertually five years away from being five years away from being ready for commercialization. There is currently no functioning fusion-based power plant at the moment, it’s solely existed as a “concept of the future” since the 1950’s. That said, maybe AI’s compute shortage (and implied power bottleneck) might unlock the capital needed to push this tech into functionality. Every alternative power concept out there is a getting look due to the AI-industrialization of society and it drips down to the startup level. Among 8 different fusion startups, Helion appears to be the valuation leader at their latest $15.5 billion round, which is being led by a tier-1 VC name in Thrive Capital. They claim their 50 MW Orion plant will be up and running in 2028. I am intrigued but I have no idea how to assess this. I don’t want to be the guy undertaking science risk when I’m not a scientist by trade.

Kalshi/Polymarket — these two are quite widely available–I have been getting offers to invest for the last 12 months at several different platforms. I kept passing on them and they just kept going up. I have no regrets because I’m just not excited about their concepts. I think the crypto/gambling/prediction phase of my life is over now and I no longer glorify its cultural impact in our society. The reduced friction of gambling in the past decade is a trend that will rot our society, particularly our youth. Young Pete actually wanted gambling to become more legal but I never thought that would result in today’s lame-ass society where zoomers bet $20 on their pocket computers on pre-season games between the Titans and Panthers and then stream it on Twitch. I now believe gambling should be legal but it should be somewhat inaccessible to the masses, like when you had to find a back alley bookie or travel to Vegas. Seeing Kalshi advertise literally everywhere just puts a bad taste to my mouth. I also think their latest $40 billion round (not yet closed) is at a preposterous valuation–it appears to be based on their World Cup-driven revenue surge. But you cannot base ARR off a once-every-four-years event, that’s so dumb. I missed out here but I’ll continue to miss out.

Harvey/Glean/Open Evidence — Harvey is basically an AI-driven legal assistant for attorneys. Glean is an AI-driven enterprise search company. OpenEvidence is an AI-driven support assistant for healthcare professionals. One key investment theme I haven’t yet made a bet on is the AI application layer. I’m just not sure about the strength of the moat of these companies and how well they can compete with OpenAI and Anthropic or rival wrappers (like Legora). There was probably a price where these names were attractive but they still trade at pretty steep multiples to revenue. For example, Harvey is starting a round at $15 billion post money and their rumored ARR is around $300 million. That is 50x revenue. I’m not overall bearish or skeptical on the application layer but I think there’s been better risk-reward in AI Infra. So these names are like a wait and see for me.

Bunch of stuff I passed on for boring reasons like pricing, fee drag, counterparty, or just not having enough bandwidth to research them and have a strong feeling about it.. but they do seem interesting and I could change my mind with more intel or better deal structure: Baseten, ShieldAI, Suno, Mercor, Higgsfield, Boom Supersonic, Fluidstack, Starcloud, General Compute, CuspAI, Panthalassa.

Some updates for current holdings

Valuation bumps and IPOs:

Databricks just closed their latest Series M at $190 billion. My position is sitting on 3x.

Crusoe is about to close their Series E at $30 billion. My position is sitting on 2.5x.

Anduril is in talks for a new round at $100 billion. That would have my position marked at 2x.

Erebor is in talks for a new round at $8 billion. That would have my position marked at 3x.

Agility Robotics is going public via SPAC at a $2.5 billion valuation ($3.3 post money). My position at the current price the SPAC company, CCXI, is about 1.9x. The deal is expected to close in the fourth quarter of 2026 and the new symbol will be AGLT. If humanoids can catch the same hype bid that quantums/nukes caught in 2025, I think this could go to $50 or more. I’m hoping Unitree’s 8000x oversubscribed IPO over in Hong Kong can become an inflection point for accelerating sentiment in humanoids.

Anthropic is in talks for an October IPO at $2 trillion. That would have my position marked at 7x. This would surpass SPCX as the largest IPO in the history of civilization. Two years ago, did you ever think we’d have multiple trillion dollar IPOs in one year?If someone told me that and to guess what year, I would’ve guessed 2045, adjusted for inflation. Exciting times.

Other updates:

Saronic is already trading at 50% premium to its latest round that closed in April. One of their autonumous vessels did a rescue job for two ejected pilots in the Iran War and I think it’s giving them a lot of goodwill.

Hark, which I had previously told you had no product when I invested in it, has just introduced its first product–Hark Handoff.

Google Deepmind just came out with Gemini Robotics 2. Their demo video has Apptronik‘s Apollo robot on it. GR2 would be considered a credible rival to my other generalized robotics brain startups like Physical Intelligence and Skild.

Canva is reportedly struggling to defend its moat in the AI era. This might be the only holding that I have that would be considered “a bag” at this point. Oops.

Some updates for your host Pete

Right now, I don’t feel like I have to do anything. Like yeah, I could invest more but I don’t have to force it. I could trade again if I feel like it but that time has yet to come so I haven’t forced that either. I can do whatever I want and it’s totally fine and earned. No guilt for choosing pleasure over business. Just feeling the peace and contentness in regards to my career and my wealth.

I am getting more and more AI-pilled by the day. I honestly don’t care about corrections (like the latest dip in AI names, punctuated by the demise of Situational Awareness) that much although it’s easier to say that when my most beta-sensitive holdings aren’t mark to market. I think AI has surpassed expectations from a capabilities and revenues standpoint, compared to what analysts/experts/consumers were expecting 1-2 years ago. I think we’re still early. I’m of the belief that we are in a once-in-a-lifetime technological revolution and the biggest mistake is to not participate. I’d like to think there’s well-trained bubble trader inside of me that will wake up when the multiples are just too big, the charts are just too extended, and the risk is just too great to keep going. I’m just not at that point yet.

On a personal note, I am a little scared of what our future will look like as AI becomes more widely adopted and more powerful. I am now a father of one and expecting to be a father of two next year. How will my kids go through school and learn how to think for themselves when they have this super intelligence cheat code in their pocket? How do I teach them that becoming dependent on this device will make them dumber? Will it matter? Will they listen? I’m currently compartmentalizing these worries a bit as they’re still young/unborn but at some point it becomes a serious downer to think about.

If anyone has any questions about pre-IPO or wants a referral (you have to be an accreditor investor), e-mail me at churning.burning.blog@gmail.com. Happy hunting.