The frame running through every item today:
Accelerating the ‘musical chairs’ game around AI startup investments.
This is a familiar game of musical chairs that’s present in almost every tech wave. I was a partner at Goldman many years ago and led the internet research effort there through the nineties — we saw it in the internet wave, the PC wave before that, social media, cloud, mobile.
The difference in this AI tech wave is that the numbers are larger, the pace of follow-on rounds is quicker, and the participants are a circular mix of financial, corporate, and cross-over investors at a scale not seen in prior waves.
Three mega AI tech IPOs are lined up for end of year: Elon’s SpaceXAI (just renamed) at $1.75T+ in June, OpenAI and Anthropic each shooting to be first by year-end at $1T+ valuations. Even seed and Series A rounds are now raising hundreds of millions on companies a few months old.
Today’s three Takes underline what’s actually different about this tech wave’s musical-chairs game — plus a Gadget AI on gaming hardware getting hit by the AI memory crunch.
Three Key Takes today:
(1) AI Startups Follow Tried-and-True Formulas — Just Faster, Bigger, Circular-er. The “fast-follow” round economics have a familiar shape — but the cycle time is compressed and the participant mix is unprecedented. Established companies are doing rapid-fire raising (raise $150M this month, then announce next $150-200M next month at 20-30-40% higher valuations). Sophisticated investors are doing it with enthusiasm. Mix of financial + corporate + strategic + crossover investors. Nvidia keeps showing up across cap tables of companies it’s also a vendor to — call it the “circular” or “boomerang” deal pattern. Crossover funds (institutions playing both private and public) are buying at private prices that increasingly look like late-stage public-market multiples. The reporting: The Information — The Problem with Fast Follows. Standing thesis on Nvidia’s Kingmaker role: AI-RTZ #1036 — Nvidia Accelerating Role as the Kingmaker.
MP Take: “This is a familiar game of musical chairs that is present in almost every tech wave. Difference in this AI tech wave is that the numbers are larger, the pace of follow-on rounds is quicker, and the participants are a circular mix of financial, corporate, and cross-over investors at a scale not seen in prior waves. Everyone understands the game being played and the music’s playing on longer than normally expected. And everyone generally knows that it can stop at any time.
The secular, bottoms up technology opportunities to be addressed by the startups are real enough. The pace of valuation jumps and the scale of those jumps, not so much. Over 30 plus years across tech waves I’ve watched, investors typically tended to be in one bucket or the other — they would wait for companies to evolve from one capital-raising segment to the next. No more. There’s a ‘everyone in on the action on a cross-investing basis’ that’s a key trend driving the momentum here.”
(2) Ex-OpenAI Researcher’s Six-Week-Old Startup at $4 Billion Valuation. A six-week-old startup founded by an ex-OpenAI researcher is in talks for a $4 billion valuation. The current hot name is Core Automation in the Valley — attacking AI agents on skills and automation. Talent dispersal from OpenAI, Anthropic, and Google DeepMind is now the dominant private-market deal flow in AI. Classic example: Character.ai — ex-Google researcher (one of the prominent co-authors of the famous “Attention Is All You Need” 2017 AI paper), left Google with co-founders, built a multi-billion-dollar private company, then got bought back into Google in an acquihire for several billion dollars. He’s back at Google driving a lot of the Gemini changes. The reporting: The Information — Ex-OpenAI Researcher’s Six-Week-Old Startup Targets Funding at $4 Billion Valuation. Standing thesis on AI talent dispersal: AI-RTZ #436 — Another Possible Key AI Talent.
MP Take: “This area has bottom-up opportunity given the range of areas that need bottom-up research innovations around AI. At every layer of the tech stack. Four years post-ChatGPT, the work has barely begun. We’ve seen this in other tech waves, but the runway of opportunities to innovate is longer than almost any other tech wave in the past four decades.
Flipside is that high valuations up front aren’t doing the founders and investors any favors long term. Especially since the larger valuations today point to big tech acquihires and M&A as the most near-term exit opportunities. The acquihire is the new innovation this tech wave vs prior tech waves. And it’s a clever twist on traditional M&A that creates more probabilistic outcomes than ever before.
Employees at companies like OpenAI, Anthropic, Google have very quickly recognized there’s a lot of VCs chomping at the bit to give them money to leave, start a startup, and see if it works. If it doesn’t work, they typically come back to their companies or get another job. More often than not, they get acquihired back. Character.ai is the canonical example.”
(3) Defense AI — A Hot Adjacent Area in the Physical World. Anti-drone AI is the fresh signal: a startup in talks for a $2 billion valuation. Silicon Valley typically hadn’t invested in defense for a long time, but AI is dramatically changing the defense segment — just as it is finance, healthcare, and so many other domains. Defense is at a pivotal point akin to when airplanes first arrived in the early 1900s and battleships gave way to aircraft carriers as the dominant Naval form. Same thing now with drones — in the air, on water, on the battleground in the form of robotic soldiers. Lots of politics, ethical questions, sci-fi headwinds (Skynet, Terminator, War Games). The reporting: The Information — Anti-Drone AI Startup in Talks for $2 Billion Valuation. Standing thesis on AI on land, sea, and air: AI-RTZ #348 — AI on Land, Sea, and Air. Standing thesis on defense AI politics: AI-RTZ #344 — Shall We Play a Game?.
MP Take: “Defense globally is seeing a secular rethink akin to the dawn of the airplane. Fundamental precepts have to be rethought. And the space has to contend with the sci-fi headwinds of Skynet from Terminator and movies like War Games that have, rightfully, made an impression on people that AI not thought through could be bad for a lot of us. The technology-driven changes and opportunities will win out, but it’s a long political road with lots of headwinds.
Unlike defense waves of the last 20 or 30 years, where there were typically two or three clusters — US, Europe, against the Soviet Union — now it’s not just the big superpowers in the game. A lot of smaller countries can do a lot more with a lot less. Iran is a great example — their drone technology, even before the current war, was a huge input for the Russian side against Ukraine. Shahid drones were a key reason Russia and Putin were able to counter Ukraine.
And Ukraine now, because of their sophistication that has literally been developed at the expense of a lot of lives and better experience on the field, has strengthened Ukraine’s case for potential entry into the EU at some point and down the road NATO because of this experience with AI technology — specifically drones in all manners of speaking. Turkey is another country with a notable ramp in AI drones.”
Plus: Gadget AI — Gaming Hardware Officially Hit by Memory Price Hikes. Continuing a theme MP has been highlighting — Microsoft Surface laptops up 15-20%, Samsung up 15-20%. This week, two gaming hardware companies independently raised pricing. Sony’s Q4 release was generally promising but PlayStation sales are down because they had to increase pricing — and they talked about additional price increases ahead due to memory shortages. Nintendo’s Switch 2 forecast disappointed for the same reason. Gaming has historically been a leading indicator of broader consumer-tech cost dynamics — Nvidia’s roots are in gaming GPUs that later spilled into AI compute. The reporting: CNBC — Sony Q4 Earnings · PS5 Gaming Sales · Memory Price. Nintendo Switch 2 angle: Bloomberg — Nintendo Profit Forecast for Second Year of Switch 2 Disappoints. Standing thesis on memory shortages: AI-RTZ #1039 — Global Memory Shortages of 20….
MP Take: “Gaming is but one subset affected by AI-driven memory and chip shortages and price increases. But it’s an important one to watch due to the high incidence of core innovations here in hardware and software, that later find applications in so many larger tech opportunities. Exhibit 1 is Nvidia itself, of course. Any dampening of gaming industry growth has unintended consequences of impact on the broader tech industry not fully anticipated.
On a broader note, these shortages also slow down the development of local AI vs ‘mainframe’ AI — a thread I’ve discussed at length. The local-AI-on-device ramp depends on cheap, abundant memory; the mainframe-AI-in-the-cloud ramp tolerates expensive memory in centralized scale. Right now, the cost dynamics tilt the balance toward mainframe AI for longer than the local-AI-on-device thesis would prefer. Yesterday’s tease about OpenAI’s potential AI smartphone — that hardware is going to cost a lot more given these component dynamics. Headwinds, not all visible from the headline capex numbers.”
Today’s AI-RTZ #1080 covers Anthropic’s AI compute constraints throttling growth — and the SpaceX/xAI ‘Colossus’ deal of sudden convenience. A deeper read on the multi-billion-dollar AI compute deal Anthropic announced this week with Elon’s SpaceX/xAI Colossus AI Data Center. Today’s transcript adds a fresh data point: Anthropic is now at a $45B ARR run rate — up from ~$30B “two or three weeks ago” — and is finishing a round of $50B+ at a $900B pre-money valuation. That’s the trillion-dollar mega-IPO bracket arriving live, with similar OpenAI news likely soon.
Closing Questions —
-
What’s the one gaming hardware innovation MP is most looking forward to in 2026? Valve’s Steam Machine. A Linux-based gaming console with potential to make a dent in Microsoft’s Windows gaming dominance — both on the hardware side (vs Xbox) and the software side (vs Windows-PC gaming). Originally targeted earlier this year; memory shortages have pushed pricing from $400-500 up to $700-800+, and timing’s now uncertain. But the broader implications for the next few years on console-style gaming and handheld gaming are substantial. Anything that meaningfully chips at the Windows-PC-gaming tax is interesting structurally — both as a gadget freak and as part of the OS-clash-with-AI-Agents thesis from yesterday’s Ep 71. Source: The Verge — Valve Steam Machine TV Gaming PC Wanted.
-
What’s the one software game innovation MP is most looking forward to in 2026? GTA VI, of course. Long-time GTA 5 fan; the sequel is one of the biggest media properties in the history of humanity — bigger than most movies or any other media form. Long-anticipated, multiple delays, still hoping it ships this year. Beyond the cultural moment, game-engine innovations from Rockstar typically spill into broader interactive-media and AI-adjacent fields — procedural worlds, NPC behavior, real-time content scaling. The 2026-economy boost from a single AAA title shipping at this scale is also non-trivial — this is a year-shaping cultural release. Source: Yahoo Finance — GTA 6 Could Help 2026.
Thanks again for joining us today. Have a great weekend. Stay tuned.
(NOTE: The discussions here are for information purposes only, and not meant as investment advice at any time. Thanks for joining us here)
Short — The Acquihire Is the New Tech-Wave Innovation Ex-OpenAI researcher’s six-week-old startup is in talks for a $4 billion valuation. Talent dispersal from OpenAI, Anthropic, Google DeepMind is now the dominant private-market deal flow. Classic example: Character.ai — ex-Google researcher (one of the “Attention Is All You Need” co-authors) left, built a multi-billion-dollar company, got acquihired back into Google for billions, now drives Gemini.
MP Take: High valuations up front aren’t doing the founders and investors any favors long term. The larger valuations today point to big tech acquihires and M&A as the most near-term exit opportunities. The acquihire is the new innovation this tech wave vs prior tech waves. A clever twist on traditional M&A that creates more probabilistic outcomes than ever before.
Short — Defense AI Is Going Global — Not Just the Superpowers Anti-drone AI startup in talks for a $2B valuation. Defense AI is the hot adjacent in physical-world AI — and unlike past defense tech waves, this one’s global. Ukraine’s AI-drone learnings are recasting NATO doctrine. Iran’s Shahid drones (supplied to Russia) are a major input on the war’s other side. Turkey is running fast on AI drones.
MP Take: Defense globally is seeing a secular rethink akin to the dawn of the airplane. Fundamental precepts have to be rethought. Tech-driven changes will win out, but it’s a long political road with sci-fi headwinds — Skynet and Terminator have rightfully made impressions. Unlike historical defense waves, this one isn’t restricted to superpowers. Smaller countries doing more with less. Ukraine’s EU-entry case is bolstered by its AI-weapons experience.
Short — Gaming Hardware Is the AI Memory Crunch Bellwether Sony Q4 earnings flagged it. Nintendo Switch 2 forecast disappointed for the same reason. AI-driven memory and chip shortages are now translating into gaming hardware price hikes. Microsoft Surface laptops up 15-20%. Samsung up 15-20%. Now PS5 and Switch 2 take the hit. Gaming has historically been a leading indicator of broader consumer-tech cost dynamics — Nvidia’s roots are in gaming GPUs.
MP Take: Gaming is but one subset affected by AI memory and chip shortages. But it’s important due to high incidence of core innovations in gaming hardware and software that later find applications in larger tech opportunities. Exhibit 1 is Nvidia itself. These shortages also slow the development of local AI vs mainframe AI — local-AI-on-device ramp depends on cheap memory.
Short — AI’s Musical Chairs Game Plays On — For Now Three mega AI tech IPOs lined up for end of year — SpaceX/xAI ($1.75T+, June), OpenAI ($1T+ year-end), Anthropic ($1T+ year-end). Venture money has mushroomed across the AI startup universe. Even seed and Series A rounds raising hundreds of millions on companies a few months old. There’s a momentum element to this — and that’s the musical chairs.
MP Take: Right now we have a happy circumstance where lots of secular reasons drive enthusiasm — fourth year of ChatGPT, lots of innovation still to do. Most startups are interesting. Just the valuations are high and exit scenarios are increasingly driven by strategic acquisitions and the acquihire phenomenon. As an investor, direct or through funds, things to keep in mind. Music plays on — for now.
Ep 72 scope: 1 Main + 4 Shorts — MP’s pre-recording scope decision. No Segments or Hooks (matches Ep 62-71 4-clip pattern — 11th consecutive episode).
Both are daily. Both are free. Both are about AI. But they’re different mediums carrying different messages.
AI-RTZ is the morning text — a deeper written take on one idea, published by at least 5 AM EST. Today: post #1080 — Anthropic’s AI compute constraints throttling growth, and the SpaceX/xAI ‘Colossus’ deal of sudden convenience.
AI Ramblings Daily is the afternoon video + podcast — my ad hoc takes and perspective on the day’s AI issues & news flow, around 16 minutes today, with short 1-minute clips for quick topic views. Today: episode #72.
Subscribe to either or both on michaelparekh.substack.com. They run as separate Sections you can opt into or out of.
Take 1 — AI Startups Follow Tried-and-True Formulas (Fast-Follows):
Take 2 — Ex-OpenAI Researcher’s Six-Week-Old Startup at $4B:
Take 3 — Defense AI Hot Adjacent (Anti-Drone AI Startup at $2B):
Gadget AI — Gaming Hardware Hit by Memory Price Hikes:
Q1 + Q2 — Gaming what-MP-is-most-looking-forward-to:
Companion text: