A crypto-native firm just put $1.2 billion behind AI and robotics — and forget who led for a second. Did the LPs who signed up for token economics know they were funding a robotics mandate? This is Startup Fundraising. It's Thursday. Today: Paradigm crosses lanes, a fusion reactor with Google money behind it, and a defense startup selling paperwork instead of hardware. Tap follow so the next episode finds you. Helene Braun, writing in CoinDesk:
Paradigm has raised $1.2 billion for a new venture fund focused on artificial intelligence and robotics, according to a blog post Wednesday. The firm's third venture fund reflects growing interest in AI and defense technology, though it does not signal an exit from digital assets.
Paradigm — the crypto shop — just closed $1.2 billion for AI and robotics. It's the firm's third venture fund, per CoinDesk, with crypto-native LPs re-upping into a mandate that looks nothing like the 2021 and 2024 crypto funds they signed on for. Right, and that's the whole thing for me. What does a firm built on token economics actually know about how to diligence a robotics go-to-market? They've already put money into Zipline and True Anomaly — drone delivery and space defense. That's a long way from a mempool. Palmedo's line to Bloomberg was basically "crypto's still exciting, but there's so much else happening we can't ignore it." Pretty honest way to say the returns are somewhere else right now. The size matters, too. A billion-two lands the same week you've got SambaNova pulling a billion at eleven billion — generalist, specialist, everybody's sprinting into the same infrastructure layer. So Paradigm's edge is what, exactly? Being early to crypto in 2018 doesn't price a robotics deal in 2026. All week, capital's been jumping category lines — and this is the cleanest single-firm version of it: a crypto GP widening the mandate in public. At this point, it's a strategy announcement. This one's from VentureBeat:
Arkenstone Defense emerged from stealth today with $35 million in seed funding to eliminate the compliance, security, workforce, and operational burden that keeps many commercial technology companies from selling to the U.S. government. The round was led by J2 Ventures, with participation from Susa Ventures, Granite Hill Capital Partners, and Artis Ventures.
Arkenstone, $35 million seed, and the whole pitch is a GovCon back office in a box — compliance, payroll, personnel security, accreditation. That's a services business wearing a platform costume. It's the first one this week selling the paperwork stack instead of the capability stack. J2 led, with Susa, Granite Hill, and Artis in the round. Everyone else has been funding the weapon — these guys are funding the filing cabinet behind it. And I actually kind of like it? Peter Dixon's line is right — startups spend years building a federal back office before they win a nickel. Still, show me the recurring customer. Is this SaaS margin, or is this a consulting shop with a login screen? The largest single check reported in this back-office category, per VentureBeat. Which means nobody's built the on-ramp yet — or nobody's proven you can. Here's the tension, though. The whole reason the slow buyer is a problem is that the buyer is slow. Arkenstone can shortcut the setup, sure — but it doesn't make the Pentagon sign faster. You've moved the bottleneck, not killed it. Leah Hodgson, writing in PitchBook:
Energy startup Proxima Fusion has raised €411 million from an investor group that includes Google and German energy giant RWE, to build Europe’s first commercial fusion reactor. The deal, which values Proxima at €2.4 billion (about $2.7 billion), is the largest in a series of VC energy deals in Europe this year, driven by the continent’s push for energy sovereignty.
€411 million into a company that raised €130 million last year — and Google's on the cap table. The stellarator schedule is secondary. Google needs power for data centers, full stop. Right — RWE and Google are the strategics, XTX and East X are leading, and Balderton and Cherry are re-upping from the Series A. The energy giant and the hyperscaler matter more here than the venture names. And that's the part that worries me. When your strategic buyer is also the company that needs the offtake, Google's interest becomes a ceiling. Who else gets to buy this company once Google's fingerprints are all over the reactor? It rhymes with the Paradigm fund earlier — capital crossing category lines. Google is showing up less as a fusion investor than as a demand signal with a checkbook. This is the biggest European energy VC round of the year, per PitchBook, and AI's power hunger is underwriting it. $2.7 billion for a Max Planck spinout founded in 2023. Three years old. For this to make sense in two years, they need a working stellarator plant with RWE; a rendering and a sovereignty press release won't do it. From TechCrunch:
Chemistry Ventures, the VC firm launched two years ago by Bessemer, Index Ventures, and Andreessen Horowitz alums, is raising $500 million for its second fund, according to an SEC filing. Founded by Mark Goldberg, Ethan Kurzweil, and Kristina Shen, Chemistry launched with a $350 million fund and invests in early-stage startups building in AI, including infrastructure and apps.
Chemistry's going back to market for $500 million on Fund II — up from a $350 million debut, two years in. And notice the sourcing: TechCrunch found this in an SEC filing. No splashy announcement, no press release. Bessemer, Index, a16z alums. So the pitch to LPs is basically the pedigree — three people who used to work somewhere famous. What does that get you that just buying the a16z index doesn't? The Journal says it's already oversubscribed and closing soon. That's a 43 percent step-up before most two-year-old funds have even finished deploying Fund I. Right, that's the part that bugs me. You raised $350 million, you've got Granola and Decagon in the book, and you're back with your hand out before we know if any of it works. Oversubscribed just means LPs are chasing the logos, not the returns — because there aren't returns yet. They invest in AI infrastructure and apps — which, in 2026, is roughly the entire venture map. The portfolio is Granola, Decagon, Persona, Serval, Nova. Early enough that none of that has marked up in a way an LP can bank. If fundraising is your world, you might also like The Data Center Daily, a daily briefing on AI compute, hyperscaler capex, power, chips, and energy markets. Find it wherever you listen to podcasts.
You’ll find links to every story we covered today in the show notes. If something stuck with you, they’re there for a closer read.
That’s Startup Fundraising for today. This is a Lantern Podcast.