← Startup Fundraising

AI Megarounds Stretch From Code Agents to Memory Chips (June 01, 2026)

June 01, 2026 · 9m 16s · Listen

A $1B code agent, a $135M memory-chip bet out of Seoul, and a $50M seed round that used to be a Series B. AI megarounds are stretching in every direction this week, and at least one of them finally has a number attached. Welcome to Startup Fundraising. We’re winding down a week that bounced from valuation gaps to cap-table geography — Cognition gets a product metric, XCENA gets a stress test, and byFounders gives us the one fund story in a week full of company rounds. Cognition finally gave me a number — Devin writes 89% of its own code, per today’s piece in The AI World. I’ve been asking for something countable since Wednesday. Now I’ve got a new question: 89% of whose codebase? Yeah, that’s where we’re starting. Let’s get into it. Awesome Agents, with Sophie Zhang:

Every AI infrastructure conversation in 2026 eventually circles back to GPUs - whether NVIDIA can make enough of them, who gets allocation, and how long training runs take. XCENA thinks that framing misses the actual choke point. The South Korean startup closed a $135 million Series B at a $570 million valuation on May 29, 2026, led by Seoul-based VCs Atinum and IMM Investment with participation from Corstone Asia.

XCENA just closed a $135M Series B at a $570M valuation, led by Atinum and IMM Investment, both based in Seoul. That’s a very different syndicate geography than anything else we’ve logged this week, and the thesis is taking a direct swing at the GPU-allocation conversation that’s dominated AI infrastructure for two years. The MX1 chip does KV cache and vector search at the memory tier — no CPU or GPU round trips — and they’re saying one MX1-equipped server replaces ten machines. That’s a serious ops claim. But mass production is late 2026, revenue is 2027, and the round closed May 29. So the $570M valuation is sitting entirely on the thesis, not on deployment. Worth flagging: Korean LPs leading a semiconductor-adjacent AI infrastructure round is not nothing. Atinum and IMM are betting the memory bottleneck story has legs while US capital is still arguing about GPU allocation. That’s a real point of view, not just a follow-on. Same week Cognition raises a billion dollars on an agentic coding story — I want to know whether XCENA’s memory architecture is actually inside any of those agent stacks, or whether these two rounds are just talking past each other in adjacent PowerPoints. Here's Barratt Dewey at Tectonic Defense:

This morning, the El Segundo-based hardware and software integration startup announced in an exclusive release to Tectonic that it raised $45M in a Series A led by Bessemer Venture Partners to expand its product family, increase deployments, and take its integration tech into new operational domains.

Credit where it’s due — Barratt Dewey and Tectonic Defense had this one as an exclusive. Picogrid, El Segundo, $45M Series A, Bessemer leading. And here’s the number that actually matters: the company has been running on $12M in seed since 2020, it’s cash flow positive, and the CEO says they’ve returned two dollars in revenue for every dollar raised. That NatSec100 stat is the most damning thing I’ve read all week — top 50 on $12M, nearest competitor raised $140M. That’s not an anomaly. That’s an indictment of how the rest of defense tech has been spending. And Bessemer leading is worth flagging — earlier this week Lightspeed was in a defense-adjacent seed, now Bessemer is writing a Series A check into defense hardware-software integration. Two tier-one names in that sector in the same week is not a coincidence. Meanwhile Cognition is eating every headline at $26 billion. Picogrid: profitable, deployed, $12M in. Cognition: one self-reported metric, $1 billion raise. I know which bet I’d rather underwrite. This one's from The Next Web:

London-based AI lab Inherent emerged from stealth on Wednesday with a $50 million seed round co-led by Index Ventures and Radical Ventures. Nvidia’s venture arm NVentures also participated, alongside Ex/Ante, Metaplanet, Macroscopic Ventures, and Mythos Ventures. It is among Europe’s largest AI stealth-to-launch rounds in 2026.

Inherent comes out of London with fifty million dollars, co-led by Index and Radical, and NVentures is also on the cap table. The Next Web has it. This is labeled a seed round, and I do need to say that out loud: fifty million dollars is a seed round now. NVentures showing up again — we flagged them in OpenRouter, now they’re in a foundational research lab that hasn’t shipped anything publicly. Nvidia is building a posture across scientific AI and routing infrastructure, and it’s not subtle anymore. The product is called Faraday, and the framing is “self-improving AI that decides which scientific questions are worth asking.” Also, the Carta Series B median we cited Friday is just irrelevant to how this round was put together. Index and Radical didn’t look at a staging playbook. They just wrote the check. Ex-DeepMind team, Biden White House policy pedigree, Matt Clifford advising — that’s an incredible credential stack for a lab that still hasn’t told us what Faraday has actually done yet. “Figures out which questions are worth asking” is a mission statement, not a product milestone. The AI World writes:

Cognition AI just closed a $1B Series D at a $26B valuation. Its autonomous coding agent, Devin, now writes 89% of the company's own code and it's already live in production at Citi, Goldman Sachs, Mercedes-Benz, and the US Army. Enterprise usage grew over 10x in 2026, with annualised revenue at $492M. AI-native software development is no longer theoretical.

The AI World closes the loop on Cognition today — and they brought an actual number with them: $492M annualized revenue, enterprise usage up 10x in 2026, Citi, Goldman, Mercedes, the US Army all in production. That’s the denominator we’ve been missing since Wednesday, when the $26B headline landed without a business attached to it. Okay, $492M ARR against a $26B valuation is a 53x multiple — which is either visionary pricing or a dare. And the 89% self-written-code stat: the article says Devin writes 89% of Cognition’s own code. That’s an internal repo metric. That is not the same claim as Devin handling 89% of Mercedes’s production commits. The Mercedes story is the part doing the heaviest lifting here — eight months of engineer time compressed to eight days is a real claim, but it’s still one anecdote attached to a $26B number. I’ve been asking for a countable fact on this valuation since Wednesday. They gave me one — 89% — and now I need to know whether that percentage is measured in lines committed, PRs merged, or just vibes inside Scott Wu’s internal Slack. One self-reported metric on a $26B round isn’t relief. It’s a new attack surface. Here's FemWealth:

Co-led by Sara Rywe, byFounders is a community-powered early-stage venture fund investing in globally ambitious teams connected to the Nordic and Baltic countries. The Copenhagen-based VC firm announced an oversubscribed €130 million+ Fund III.

Closing out the week with the only fund-level story we’ve seen: byFounders Fund III, €130 million plus, oversubscribed, Nordic and Baltic focus, co-led by Sara Rywe out of Copenhagen. FemWealth had it first. Community-powered, early-stage, geographically specific — that’s either a differentiated LP pitch or it’s nothing. €130M is real money for that mandate, but the question is whether “globally ambitious teams connected to Nordic and Baltic countries” is a sourcing edge or just a nice sentence. This is exactly the emerging-manager story we kept circling this week — the concentration question doesn’t get solved by one €130M close, but Rywe getting an oversubscribed Fund III out of a community-powered model at least shows the LP appetite exists outside the US megaround circuit. If Startup Fundraising helps you make sense of the raise, take a moment to subscribe or leave a review wherever you’re listening. Reviews really do help other founders and operators find the show.

We’ve put links to every story from today’s briefing in the show notes, so if one of the deals, funds, or signals caught your attention, you can follow it there and read further.

That’s Startup Fundraising for today. This is a Lantern Podcast.