A billion more for fusion, with not a single investor willing to sign their name — and somehow that's not even the wildest number on the tape today. For anyone joining us midstream, here's the short version: AI money's been piling into whoever can claim a grip on the next compute stack — model platforms, infrastructure, efficiency plays. Recent checks went to Fly.io, Enigma, and Multiverse Computing. That last one stacked a $570 million Series C at a $1.7 billion pre-money — and then Moonshot AI walked in and made all of it look modest. This is Startup Fundraising. Today: fusion with a deadline, an AI valuation that quadrupled in eight months, and a seed round that doesn't look much like a seed. We're staying with this story: AI venture market concentration. Follow the show and you won't miss what comes next. From Aaron Pressman at The Boston Globe:
The largest venture capital deal in Massachusetts in almost three years, the funding will go towards finishing SPARC, Commonwealth’s proof-of-concept fusion machine in Devens, expected to come online sometime next year, as well as supporting the planning for the company’s first commercial device, dubbed ARC, at a site in Virginia.
Finally — a billion with a deadline. SPARC comes online next year in Devens, and Mumgaard says this check gets the machine finished. I can actually point to where the money goes. It's the largest VC deal in Massachusetts in almost three years, and Mumgaard declined to name a single investor or put a valuation on it. So we've got the biggest number in the state, backed by an invisible cap table. But split the billion in two, Hope. Half of it finishes a real proof-of-concept. The other half goes toward planning a commercial plant, ARC, on a site in Virginia that's — what — a rendering right now? SPARC has to actually put out more energy than it eats before anybody breaks ground on that. The finance function tells you more than the press release. They're building the capital structure for a multi-billion-dollar program. They've raised four billion, with more coming — though Mumgaard says not tens of billions. This company is preparing to be underwritten like infrastructure instead of pitched like a startup. In three days on this desk, we've watched hard-infrastructure rounds stack up. This is the first where the money buys a finished machine you can point to. No squinting at a thesis. Here's Mark Rutherford at TechTimes:
Moonshot AI, the Beijing-based maker of the Kimi K3 model, confirmed Tuesday that it has closed a $3.5 billion financing round at a $35 billion post-money valuation — more than triple its $4.3 billion valuation from eight months ago, and far above the $1 billion to $2 billion it originally set out to raise, Bloomberg reported.
Eight months ago this company was worth four-point-three billion. Today it's thirty-five. They went out to raise one to two billion and closed three-point-five. Call that a fundraise? Actually, someone tore up the price sheet mid-auction. Bloomberg's story is really the cap table: state-backed and strategic Chinese money in the mix. That 8x markup in eight months comes down to who's writing the check and why, not pure venture conviction. They're already circling a follow-on at a fifty-billion pre and teeing up a Hong Kong IPO they want done this year. So the private markup is really the on-ramp to the listing. And one number didn't make the benchmark chart: independent testing pegged Kimi K3's hallucination rate at fifty-one percent on factual queries. Fifty-one. You're paying thirty-five billion for a coin flip that talks. Add that to the AI capital-concentration tally we've been keeping all week. It's another giant marker, right on top of the fusion billion we just covered. Yeah, but the fusion people have a machine to finish next year. What does this open-weights bet have? An IPO clock and three Beijing data laws that follow the servers wherever you park them. I know which billion I trust. Maria Deutscher, writing in SiliconANGLE:
Groundcover Inc., a startup with a platform for monitoring cloud environments, today announced that it has raised $100 million in funding. One Peak led the Series C round. It was joined by Morgan Stanley, Zeev Ventures, Angular Ventures, Heavybit and Jibe. CTech reported that the deal values Groundcover at $500 million.
Finally, a company that does a thing. Groundcover raised a hundred-million-dollar Series C led by One Peak, and the pitch is a pricing model that doesn't bill you per gigabyte of telemetry. That's an actual operator gripe they're solving. And credit where it's due — CTech had the five-hundred-million valuation before Groundcover said a word. The company announced the round; CTech announced the number. Five hundred million, though. Call it 5x revenue and you're in a room with Datadog, plus Dynatrace and New Relic. They're all public and entrenched. Now Groundcover needs to show customers actually switching, not just a cleaner eBPF story. Morgan Stanley's on the cap table too. When the bank shows up at a Series C, somebody's already sketching the exit slide. After a billion for fusion and thirty-five for Moonshot, a hundred million for people watching your Linux kernel feels almost quaint. Quaint, and I'll take it. The pitch is simple: companies shouldn't have to go blind on their dev environments just to save money. If that lands, the multiple takes care of itself. Business Insider, with Geoff Weiss:
Two founders who sold their first startup to Cisco for $500 million have raised $34 million in seed funding for their next act. Last year, CEO Nitzan Shapira and CTO Ran Ribenzaft launched their new startup Harmony, which uses AI to field workplace requests.
Thirty-four million dollars. Seed round. For a company that's been alive since last year and does password resets inside Slack. The $500 million Cisco exit is what's pricing this round, Adam. Shapira and Ribenzaft sold their last company, so investors are buying the founders more than the metrics. Right, and that's my problem — a $34 million seed is a Series A in everything but the label. Somebody's already betting Harmony is a platform, while the product handles onboarding and software access across a hundred apps. Useful? Sure. Category-defining? Show me the retention. When a seed is sized like a growth round, pedigree is doing the underwriting. They've already proved they can build. Cover up the founders' names, then ask whether the numbers still justify it. Onboarding and HR tickets are exactly the unsexy plumbing I like. So build the boring, sticky thing — just don't let the exit story write checks the ARR can't cash. Have feedback, a story idea, or a correction? Email us at startupfundraising at lantern podcasts dot com. We’d love to hear from you.
Looking ahead, Commonwealth Fusion Systems expects SPARC, its proof-of-concept fusion machine in Devens, to come online sometime next year. Moonshot AI, meanwhile, is targeting a Hong Kong initial public offering as soon as this year.
You’ll find links to every story in today’s show notes if you want to dig deeper on any of them.
That’s Startup Fundraising for today. This is a Lantern Podcast.