Anthropic is closing in on a thirty-billion-dollar raise that would push its valuation past nine hundred billion. And somehow that’s not even the weirdest part of today’s board. This is Startup Fundraising. Today we’ve got Anthropic, DeepSeek’s first outside raise, and a self-improvement AI company with Nvidia on the cap table — which, somehow, is already the fourth time this week Nvidia has written a check into a company whose revenue is basically still a promise. And one of those founders is on the record saying he’ll put frontier research ahead of revenue. Apparently you can just say that out loud when you’re raising ten billion dollars. Sequoia, Dragoneer, Altimeter, Greenoaks — they’re all in the Anthropic syndicate. At nine hundred billion, the bigger question is who actually set that price, because I’d really like the answer to be cleaner than this. This one's from NewsX.io:
Anthropic is set to close its latest round of funding, which may top US$30 billion at a valuation above US$900 billion, as soon as next week, according to people familiar with the matter, vaulting ahead of rival OpenAI to become the world’s most valuable AI startup.
The mega-round ceiling we were watching after Hark just got blown past. Anthropic is now tracking north of thirty billion dollars in a single raise, at a valuation above nine hundred billion, and it could close as soon as next week. Sequoia, Dragoneer, Altimeter, and Greenoaks are each putting in about two billion, so you’ve got four co-leads at the same check size — which makes the usual ‘who priced this?’ question a lot harder to pin down. Four co-leads at two billion each gets you eight billion of the thirty-plus. The rest is Founders Fund, General Catalyst, and a bunch of existing investors the article hasn’t named yet. What I want to know is the revenue run rate against a nine-hundred-billion-dollar valuation, because Claude is a real product, but ‘strong investor demand’ does not show up on a P&L. The other detail that matters: this came together in a matter of weeks after Anthropic got inbound proposals in late April. That’s not a company running a process — that’s a company fielding offers. Whether that speed says conviction or FOMO on the investor side is the real read here. Rounds that close in weeks off inbound demand at nine-hundred-billion-dollar marks are exactly where the terms get soft — liquidation preferences, governance rights, anti-dilution. The headline number is real. What’s underneath it is still the part we don’t have. From The Next Web:
DeepSeek founder Liang Wenfeng has told prospective investors in the company’s first outside funding round that the lab intends to pursue artificial general intelligence as a primary goal, and will keep releasing open-source models rather than chase short-term commercialisation.
DeepSeek’s first external raise is seventy billion yuan — roughly ten billion in valuation, with at least a three-hundred-million-dollar check size — and Liang Wenfeng is walking into that room telling investors the lab will prioritize frontier research over revenue. That’s not a side comment. When a founder says that to prospective investors, the term sheet becomes a governance conversation fast. High-Flyer Quant funded this thing entirely until now. Liang had kept outside money out specifically to avoid pressure on the product roadmap. So if he’s opening the cap table now, it’s because the training runs got too expensive even for a profitable hedge fund. That’s the headline — the cost curve finally caught up with him. The Information says the longer arc is seven billion or more as the lab moves toward recurring revenue, which means the ‘no near-term revenue focus’ line has a shelf life. What I’m watching is the gap between the three-hundred-million close and the point where recurring revenue is supposed to arrive — and what the liquidation prefs and information rights look like in that gap. Liang is basically asking investors to price the round on mission, not metrics. That’s either the most honest pitch of the year — because he’s telling you exactly what you’re buying — or it’s the most dangerous one, because there’s no invoice stack to audit. I’ve spent the whole week asking for the invoice stack, and the market keeps sending back vision decks. DeepSeek is just the version where the founder says the quiet part first. Here's Phemex News:
AI startup Recursive has emerged from stealth mode, securing $650 million in funding at a $4.65 billion valuation. The funding round was led by GV and Greycroft, with contributions from AMD Ventures and Nvidia. Recursive aims to develop recursive self-improvement technology, which it believes is the fastest route to achieving superintelligence.
Recursive is out of stealth: six-fifty million raised at a four-point-six-five billion valuation, with GV and Greycroft co-leading. The strategic-LP thread from all week just got its sharpest version yet — AMD Ventures and Nvidia both on the cap table of a company whose whole thesis is AI that improves AI. Nvidia is now on four cap tables this week where the revenue line is still a future-tense sentence. At some point that stops looking like conviction in the company and starts looking like a picks-and-shovels hedge dressed up as a Series A. The Nvidia conflict is worth saying out loud: they’re selling the chips that train these models, and they’re also on the cap table of a company whose product is meant to make those models need more chips. That is not a passive LP position. Richard Socher and Tim Rottäschel have real credentials — OpenAI, Meta, Uber AI — but ‘recursive self-improvement as the fastest route to superintelligence’ is a roadmap with no first stop. What does Recursive actually ship in the next eighteen months that somebody can invoice for? Here's Cristian Dina at The Next Web:
Moment, the fintech company founded by a cohort of former Citadel Securities quantitative traders and researchers, has raised $78 million. The round was led by Index Ventures with participation from existing investors Andreessen Horowitz and Avra. The company last raised $36 million in July 2025.
Moment — former Citadel quants, $78 million raised, led by Index Ventures with a16z and Avra following in. The client list is the part I’d actually hold onto: Edward Jones at $2.1 trillion in AUM, LPL at $1.7 trillion. Those aren’t pilots. Those are institutional commitments, which means Index wasn’t pricing this off a deck. Edward Jones and LPL as named partners is the detail that matters — that’s $3.8 trillion in AUM on the hook. But ‘AI operating system for wealth management’ is still a category claim, not a revenue line. What’s Moment’s take rate on fixed-income and equities execution through those accounts, and did that move when they doubled the round from $36 million to $78 million? The jump from the July raise is worth flagging: $36 million to $78 million in under a year, with the same investor base plus Index stepping in as lead. That means Index is the one setting the price this time, not a16z, even though a16z was in the prior round. That’s the valuation move I’d keep tracking. This one's from Fund Momentum:
Beijing-based BAI Capital, founded by Anna Long and spun off from Bertelsmann Asia Investments in 2021, has held a US$600 million first close on a new US$800 million fund — the firm's largest USD vehicle ever and one of the largest dollar-denominated China-focused growth funds raised by a non-mega-fund manager in the current cycle.
BAI Capital is Beijing-based and spun out of Bertelsmann Asia Investments in 2021. They’ve hit a $600 million first close on an $800 million target, and Fund Momentum says it’s one of the largest dollar-denominated China-focused growth funds from a non-mega-fund manager in this cycle. The LP list still isn’t named. Same opacity pattern we flagged on Armada’s $230 million ghost cap table on Monday — when the thesis is cross-regional and the LPs are unlisted, you’re taking the fundraise on faith. The pitch is ‘Chinese champions expanding overseas’ — which is a pitch, not a portfolio. Twenty-two IPOs in the track record is a real number, I’ll give them that. But $600 million in a market where LPs have been pulling China exposure since 2022 means somebody opened a big checkbook, and I’d really like to know who before I start grading the conviction. And that lands right in the Galleon Forge thread Adam brought up Wednesday — who’s actually contracted versus who’s still in the pipeline on cross-border infrastructure. BAI’s ‘international companies leveraging the Chinese market for scale’ bucket is exactly where that question lives, and we still don’t have named portfolio companies or LP commitments to check it against. Got a fundraising question, a founder story we should follow, or a correction to send our way? Email us anytime at startupfundraising at lantern podcasts dot com. We’d love to hear what you’re watching.
You’ll find links to every story we covered today in the show notes, so if something caught your ear, you can dig into the source and spend a little more time with it.
Thanks for listening. That’s Startup Fundraising for today. This is a Lantern Podcast.