Nora: An inference company nobody outside a data center could name just closed one-point-five billion dollars — at a seventeen-and-a-half billion valuation. [dry] For running other people's models faster.
Dex: [excited] And that's the SMALL number today, Nora. That's the appetizer.
Nora: [laughs] God help us. This is Nora.
Dex: This is Dex, and it's Sunday, July nineteenth — three stories, all soaked in AI money.
Nora: Let's get into it before the valuations go up again mid-sentence.
Dex: First up — Fireworks AI grabs a billion and a half in a Series D at seventeen-point-five billion.
Nora: Then DeepSeek — yes, again — chasing another one-point-five billion at seventy-one billion, and now flirting with an A-share IPO.
Dex: And to close us out, an Indian coding startup called Emergent hit unicorn status a year after launch.
Nora: A year. I couldn't get a lease signed in a year.
Dex: So — Fireworks AI. Enterprise inference platform, open-source and specialized models, and they just closed one of the biggest infra rounds of the year.
Nora: Okay, define "inference" for the humans, Dex. To me it's the part after the model already learned everything.
Dex: Exactly that — it's the serving layer. You've got a model, Fireworks makes it run fast and cheap at scale for enterprises.
Nora: So they're the toll booth on the AI highway.
Dex: [laughs] Basically! And toll booths at seventeen-point-five billion print money.
Nora: Do they, though? Inference is brutally competitive. Everybody and their GPU is racing to serve models cheaper.
Dex: That's the bet — they're leaning hard into open-source and specialized models, not just being a generic wrapper on the big labs.
Nora: [dry] "We're differentiated" is what every commodity business says right before margins hit zero.
Dex: Hot take, though — inference is where the real AI money lives. Training is a bonfire. Serving is a subscription.
Nora: That I'll grant you. Recurring beats one-time. But seventeen billion prices in a lot of subscriptions that don't exist yet.
Dex: Sure, but every company deploying open models needs someone to run them without hiring a whole infra team.
Nora: So what's this mean for the person listening who isn't a VC?
Dex: If you're building on AI, this is good news. Cheaper, faster inference means your app's API bill drops.
Nora: And if you work at a rival inference shop? Update the resume. Consolidation's coming when the music stops.
Dex: Next — DeepSeek. China's darling is in talks for another one-point-five billion at a seventy-one billion valuation.
Nora: Wait — didn't they just raise seven billion? Like, two months ago?
Dex: Two months ago. [beat] And now they want more, and they're eyeing an A-share IPO this year.
Nora: Seven billion doesn't last eight weeks? What are they doing, laminating the GPUs in gold?
Dex: AI infrastructure at national scale, Nora. Chips, data centers, the whole buildout — it eats cash like nothing you've seen.
Nora: The A-share IPO is the part that raises my eyebrow. That's the domestic Chinese market — retail investors, tight state oversight.
Dex: Which is exactly the point — going public at home, on home rules, funded by home money. It's a statement.
Nora: [dry] A statement, or a way to raise capital when Western money is off the table.
Dex: Can be both! Seventy-one billion for a company that shook the whole industry with a cheap model — that's not nothing.
Nora: It's the velocity that scares me. Seven billion, then another one-and-a-half, then IPO — all in one year. That's not growth, that's a sprint away from something.
Dex: Or a sprint toward something — dominance. They're trying to be the AI backbone of an entire economy.
Nora: So what's the takeaway for someone watching from outside China?
Dex: The AI race isn't just two American labs anymore. If you're planning a global product, DeepSeek's stack is a real option now.
Nora: And if you're an investor eyeing that A-share IPO — read the fine print twice. Domestic listings don't play by the rules you know.
Dex: Last one — Emergent. Indian AI coding startup, one-thirty million Series C, one-point-five billion valuation, unicorn in just over a year.
Nora: A vibe-coding company. Say it with me — vibe-coding.
Dex: [laughs] Say what you want, the category is on fire — Cursor, Lovable, and now Emergent right there with them.
Nora: I lived the founder life, Dex. A year to a billion-five means the product's barely aged past a prototype.
Dex: Or the product's that good. You describe what you want, it builds the app. Non-engineers shipping software.
Nora: Which is amazing until they need to debug the thing the AI wrote and nobody understands it.
Dex: Sure, but they're winning out of India — that's the story. This isn't a Silicon Valley round.
Nora: That part I love, honestly. Global talent, global capital. Good.
Dex: Hot take — vibe-coding is the biggest onramp to software since the spreadsheet.
Nora: [sarcastic] And the biggest offramp to a security breach since the spreadsheet.
Dex: [laughs] Fair. But the wave's real — three unicorns in one category tells you it's not hype.
Nora: Or it tells you VCs are pattern-matching each other into the same trade.
Dex: So what's it mean for you? If you've got an idea but can't code — try one of these tools this week. Seriously.
Nora: And if you do code for a living — learn to review AI-generated code fast. That's the job now.
Dex: So — Fireworks lit up seventeen-point-five billion running everyone's models at speed.
Nora: DeepSeek's back for another one-and-a-half after eight weeks, eyeing a hometown IPO.
Dex: And Emergent went unicorn in a year, riding the vibe-coding wave.
Nora: Three rounds, one word: AI. Shocking, I know.
Dex: Before we go — I added it up. Just today's three stories, that's over three billion in fresh capital.
Nora: On a Sunday. Nobody's even at the office.
Dex: [laughs] The term sheets sign themselves now, Nora. It's the vibe-funding wave.
Nora: [dry] Don't. Don't make that a category.
Dex: That's the show! Three billion before lunch — see you tomorrow, folks.
Nora: And if DeepSeek raises again before tomorrow — Dex, you're doing an emergency episode. I'll be the toll booth.
This show is made with AI: the hosts’ voices are synthetic and the scripts are AI-assisted. Every story links to its original source.