Startup Fundraising · 2026-06-25 · 6 min
Key moments - from our scoring
Substance score
50 / 100
Five dimensions, 20 points each
Three infrastructure and deep-tech rounds close this week, each exposing different pressures shaping the 2026 fundraising landscape. Baseten, an AI infrastructure startup, closed $1.5bn at $13bn valuation in its fourth fundraise in eighteen months - raising questions about whether mega-rounds reflect genuine product demand or window-grabbing behavior when led by Sands Capital and Wellington Management rather than traditional venture powerhouses. Engram emerged from stealth with $98M for a learned memory layer addressing wasted AI tokens, backed by General Catalyst, Kleiner Perkins, and Sequoia (with angels including Andrej Karpathy and Assaf Rappaport), but faces margin pressure from Microsoft's Azure pricing and the risk that the category itself remains undefined. Against this backdrop, Partly - a New Zealand auto parts software company using AI for collision repair cataloging - raised $50M at $500M valuation and is opening its first US office, representing the only non-US-founder bet on the tape and the most defensible moat: four years building proprietary vehicle parts data covering 91% of vehicles with 50+ manufacturer agreements. The episode examines concentration risk, unit economics sustainability, and what separates real operational moats from AI-label arbitrage.
Baseten closed a $1.5 billion Series F round at a $13 billion valuation, led by Sands Capital and Wellington Management, with participation from Blackbird (Australia). This was the company's fourth fundraise in eighteen months.
Engram built a learned memory layer to fix wasted tokens in AI systems and emerged from stealth with $98M backed by General Catalyst, Kleiner Perkins, Sequoia, and angels including Andrej Karpathy (OpenAI co-founder) and Assaf Rappaport (Wiz CEO). Its initial partners include Microsoft, Notion, and Harvey.
Partly spent four years and $10 million building proprietary vehicle parts data covering 91% vehicle coverage with 50+ manufacturer agreements, using government records and manufacturer feeds annotated by hand. The company raised $50M at $500M valuation and is opening its first US office to serve the collision repair sector.
Multiple rapid rounds create severe dilution for early investors and raise questions about whether the $1.5B raise reflects genuine product-driven capital needs or opportunistic capital-raising while investor appetite for AI infrastructure remains high.
If Engram's memory layer runs on Azure, Microsoft's own pricing for similar functionality could compress margins, similar to the pricing pressure facing Baseten as larger cloud providers build competitive offerings into their platforms.
Our reviewer’s read on each dimension, with quotes from the episode.
For a 6-minute commentary show the hosts surface several real, non-obvious observations - serial dilution at Series F, the committee-vs-lead read on Engram's syndicate, and the distribution-margins-not-AI-label test for Partly. But the density is uneven; the Baseten segment in particular leans on questions rather than answers.
Every round leaves a dilution trail so by Series F the early checks have been ground down to something you'd need a microscope to find on the cap table.
Hard to call that a lead investor; it reads like a committee.
A few genuinely sharp framings elevate the commentary above standard deal-recap fare - notably the 'round IS the product' hook and the 'beautifully decorated category that doesn't exist yet' cut on Engram. The geographic concentration thesis is an underreported angle. Most takes, however, are analytical rather than contrarian.
At some point, the round IS the product.
It's a beautifully decorated category that doesn't exist yet.
There are no guests at all - this is a two-voice news commentary format. The hosts demonstrate some market literacy but neither is established on the transcript as a practitioner who has raised or deployed capital at scale; they are analysts of others' fundraises, not operators sharing first-hand experience.
Three deals on the tape today - a mega-infra raise a stealth memory play coming out swinging and a New Zealand auto parts company nobody's heard of.
Impressive specificity for a 6-minute show: named lead investors, exact valuations, check sizes, round labels, and granular product metrics like vehicle coverage percentages and manufacturer agreement counts. The Partly segment is the strongest, citing four years of build time, a $10M spend figure, and concrete data-moat details.
They spent four years and $10 million building Interpreter on vehicle parts data - government records manufacturer feeds their own teardowns parts interpreters annotating by hand.
91% vehicle coverage, 50-plus manufacturer agreements.
The hosts pose pointed diagnostic questions and maintain a healthy skepticism throughout - the 'product need vs. open window' framing on Baseten and the Azure margin squeeze on Engram are genuinely probing. However, the format is news commentary with no interviewee to push back on, so craft is limited to internal debate quality rather than interview discipline.
Are they raising $1.5bn because the product needs it, or because the window's open and you grab the money while you can?
Here's my problem with the pitch - they're selling a memory layer to fix wasted tokens. But if that layer runs on Azure, Microsoft's pricing can eat the margin.
Computed from the transcript - who did the talking, and the words that came up most.
Baseten’s $1.5 billion Series F puts AI infrastructure back at the center of startup fundraising, while Engram and Partly show investors still paying up for narrower, cost-saving AI systems. In this episode AI infrastructure startup Baseten raises $1.5bn at up to $13bn valuation - The Next Web AI infrastructure startup Baseten raises $1.5bn at up to $13bn valuation The AI infrastructure startup raised $1.5bn in a Series F, with Australia’s Blackbird VC putting in what may be the largest single outlay yet by an Australian firm.
Transcribed and scored by The B2B Podcast Index.
Baseten just closed a billion-five at a thirteen-billion valuation - and it's their fourth fundraise in eighteen months. At some point, the round IS the product. This is Startup Fundraising. Three deals on the tape today - a mega-infra raise a stealth memory play coming out swinging and a New Zealand auto parts company nobody's heard of.
From Ana-Maria Stanciuc at The Next Web: Baseten has raised $1.5bn in a Series F round that values the AI infrastructure startup at up to $13bn a number that arrives barely 18 months and four fundraises into the company’s current growth spurt. The round was led by the US investors Sands Capital and Wellington Management, Four fundraises in eighteen months to get to $13bn - that's the part I keep circling. Every round leaves a dilution trail so by Series F the early checks have been ground down to something you'd need a microscope to find on the cap table.
Right, and that velocity is what I want explained. Are they raising $1.5bn because the product needs it, or because the window's open and you grab the money while you can? Sands Capital and Wellington led - which matters.
Not a16z, not Founders Fund. And the headline hook is Blackbird out of Australia writing what they're calling their biggest check ever. Biggest check ever, amount undisclosed. Love a record you don't have to put a number on.
It's still a clean lead/follow read, though - Sands sets the terms, Blackbird gets the local-hero story because the founders are Australian. Two different jobs on one cap table. So over the next two years this has to hold up: Baseten sells the plumbing under everyone else's AI pitched as cheaper than the big providers. At $13bn, that margin had better be real, not a discount they're eating just to win logos.
From PR Newswire: Engram the company building the learned memory layer for AI emerged from stealth with $98M in funding from General Catalyst Kleiner Perkins Sequoia Capital Factory Modern Amplify Partners Neo and notable angels and advisors including Assaf Rappaport co-founder and CEO of Wiz Andrej Karpathy co-founder of OpenAI and Pieter Abbeel AI and robotics pioneer and co-director of the Berkeley AI Research Lab. Engram comes out of stealth with $98M, and the partner list is Microsoft, Notion, Harvey.
They're partners, not customers, and that word can cover an empty revenue line. And the syndicate is General Catalyst, Kleiner Perkins, and Sequoia all on one cap table at launch. Hard to call that a lead investor; it reads like a committee. Somebody set those terms and the others chased.
Here's my problem with the pitch - they're selling a memory layer to fix wasted tokens. But if that layer runs on Azure, Microsoft's pricing can eat the margin. Right after the Baseten round we just hit - same gravitational pull. The biggest US names pile into the splashiest infrastructure-adjacent bet and the foreign founder gets the smallest check later in the show.
Karpathy and the Wiz CEO on the angel list the 'learned memory layer' framing - it's a beautifully decorated category that doesn't exist yet. I want to see the unit economics two years from now, when every enterprise already pays Microsoft for a version of the same thing. This one's from SiliconANGLE: Partly Group Ltd. a New Zealand startup using artificial intelligence to change up the automotive parts business announced today that it has raised $50 million new funding at a $500 million valuation.
It’s also opening its first U.S. operation, betting that the country’s collision repair sector is ready to buy software it has so far had to do without. Okay, this is the one I've been waiting for all episode.
Partly - $50M, $500M valuation, AI for auto parts in collision repair. Out of New Zealand, first US office just opening. After Baseten at $13bn and Engram's whole stealth-launch routine, this is the round I'd actually root for. Auto parts cataloging is operationally brutal and deeply unsexy - nobody pitches a Series whatever on collision repair to look cool.
And after a day where the smallest check on the tape belongs to the only non-US founder - there's your concentration issue in miniature. Partly raises $50M, opens in the States, and if their numbers are anywhere near the round size, you're looking at a ten-x revenue multiple. What I want is the cross-border read. A New Zealand cap table raising US dollars at a US valuation, then planting an office here.
There's a currency and jurisdiction question sitting under that $500M number that nobody's printing yet. They spent four years and $10 million building Interpreter on vehicle parts data - government records manufacturer feeds their own teardowns parts interpreters annotating by hand. You don't fake that with a wrapper. The moat is all the grunt work.
91% vehicle coverage, 50-plus manufacturer agreements. The valuation either makes total sense or it's insane, and the answer's in auto parts distribution margins - not the AI label. If Startup Fundraising helps you think more clearly about fundraising take a second to subscribe and leave a review wherever you're listening. It really helps other founders and operators find the show.
You'll find links to every story we covered today in the show notes. So if something's especially relevant to your raise, your board, or your next investor conversation, it's there to dig into. That's Startup Fundraising for today. This is a Lantern Podcast.
Other episodes covering the same guests and topics, from across The B2B Podcast Index.
Platform Shifts, AI, and the Future of Consumer InvestingVenture Unlocked · on General Catalyst85 / 100
Andy Chen: The Convenient Cofounder PenaltySand Hill Road · on Kleiner Perkins82 / 100
117. Infrastructure at the Edge of Everywhere with Armada, Microsoft, and DISAAll Quiet on the Second Front · on Microsoft Azure81 / 100
The Defence Capital Supercycle: Europe's New Venture Capital InfrastructureSecurity Sutra · on Sequoia Capital80 / 100
Treat Metering Like Finance: Building Data Platforms for Consumption EconomicsData Engineering Podcast · on Microsoft Azure79 / 100
Best of Build Mode: Think like a VCEquity · on General Catalyst78 / 100
Kling’s $3B Raise Keeps AI Funding in Overdrive - July 03, 202678 / 100AI Infrastructure Megarounds Keep Coming - July 02, 202671 / 100Nigeria’s Startup Fund Meets the AI and Space Capital Rush - July 01, 202698 / 100Sharon AI’s $1.6B check puts compute back at the center - June 30, 202675 / 100The Seed Round Has Swallowed Series A - June 29, 202673 / 100