Newsletter · · Ashutosh Agarwal

A Medtech Founder Closes His Series A With $20,000 Left in the Bank - The Raise - Week of September 28, 2026

The Raise for the week of September 21 to 27, 2026. Aortix CEO Jordi Martorell on spending two years and nine months raising a Series A and closing with $20,000 in the bank, set against a week when TypeSafe went from a $200 million seed valuation to $10 billion-plus talks and Snorkel, Enveda, Solkoa, Factory and OpenEvidence all raised.

The Raise

Week of September 28, 2026: A Medtech Founder Closes His Series A With $20,000 Left in the Bank

Week of Sep 21–27, 2026: a seed-stage startup went from a $200 million valuation to $10 billion+ talks in about seven days. Factory tripled its valuation, OpenEvidence hit $15 billion, and Anthropic's IPO slipped. Our founder story runs the other way: 2 years and 9 months to close a Series A, with $20,000 left in the bank.

On Thursday, September 25th, Tech Brew Ride Home covered a story that we had to read twice.

TypeSafe AI, the company behind a new kind of AI model called Jev, came out of stealth the week before with a $40 million seed round led by DCVC. PitchBook put the valuation at $200 million. Seven days later, The Information reported that TypeSafe was in talks to raise more than $1 billion at a valuation above $10 billion. The host couldn't quite believe it either:

"Last week, it was valued at $200 million. This week, possibly more than $10 billion."

"this would without a doubt be the fastest fundraising glow-up of all time."

That's a 50x markup in one week. No new revenue, no new product. The only thing that changed was that people were talking about it.

What makes it stranger is how un-flashy Jev is. The Elon Musk Podcast (Sep 27) walked through it: Jev deliberately refuses to write sentences. It isn't a chatbot. It picks from preset answers and attaches a confidence score, the way a triage nurse sorts patients. The pitch is that most AI work inside companies is really sorting (yes/no, which bucket, flag or don't flag), and that doing it with a model that writes paragraphs is slow and expensive. The founding team has real pedigree: former OpenAI researcher Diogo Almeida, former Meta research engineer Sasha Sheng, and entrepreneur Eric Gaffney. Snorkel AI's Alex Ratner, who has been in machine learning for a decade, joked about it on TBPN:

"I remember when a Jev was called a classifier"

He has a point. "Classifier" is the old, boring name for exactly this kind of model.

Here's our view. We don't doubt the product works; people we trust say it's fast and cheap. But $200 million to $10 billion in a week isn't price discovery. It's a stampede. Jason Lemkin said as much on 20VC: he called the launch "a master class in launch PR". The Elon Musk Podcast went further and described a coordinated wave of founders and engineers all posting about Jev at the same moment, "manufacturing the perception of an absolute consensus."

The best explanation of why investors keep doing this came from Rory O'Driscoll on 20VC. He was reacting to a widely shared memo from Menlo Ventures about "playing the game at the top of the cycle":

"Some part of the increase in size is justified by math. But let's be honest, I think some part of it is justified by FOMO and the fear of missing the next cursor."

That's the whole week in one sentence. Every investor who missed Cursor (the AI coding tool whose value exploded) is now terrified of missing the next one, so they pay almost any price for anything that looks like it. Keep that in mind as you read the rounds below. And then read our founder story, which is the exact opposite.

Tech Brew Ride Home, "Is Copilot Microsoft's Everything App?" (Sept 25); Elon Musk Podcast, "Why Jev Refuses to Write Sentences" (Sept 27); 20VC, "Meta's Muse Hits No. 1 … Menlo Sounds the AI Bubble Alarm …" (Sept 24).

This Week's Rounds

  • TypeSafe AI: $40M seed at a ~$200M valuation (per PitchBook), led by DCVC. Now reportedly in talks to raise $1B+ at $10B+. (The seed is announced. The $10B round is only talks.) It builds Jev, a "classifier" model that answers with a pick and a confidence score instead of text. Per the Elon Musk Podcast, it runs roughly 89–400x cheaper and about 200x faster than general-purpose language models on sorting-type jobs. Tech Brew noted the buzz "may be driving up its compute costs," which gives it a real reason to raise more. Tech Brew Ride Home (Sept 25); Elon Musk Podcast (Sept 27).

  • Factory: $200M at a $5B valuation. That's triple its last valuation. Factory sells AI coding agents (software that writes and fixes code on its own) to big companies. Harry Stebbings is an investor. On 20VC's mock investment committee, both Jason Lemkin and Rory O'Driscoll voted yes. Lemkin's reason was data sovereignty, meaning companies keeping control of their own data. Having just come back from Salesforce's ($CRM) Dreamforce conference, he said executives "don't trust Anthropic and open AI with their data. They don't trust it." So they want a coding tool from a company that doesn't also sell the AI model underneath. He also pointed to Brad Gerstner as a reason for confidence. O'Driscoll's reason was simpler:

    "coding is the mother load. It's that simple."

    With Cursor now off the table, he said the only independents still selling this to large enterprises are Factory and Cognition. The show didn't name the round's lead investor. 20VC, "Meta's Muse Hits No. 1 … Factory Triples Its Valuation to $5 Billion …" (Sept 24).

  • OpenEvidence: $250M at a $15B valuation (up from $12B in January), per Business Insider as relayed on Tech Brew Ride Home. It's an AI search engine for doctors, and according to CNBC more than two-thirds of U.S. doctors use it for diagnosis and treatment advice. It has raised over $1 billion in the past year from Thrive Capital, DST, GV, Kleiner Perkins, and Sequoia, and NVIDIA ($NVDA) is also an investor. The interesting part: it "could be open to selling itself." OpenAI and Anthropic are pushing into healthcare, and buying OpenEvidence would give either one instant access to doctors. Tech Brew Ride Home, "Is Copilot Microsoft's Everything App?" (Sept 25).

  • Snorkel AI: $350M round. Snorkel makes specialized training data, the carefully labeled examples AI models learn from. CEO Alex Ratner came on TBPN to announce it (and got the gong). He said AI is shifting from "data 1.0 to data 2.0": early on, models just need huge amounts of data, but as they mature they need smaller amounts of expert, specialized data. The show didn't give a valuation or lead investor. TBPN, "Harvey's Margin Whiplash, Human Powered Agents, McLaren Rebrand, Insects > Humans?" (Sept 22).

  • Enveda: $311M round. CEO Viswa Colluru announced it on TBPN (the transcript spells the name "Inveda"). Enveda uses AI to find new medicines hidden in natural compounds. He described it as "a sequencer, but for life's chemical code instead of life's genetic code." His argument is that humans have identified about 400,000 natural compounds out of an estimated 1–10 billion. The first two drugs: a non-steroid pill for eczema and asthma (he said no new non-steroid oral asthma drug has been approved in over 25 years), and a once-daily pill based on a hormone the body releases after a sprint. He called that one "the chemistry of exercise," aimed at helping people keep weight off after they stop taking GLP-1 drugs. TBPN, "4 Tech Arguments (Number 3 Will Shock You), Mansion Section, John Arnold Joins" (Sept 25).

  • Solkoa: $75M for its first plant, "Solkoa One." Solkoa turns refined rare-earth ore into the metals used in magnets for EVs, fighter jets, data centers, and wind turbines. The 500-ton-per-year plant is in Nevada, with commissioning set for July 4th next year. The CEO on TBPN:

    "we are completely unable to make these metals in the West. Solkoa is changing that."

    One product is samarium, which is used in F-35s and guided missiles. The CEO pointed out that China is one of the only producers and "banned the export of it," which has left the defense market badly short. About a year ago Solkoa switched from recycling scrap to working directly with mines, because recycled material just doesn't come in big enough volumes. TBPN, "Meta Connect Reactions, Zuck's Beer Pong Controversy, New Bentley EV" (Sept 24).

  • Legora: reportedly raising at $11B after hitting $200M in annual recurring revenue. (A round in progress, not closed.) Legora makes AI software for lawyers. On 20VC, Lemkin was a fan but held back. The Information had just reported that rival Harvey's gross margins had fallen to around minus 50%, meaning it costs more to deliver the product than customers pay. Lemkin said that if margins are "minus 50% and going down, I might be slightly nervous", enough that he wouldn't lead the next round. O'Driscoll put it more bluntly: "No, I won't do it at 10 billion." 20VC (Sept 24).

  • Instinct: an update on last week's $1B-at-$10B story. On This Week in Startups, Jason Calacanis showed the round-by-round history from the data provider Harmonic: a $25M seed at $50M post-money in April 2026, a $75M Series A at $500M in August, and a $250M Series B at $2.5B led by Benchmark and Index Ventures. Reportedly about $350M of the targeted $1B is already raised, with Sequoia and Benchmark "circling." (Calacanis said on air that the dates "need to get checked," since the A and B look like they came about a month apart.) The founder, Noah Shin, is 23, and Business Insider says the app has 100,000 users. Calacanis wasn't impressed:

    "I do get a Clubhouse vibe right now in terms of the playbook here"

    His guest was harsher about the product: "We have a couple of people who've used Instinct, and the experience was just not great." This Week in Startups, "VCs Would Bet on Open-Source AI Over OpenAI and Anthropic | E2341" (Sept 23).

  • Crusoe: an update on last week's $3.9B round at $30.9B. New detail from 20VC: Crusoe has about $140 billion of total contracted value, meaning customer commitments it has signed. Lemkin still passed on the mock deal: "this is all a spreadsheet investment… It's just a hint too expensive to hit our margin, despite the growth." O'Driscoll's warning is the one worth remembering. Data-center companies borrow heavily, so they are very exposed if AI growth slows. In his words, a "slowdown when you've got like four or five to one leverage can be pretty brutal." 20VC, "… Crusoe's $3.9 Billion Round. Is the Data Centre Trade Overheating?" (Sept 24).

Founder Story of the Week

Jordi Martorell, Aortix Global Medical Device Podcast, "#473: The MedTech Odyssey: Bridging Academic Science to Series A Success" (Sept 21)

While TypeSafe went from $200 million to $10 billion in a week, a founder in Barcelona spent two years and nine months raising a Series A. When the money finally arrived, his company had $20,000 in the bank and $250,000 in debt.

Jordi Martorell, CEO and co-founder of Aortix, told the story on the Global Medical Device Podcast (episode #473, Sep 21). It's the most honest look at how a raise actually goes that we've heard in a while.

The science came first. Martorell did his PhD at the Harvard–MIT Biomedical Engineering Center, studying how blood flows through arteries, and then went back to Barcelona to teach and research. Aortix makes a patch that dissolves inside the body. It's delivered through a catheter to repair aortic dissection, a tear in the body's main artery. He explained why it matters in plain terms:

"In the end, we have a patch to cover a hole, you know? Yeah. And it's a hole in your main artery. If you have a hole in your main artery, you die."

He didn't officially start the company until October 2018, after years of grants and research. The device is complicated. It has seven major components, and he joked about how many attempts it has taken:

"Overall, we have done more than 350 iterations of the device, you know, combined. Yeah. So, I mean, we suck."

He also noted they haven't changed the core idea once since 2018.

Round one: 75 friends in a room. In 2019, with no VC interest, he raised from people who knew him. Friends had been asking him for years when the poor scientist would finally make some money, so he gathered about 75 people in a room, pitched them, and was completely upfront about the risk:

"telling a hundred percent of your investors that once you have given me a dollar, you have lost it. Forget about it. Like it's, it's binary."

"our first seed round was in 2019. Okay. And it was, we raised 650,000 from, from, from, from just friends."

Round two: crowdfunding, run as a PR campaign. By 2021 VCs still considered Aortix too early for a Series A, and there were "literally three VCs at the time who could invest in us." So he raised a bridge round (a smaller round to get the company to its next big raise) on Capital Cell, a Spanish crowdfunding platform for health and science companies:

"we actually raised 2.3 million out of that with 500 plus investors."

He calls this "more like a PR strategy" than a financing, because it spread the word and brought in hundreds of backers. Along with 2.5 million from the EU's EIC Accelerator program, that money kept the company alive for more than four years.

The clever part: cap table engineering. You'd think 600 small investors would scare VCs off. Martorell set it up so it didn't:

"We have now like about 600 investors globally, which there's a lot of, I'd say public companies that don't have so many investors."

"So the first 60, they were all syndicated in me. Right. So I signed for everyone."

The first group gave him power of attorney (legal permission to vote their shares). The second group invested through a single company that he alone runs, which he described as "kind of like a mini spec." So when VCs showed up, they only had to deal with the founders:

"The VCs, in reality, they were only negotiating with us."

If you're a founder thinking about a community round, this is the part to copy.

The Series A grind. He started the process at a medtech conference in September 2022 and closed in May 2025. Some investor conversations lasted 14 months and still ended with the standard medtech answer: come back when you have first-in-human data (results from the first trials in people). A big part of the problem was geography:

"impossible to get a series A with a foreign VC as a leading guy. Impossible."

He says Spanish VCs can support maybe "10 series A per year tops" across the whole ecosystem, while roughly one new company is founded every week. He was self-aware about how foreign investors see Spain: "we are these Spanish guys. No, you know, they play soccer very well, but what about that?" Costs made it worse. Animal studies, clinical work, and materials have all roughly doubled since COVID, and in his words "the party is expensive."

Then, in March 2024, the round fell apart. The prospective lead investor, who hadn't yet issued a term sheet, got cold feet. Martorell rebuilt the round around the existing investors who'd been following it. They doubled down, asked for (in his words) "a very, very good valuation for them," and new investors joined on those terms. With a top-up that closed a few months ago:

"it goes to more than 15 million euros. So about $18 million more or less."

And how close it came:

"we closed around with 20,000 in the bank and like 250,000 in debt."

The lesson he saved for last. Right at the end, almost in passing, he said something the host said he'd "buried the lead" on:

"100% of the investors that we got in the, as VCs or yeah, as VCs told us no at some point."

Every VC now on his cap table turned him down at least once. None of them came through the big healthcare investor conference he'd attended five years running. They came from years of turning up at every major vascular surgery conference since 2018. One of those contacts is a strategic investor, a large company in the field, and he says one of them "is going to acquire us." His advice to other founders:

"You need to be there. You need to be resilient. You need to show progress."

He was also candid about what came after. He called life with an institutional board "a tough first year having VCs," after years of reporting to nobody. And his main advice is basically "don't do what I did": "that's not a recommendation. Get to first in human as fast as you can."

Why this is the story of the week. It's the other half of the market. At the top, a coding agent triples its valuation and a seed-stage startup gets offered a $10 billion price within days. Meanwhile, a company with a real medical device and a real patient problem spent almost three years and nearly ran out of money raising €15 million. Both are "venture capital." They just aren't the same market anymore. If you're a founder outside the hot categories, trust Jordi's version of the job over anything you see on X.

Also Heard

  • Bending Spoons started with $40,000 its VC didn't want back. On All-In (Sep 23), CEO Luca Ferrari told the origin story. The company had just agreed to buy Eventbrite ($EB), after picking up Miro last week. Ferrari's first startup, an AI company launched in 2010 ("very early, too early, clearly"), failed. Rather than pay the legal costs of winding it down, the investor let the founders keep what was left: "about $40,000 in capital we'd raised from the VC." They "enthusiastically turned it into seed financing" for Bending Spoons. Their first acquisition was a keyboard-customization iPhone app for about $10,000. The model since then is to buy products people already use, run them with small, very strong teams, and pay for more deals with cash flow and debt instead of selling shares:

    "when we IPO'd, we had only quote unquote only raised about half a billion dollar in primary equity and we were at roughly 20 billion in valuation."

    He also explained why private equity firms can't copy this. They keep each company separate so it can be sold later, which means they can't share engineers or technology across their portfolio. And a stat we liked: 800,000 job applications last year, fewer than 300 hires. All-In, "Luca Ferrari, Bending Spoons CEO: The $40K Origin Story, Buying Product-Market Fit & Why Private Equity Can't Compete" (Sept 23).

  • Fleet.co bootstrapped for seven years, then let the founders cash out. On Top Founders (Sep 24), co-founder and CEO Seyvan Marjan explained how Fleet, a Paris company that rents laptops and phones to businesses on subscription, grew from zero to about 40 million in revenue across 20 countries without outside money. It survived a 2023 slump when hiring among its customers fell about 70%. When it finally took outside capital in February 2026, at about 30 million in revenue, the deal was all secondary. That means the founders sold some of their own shares, and none of the money went into the company. It was valued at 100 million, using roughly two-thirds equity and one-third bank debt (the show didn't say whether in dollars or euros). His reasoning: "So the company doesn't need outside capital." Fleet makes close to 10 million a year in EBITDA (roughly, operating profit), so it doesn't need to raise. It's a useful example of taking money off the table without giving up control. Top Founders, "How He Makes $40M/yr Renting Laptops for $50/Mo" (Sept 24).

  • Bessemer is raising $5.75 billion, the largest fund in its history. $4 billion of it is set aside for growth-stage companies. Partner Samir Dalakia explained why on Bloomberg Tech (Sep 23):

    "these companies need a lot more capital than they used to need. And they stay private longer than they ever have. And we think that's a permanent structural shift."

    He said companies used to go public after 5–6 years and now take "an average of like 12, 14." He also mentioned Bessemer has put $650 million into Anthropic across four rounds, starting with a Series D two and a half years ago. Bloomberg Tech, "Anthropic's New Model, the Robot Economy and Apple Health" (Sept 23).

  • Anthropic's IPO slips, and the price may fall with it. Per the Wall Street Journal, as discussed on All-In (Sep 26) and Pivot (Sep 22), the Anthropic IPO originally planned for October has moved to November or later. Polymarket's odds of it happening in 2026 dropped from 96% to 76%. David Sacks said the company's leaders are hurting their own deal: "You have current leadership in the company saying that there's a greater than 10% chance of causing human extinction, and their own product is unsolved in this area. So think about the risk factors that's creating." Chamath Palihapitiya expects buyers to demand a big discount: even on a $100 billion revenue run-rate, "it's probably going to clear at a much lower price than anybody thinks." He was careful to call his "one or less" (trillion, vs. the $2 trillion headline) rough numbers. On RiskReversal (Sep 25), with Jim Chanos and Gary Marcus as guests, the discussion noted that OpenAI is pushing its own IPO filing out to 2027, and one guest called many of these valuations "fantasy." All-In, "Anthropic IPO at Risk, Meta's Muse Pop, Token Prices Fall, Open Source Gains Share, Alignment Fails" (Sept 26); Pivot, "Trump's Press Crackdown, Paramount Settles, and Anthropic's $2 Trillion IPO" (Sept 22); RiskReversal Pod, "Jim Chanos & Gary Marcus: Circular AI Financing, The Agentic Economy & Doomsday Scenarios" (Sept 25).

  • NScale files to go public at ~$35B. Run the Numbers (Sep 24) went through the S-1 (the document a company files before listing) for this two-year-old UK "neocloud," a company that rents out AI chips. It plans to list on the NYSE under NSCL, targeting about $35 billion per the FT, up from about $14.6 billion in March. The headline figure is $103.4 billion of contracts, mostly from Microsoft ($MSFT, up to $43.8B) and Anthropic (up to $44.6B). The catch:

    "Only $2.6 billion of that is live today. So, about 2% of that committed value."

    About 5% of its contracted chips are actually running. Its largest customer was 73% of 2025 revenue. It has also raised $3.1 billion in convertible notes (loans that can turn into shares). As a data point on the IPO window: a company can file to go public on contracts that are 98% not yet live. Run the Numbers, "The NScale IPO | A Neocloud S1 Breakdown" (Sept 24).

  • Andreessen Horowitz launches an "Academy" with $42M. CEO Gagan Biyani came on TBPN (Sep 22) to announce a one-year, in-person school in San Francisco for young builders. The first class of 50 is tuition-free, and Marc Andreessen and Ben Horowitz are joining the board. It's a clear bet that a network and time in Silicon Valley beat a degree. TBPN, "Harvey's Margin Whiplash, Human Powered Agents, McLaren Rebrand, Insects > Humans?" (Sept 22).

  • Unilever ($UL) buys Grüns, reportedly the fastest big exit in consumer packaged goods. On The Startup CPG Podcast (Sep 26), a VP of product talked through how the three-year-old supplement-gummy brand grew to 130 employees on a "taste to retain" strategy (make it taste good enough that people keep buying) before the sale. No price was given. The Startup CPG Podcast, "How 'Taste to Retain' Built a Billion-Dollar Supplement Brand" (Sept 26).