# Invest Like the Best: The Category Nobody Wanted to Hear About Nine Months Ago Just Drew Thirteen Term Sheets Source: Realpha Blog (blog.getrealpha.com) Original article and charts: https://blog.getrealpha.com/en/blog/iltb-2026-11-06-ramtin-naimi-on-the-state-of-venture-investing-inv/ > Notes on the Invest Like the Best episode 499, published 6 November 2026, where host Patrick O'Shaughnessy talks with Ramtin Naimi, founder of the seed firm Abstract. He scores today's opportunity set a 10 out of 10 and its frothiness a 6.5 to 7 against a 10 for 2021, and says the spread between the lowest and highest price on the same company now runs 4x. Educational listening notes only — no investment advice, no price targets, and the judgements here may be wrong. Published: 2026-11-06 Locale: en Tags: venture capital, podcast-notes, valuation discipline, brain-computer interfaces, private markets ![A Silicon Valley office corridor at dusk, a row of empty chairs receding chair by chair into the depth of the frame, and at the far end the one lit glass conference room with a sketch of a head-worn device still on the whiteboard](/covers/iltb-2026-11-06-ramtin-naimi-on-the-state-of-venture-investing-inv-cover.png) > When you walk, I walk; when you trot, I trot; when you gallop, I gallop. But when you race off and leave the dust behind, all I can do is stare after you. > > —— Zhuangzi, "Tian Zifang" (Warring States period) In episode 499 of Invest Like the Best, published on 6 November 2026, host Patrick O'Shaughnessy talks with Ramtin Naimi, founder of the early-stage firm Abstract, about one thing only: what is happening on the deal table in San Francisco right now. Naimi gives two scores — against his nearly ten years in venture, he rates today's opportunity set a 10 out of 10, and the frothiness a 6.5 to 7, with 2021 as the 10. His worked example is the brain-computer interface company Conduit: when he seeded it nine months ago nobody in the Valley cared about the category, and in the eight to ten weeks before this recording the category got hot and Conduit's new round came in heavily oversubscribed with thirteen term sheets. The condition on all of it: these are the deals crossing his own desk, and he admits that the lowest and highest price on the same company now differ by 4x, so even someone looking at these every day cannot call where one will clear. ![Three bars are compared on the same zero-to-ten scale: today's opportunity bar reaches the full ten, today's bubble bar only about seven, and the 2021 bubble bar also reaches the full ten, showing that opportunity and bubble are not maxed out together this time.](/figures/iltb-naimi-two-scores-en.svg) ## Nine months ago, nobody wanted to hear about brain-computer interfaces When Naimi invested in Conduit, brain-computer interfaces were a cold category in the Valley. The phrase sounds like science fiction and describes something much narrower: getting a machine to read the signals in your head. Back then most people's mental image came from Neuralink — implant a chip in the brain, get a beautiful readout, pay for it with surgery. Conduit's approach is a device you wear on top of your head, no surgery, and a worse readout. So the mood in those meetings nine months ago was, as he puts it, plain enough: anybody who met these founders would tell you they were brilliant, then walk out of the room and do nothing. The people test passed. The market test did not. ## What they brought was a device that guesses your next sentence before you type it At that first meeting Conduit had a small amount of human training data. What the small sample showed was the interesting part: a subject would type out the sentence in their head, and before they typed it the device and the model had already inferred what they were thinking with about 80% accuracy. Put the model's guess beside what they actually typed and the words differ, the meaning does not. More to the point, the model gets more accurate as the data grows. If that line holds, the small sample from nine months ago was the beginning of a slope. By the time of this recording they had 30,000 hours of human training data and were about to run a million-hour training set. Naimi followed the slope out to a picture that made me laugh: five or ten years from now, you and I wear a small thing over the ear that understands what we are thinking, can tell an actionable thought from an idle one, and hands the actionable ones to our agent to run in the background. What one person's output then looks like — his tone when he got there sounded like he was still digesting it himself. ## Eight to ten weeks, and the whole Valley changed its answer Then brain-computer interfaces got hot. Naimi says it happened in the eight to ten weeks before the interview — same people, same approach, same slope, except now somebody is willing to believe a trillion-dollar company can come out of this, so thirteen term sheets arrive and the new round prices that trillion-dollar imagination in. What stays with me from this stretch is how small the before-and-after difference is. Over nine months Conduit did not change what it was doing; the consensus around it changed. Naimi's own read: smart people tend to understand a market well before the market starts paying attention, so the market they picked was not a bad one — they were simply early to it. ![On the left an empty dashed circle stands for zero term sheets nine months ago, and an arrow points right to a block of thirteen filled dots, with the label "same team, same method" in the middle, showing that only the count changed.](/figures/iltb-naimi-zero-to-thirteen-en.svg) ## Etched already played the same scene He gives a second example. Gavin Uberti of Etched met every good fund in the Valley, and Naimi says he cannot find a single investor who met Gavin and thought he was anything but brilliant. What most of them concluded on the way out was that a startup cannot outcompete the incumbents at AI chip fabrication. That just does not happen. Then Sequoia blessed the company at a 10 billion dollar valuation. After which everybody changed their answer: startups can run circles around incumbents in this market after all — and every other chip company on the market raised at an astronomical price, because the category was now obvious. Patrick tries to compress this into one metric and asks whether there is an ideal number of term sheets. Naimi's answer: a company with fifteen term sheets gets very expensive, a company with two or three prices reasonably, and if you are a good investor the lower competition is what lets you argue your way into a core position. Then the line I wrote down: for a company to get that many meetings with good investors, a lot of people had to vouch for the founder; so walking out of all those meetings with only a handful of term sheets usually has nothing to do with whether the founder is any good — people decided the opportunity was not interesting. And in venture the opportunity usually is not interesting; in a small subset of cases it is, and telling those apart means doing the work of learning the category, which most people skip. ![The horizontal axis is the number of term sheets and the vertical axis is price: a point in the lower left is labelled two or three sheets, a fair price and a core position still available, while a point in the upper right is labelled fifteen sheets and expensive, with only a dashed line joining the two.](/figures/iltb-naimi-term-sheet-count-en.svg) ## What the people chasing are doing Which brings us to the loudest part of the episode. Rounds move faster than ever and so do follow-ons. The first tranche of a hot company goes to a few firms at preferential pricing; the firms that lost used to write a small cheque to keep the relationship warm and wait for the next round. They no longer wait. They ask what it costs to get in today, and the answer is usually three to five times the first tranche price, closing alongside it. Silicon Valley has given that second group an unkind nickname, and Naimi says anyone who claims they have never been in it is lying. ![Three horizontal bars of different lengths run rightward from the same starting point, marked one times, three times and five times, but all end at the same vertical dashed line labelled "same day close", showing that chasers pay a multiple of the price yet enter at the same moment.](/figures/iltb-naimi-chasing-price-en.svg) He does not think the tranching itself is unhealthy — these companies are capital-intensive enough to absorb it, and founders would rather not raise all of it at the primary price. The chase is the unhealthy part: offers thrown out fast, then changed or pulled after the fact. He has seen more deals renegotiated after being offered in the last six to eight weeks than at any point since 2021. The same pressure sits on the investors themselves. In this market a markup on one of your companies is the bare minimum for keeping pace, so markups stop feeling good. Every three weeks something good happens in the portfolio and you feel fine about yourself; the next three weeks you watch good things happen to companies you are not in, and you are behind again. The money cannot keep up either. A fund raises against an eighteen to twenty-four month deployment period, and Naimi says the mega platform funds have come back to market two or three times inside that same window — deploying roughly three times faster than modelled. Two unnamed numbers: one fund raised 2.5 billion dollars, started deploying in July, and told him it would be back in market in January; another took 3 billion dollars nine months ago and is back asking for 5 billion, and will get it. The gap has now reached founders. In the last eight to twelve weeks he started seeing something new: a VC hands over a 150 million dollar term sheet and then asks whether funding it in mid-January, three months out, is acceptable — because the current fund's money is already committed to existing companies, the new fund is still closing, and the capital facility is not set up yet. Another company got a 50 million dollar term sheet where the lead committed 40 million, offered to wire 5 million at closing, and needed another eight to ten weeks for the rest. ![Within the same eighteen-to-twenty-four-month window, the top row is one unbroken planned deployment bar while the bottom row is cut into three short segments, each with an upward arrow back to a new fundraise, showing the actual pace is three times the plan.](/figures/iltb-naimi-deployment-speed-en.svg) The exit end is quieter. Some institutional investors have 30 to 50% of their assets in venture, but unpack it and that exposure sits in three or four companies — SpaceX, Anthropic, OpenAI, Databricks and the like. Until those names generate liquidity, nobody's exposure comes down; and the institutions without them have had very little cash back this cycle. ## "Everyone else made money — am I the one falling behind?" I know this feeling. Somebody else's position triples in a year, mine sits still, and my first instinct is that I picked wrong. What Naimi says to founders transfers whole: seeing numbers growing 10x, 15x, 20x year over year and concluding you have been left behind happens because you are comparing your growth to the prettiest absolute figures in the market. His way of unpacking it is to look at what that revenue is made of. Companies growing at that rate usually have something unclean in the revenue: margins that are thin or negative; growth that is violent but concentrated in a few customers; data sales where last month's number times twelve becomes annualised revenue; consumer rather than enterprise revenue; usage-based rather than contractual revenue. The ceilings and durability of those shapes differ enough that putting them on one table and comparing sizes tells you nothing. The counterexample is his own portfolio: companies growing 200 to 700% this year while margins travel from negative 15% to positive 35%. Patrick adds a figure — Anthropic said its gross margin for the first half of this year was 85%. The same day, Anthropic's S-1 became public, and Patrick calls the most glaring stat in it this: 140 billion dollars raised, 10 billion spent, 132 billion still on the balance sheet. ![A line climbs from minus fifteen percent below the zero line up to plus thirty-five percent above it, crossing that zero line on the way, showing the direction of gross margin rather than just its level.](/figures/iltb-naimi-gross-margin-crossing-zero-en.svg) So now when I see a beautiful growth rate, the first question is what kind of revenue it is and which way the margin is travelling. That question works just as well on listed companies, and it never returns the verdict "I lost" — only "these two do not belong in the same comparison." ## "Why does the safest-sounding approach make me uneasy?" Patrick asks the reasonable question: given how crazy the early stage is, should you walk to the other end of the spectrum — the companies that are already large, legible and growing fast, where competition is thinner? Naimi says that does buy you exposure to great companies, then offers an arithmetic worth copying down: in 2021 you could invest in Stripe at 100 billion dollars. Stripe was a great company and still is. Today it is 179 billion dollars — call it an 11 to 12% annual return over five years. That beats the bottom 75% of venture funds. So buying the best companies at the highest prices can probably put you in the top quartile. What makes him uneasy is how crowded that has become. Over the last eighteen months he has watched a record number of firms arrive at the same idea: there are a handful of obviously great companies that can absorb enormous cheques, and demonstrating access to them lets you raise a large fund and sit there passively. His test is blunt — he cannot think of many times in history when an easy strategy was also a good long-term strategy. I took that back to my own holdings. My semiconductor equipment and software positions were each bought for their own reason, but when I ask honestly how many people now hold them for the same reason I do, a few of the answers make me uncomfortable. Crowding does not tell you to sell. It tells you that if you are wrong, nobody standing next to you will take the other side. ## Worth a look - Invest Like the Best, episode 499, 6 November 2026 — the source for this piece - Anthropic's S-1 registration statement, made public the day this episode was recorded - The companies Naimi named, usable as a watchlist for the next few years: Conduit, Etched, Applied Compute, New Limit, Chai Discovery, Isomorphic, Physical Intelligence - The newer financing structures he mentions are worth reading up on: growth debt underwritten against customer value, and firms funding companies with compute instead of cash ## One thing to take away **The same people doing the same thing can be priced an order of magnitude apart either side of consensus; the extra money buys you other people's agreement.** The only thing that changed about Conduit in nine months is that bystanders changed their answer, and Etched waited for Sequoia to name a price before being believed. Naimi splits venture returns into two sources: reactive, where picking well gets you to the top quartile; and proactive, where backing markets that are not yet obvious is what gets you to the top 5%. The difference is whether you pay before consensus or after it. ![A distribution curve of venture fund returns has two vertical lines marked on its right tail: the left one is the top quartile reachable by paying up after consensus, and the one further right is the top five percent only reachable by backing the idea before consensus.](/figures/iltb-naimi-before-after-consensus-en.svg) Something I have tried that works outside investing too. Think of one person around you whose current project nobody is backing yet — a career switch, a book no one is waiting for, caring for someone who is hard to care for. Go find them once this week, offer no advice, ask only for details, and afterwards put the date and one sentence they said into your phone. Open it again in three months.