Newsletter · · Ashutosh Agarwal
Nobody Fears the Down Round Anymore - The VC Read - Week of July 16, 2026
Startups and venture newsletter for the week of July 16, 2026. Founders have quietly stopped fearing dilution and down rounds, and this week's podcasts argued over whether that freedom is the healthiest shift in a decade or a trap set by too much cheap money, alongside Gavin Baker's case that there are no dark GPUs and Wix's founder on running real software while the market prices it like a corpse.
The VC Read
Week of July 9–16, 2026: Nobody Fears the Down Round Anymore
Founders have quietly stopped being afraid of dilution and down rounds. This week's podcasts argue that's either the healthiest thing to happen to founders in a decade, or a slow-acting trap that too much cheap money is setting for the whole ecosystem. Plus: a marquee investor explains why "there are no dark GPUs," and Wix's founder on what it's like to run a real software business while the market prices you like a corpse.
A quick note on the week: the venture-macro conversation ran almost entirely through two shows this week, 20VC and The a16z Show. This Week in Startups, Invest Like the Best and The Meb Faber Show didn't surface fresh big-picture VC commentary in the last seven days that we could get to, so this issue leans on four strong episodes rather than a dozen thin ones. There was plenty to chew on.
The Big Debate: Is the Death of Dilution Fear Freedom for Founders or a Trap?
For most of venture history, a founder's nightmare was the down round: raising money at a lower price than last time, getting crushed on ownership, and handing tough investors the leverage to block a sale or force a fire-sale. Raising "too much" was dangerous. This week, a 20VC roundtable with Jason Calacanis and Rory O'Driscoll of Scale Venture Partners made the case that this fear has basically evaporated, and then argued about whether that's a good thing.
The Bull Case: Founders Have Been Set Free
Calacanis laid out the shift bluntly. "No one's worried about making their last round high-priced investors' money anymore. Literally no one is, because I believe investors have learned to accept 1x when it doesn't work out without drama, without blocking, without threats." In plain terms: if you raise at a $4 billion valuation and end up selling for $800 million, the late investors quietly take their money back and move on. "Founders, oh, I raised at $4 billion, but maybe I exit at $800 and I get an $80 million carve-out, they're just not worried," he said. He contrasted it with his own founding days: "My generation of founders, we were terrified of it. We were terrified of the expectations." O'Driscoll, initially skeptical, came around on air: "It's freed up the risk."
The logic is that this removes a real brake on ambition. If a big late-stage round can't hurt you on the downside but massively increases your odds of reaching a giant outcome, you take it. As O'Driscoll framed it, the real trade is "optionality versus upside. There's no doubt if the prize is $1 trillion, which it has been in at least 3 cases, it looks like, it really doesn't matter what it takes to get there. You just have to get there. And if skimping on it reduces the probability of getting there even 10%, it's a huge mistake." Calacanis pointed to Databricks, which has now done a dozen announced rounds (they cheekily named the latest a "Series M"), as founders happily raising round after round because each one only dilutes them a little. And there's data behind it: O'Driscoll noted that CARTA's numbers show dilution per round is actually going down, so founders can raise more times and still end up owning roughly the same slice.
He even used it to second-guess capital discipline. Cary Kilberg at Linear has raised just two rounds and famously refuses to meet VCs, a founder Harry Stebbings said "will return my fund one multiple times over." Calacanis admired it but wondered aloud whether extreme frugality is the right call "in 2026 when the prize is so large. If the exit's a couple billion, $5 billion, it's good. If the exit's $100 billion, then you just..."
The Bear Case: Too Much Money Is Quietly Breaking Companies
The counterweight came from Arvind Jain, co-founder of Glean (and previously Rubrik), on a separate 20VC episode. Asked what he'd most change about the startup world, he didn't hesitate: "There is too much capital available today for startups. And it's actually sometimes creating failure paths for people." His example was concrete and a little jaw-dropping, a startup that has only raised a seed round paying half a million dollars for a single engineer. "The startup founder is okay with it. The investors are okay with it. But it's just surely not a sustainable path to actually win. And they're paying it while Google is not. And Google knows that they don't need to actually buy talent like that." Cheap, abundant money, in this view, lets founders build cost structures they can never grow into.
Rory O'Driscoll's own caution rhymes with this. Even as he warmed to the bull case, he admitted he's watching the ugly side play out right now: "I'm watching a threat through my portfolio from a nonstandard VC right now that is blocking round after round after round." The "investors just take their 1x" world only holds for standard, reputable funds; the moment a founder has let a hedge fund or an opportunistic crossover into the cap table, the old blocking games come right back.
And there's a deeper worry lurking underneath the optimism: the whole "dilution doesn't matter" mindset only works if the giant outcomes keep coming. O'Driscoll was candid that the math has been warped by a handful of trillion-dollar winners. He recalled the early SpaceX backers who "did the original check when the rockets were still blowing up" and now own maybe 3% or 4% of the company, a return so enormous it makes ownership discipline look quaint. But for the other 99% of startups, he conceded, "raising too much eliminates the optionality of taking that billion-dollar exit," because a stacked preference stack can eat a modest sale alive. Twenty percent of a billion-dollar exit, he reminded listeners, "is life-changing, life-changing, especially with QSBS for now."
Where it nets out: this is the age of growth investing, as Calacanis put it, and the absence of downside blocking "adds velocity, not on the investor side, but on the founder side." Founders will raise more, more often, at higher prices, and sleep fine doing it. The open question is whether a generation that never learned to fear dilution is building companies that can survive a world where the trillion-dollar outcomes get rarer.
The Vivid Sideshow: Sam Altman's 5%
The same roundtable spent a long stretch on Sam Altman reportedly offering the US government 5% of OpenAI. It's a perfect illustration of how little founders now fear giving equity away. Calacanis, an early skeptic, talked himself into it: he compared it to how Klaviyo gave Shopify about 5% of itself so a giant partner wouldn't crush it. "It creates an unexpectedly large amount of alignment... I'm constantly shocked how much that brings you into the boardroom." His verdict: "I think as an investor, I take the dilution."
O'Driscoll thought it was a serious mistake, and his reasons are worth hearing. First, ownership doesn't buy love: "Microsoft owns 30% of OpenAI. If ownership stake resulted in besties, they'd be besties. They're not besties. They're in a stale marriage looking for a divorce, but can't quite pay the tax." Second, once you invite the government in, you don't control where it stops: "You start with pre-approval, suddenly you end up with an ownership interest, then you end up with a board member." He noted OpenAI has been publicly arguing that AI is so economically disruptive that America should overhaul its tax system, and warned that if you tell Washington you're about to wipe out half the labor force, "it's about an hour before Bernie Sanders says... maybe we should go for 50 [percent]." His bottom line: the danger narrative is exactly what let these labs raise the enormous sums they have, but "if you really are impacting a $30 trillion economy," 5% won't buy peace, it'll invite a much bigger grab.
Signals
"There are no dark GPUs." Gavin Baker, CIO of Atreides Management, gave the sharpest bull answer yet to the AI-bubble question on The a16z Show. His core point: the 2000 telecom bust was defined by "dark fiber," cable laid in the ground and never lit up. "At the peak of the bubble, 97% of the fiber that had been laid in America was dark. Contrast that with today. There are no dark GPUs." The chips being bought are running flat-out (his line: the papers are full of GPUs "melting"). He backed it with numbers: the US now has roughly $1 trillion of data centers with plans to add $3 to $4 trillion over five years; the industry has already spent more, inflation-adjusted, on data-center buildout in three years than the entire US interstate highway system cost over 40. Yet Google says it processed 150 times more AI "tokens" in the last 17 months, and the biggest spenders have seen their returns on invested capital rise about 10 points since ramping up. He also punctured the "circular financing" scare (Nvidia investing in customers who then buy Nvidia chips): it's real but small, and driven by competition with Google's TPU chip rather than a need to prop up demand. On valuation, he noted Nvidia trades around 40 times earnings versus Cisco's 150 to 180 times at the 2000 peak. His caution was aimed elsewhere: AI models will be structurally lower-margin than the 80 to 90% gross margins of classic software, because they're compute-hungry by nature. (The a16z Show, "Is AI a Bubble? Gavin Baker on Data Centers, GPUs, and the AI Economy," July 14, 2026)
Is SaaS dead? A public-company founder's view from inside the wreckage. Avishai Abrahami, CEO of Wix, gave what 20VC called his most candid interview in years. Wix trades at about a $2.8 billion market cap on $2.1 billion of revenue, and owns Base44, a fast-growing "vibe coding" product (build software by describing it in plain English) now well past $150 million in annual recurring revenue. He argues the market is pricing the core business below zero: "Because if you look at peers doing vibe coding, the multiple is much higher... on its own, it's probably at $8 billion." His defense of software's moat is about trust and complexity, not features. On Salesforce: "the biggest value Salesforce provides... is the fact that you trust SaaS providers with your data. What other platform will JP Morgan trust for their customers' data? None." And his memorable line on limits: "You're not gonna vibe-code Shopify, no matter how good you are, the business stack is too hard." Wix actually tested this: they put professional developers on rebuilding just the booking logic for a hairdresser inside Base44; after a week they'd managed a little, and a stronger team still hadn't cracked it two weeks later. His frustration with the market is palpable: "Today we're trading on other companies' news. We're not trading on Wix news. We're trading mostly on what OpenAI or Anthropic or Google are saying... I really don't care." (20VC, "Wix's Founder on What Wall St Gets Wrong About AI | Will Base44 Win the Vibe Coding Wars | The Buyback Disaster with Avishai Abrahami," July 13, 2026)
The exit market is genuinely tough, and everyone knows it. On the Glean episode, Stebbings summed up the landscape: "If you don't have a billion in revenue today, it's hard to go public. Tech acquirers are very specific about what they want to buy. PE is licking its wounds from having a portfolio that's full of Medallias." Arvind Jain pushed back on the doom, arguing that over his 25 years "it's easier to build a startup and get a good exit from it these days than it used to be." Both things can be true: giant outcomes are more achievable than ever, but the middle, the solid $200-million-revenue company hoping to go public or get bought at a fair price, is being squeezed.
Enterprises are asking "where's the ROI?" Two shows picked up on Alex Karp of Palantir going viral on CNBC. His two claims: there's never been more skepticism from large enterprises toward frontier-model providers like OpenAI and Anthropic, and real doubt inside those companies about whether AI spending is paying off. O'Driscoll, who watched the whole clip, said that beneath the theatrics "the points were spot on. Corporate America is saying, I'm spending all this money, am I getting anything?" plus a fear that handing their data to a model provider means teaching a system that will later sell their playbook to competitors. Notably, Palantir's stock rose 9% that day. Arvind Jain, who sells AI to big enterprises, largely agreed there's real "operational dependence" anxiety: "In the future, all of that institutional learning is going to accumulate in that agent... if you don't own the learning, you're basically fully dependent on these AI companies." (20VC, "Sam Altman Offers Trump 5% of OpenAI... Alex Karp Sounds the Alarm... Deepseek Building Own Chips," July 9, 2026; and 20VC, "Why OpenAI and Anthropic Won't Win the App Layer... with Arvind Jain," July 11, 2026)
AI isn't getting cheaper, yet. A quietly important data point from Arvind Jain, who sits on the buyer's side: over the last six to nine months, "every model actually increased their per-token price," even though everyone had assumed prices would keep falling. He also revealed the eye-watering cost of running agents at scale, one Glean engineering "triage" agent that handles 95% of production issues costs "a million dollars a month," and he's not sure it beats humans yet. His prediction is still that costs fall by orders of magnitude eventually, but the near-term reality (labs raising prices to look like real businesses before going public) cuts against the "software gets free" story. Wix's Abrahami echoed the nuance: building their own model only saves "between 1% to 30%" versus frontier models on complex tasks, not the 14 to 16x that Chamath Palihapitiya has cited for open source (which, Abrahami noted, applies to simpler tasks).
The buyback that aged like milk. Abrahami was refreshingly honest about Wix's big share buyback: "Obviously we had terrible timing." With about $1.5 billion sitting idle and no big acquisition planned, they bought back stock at what looked like a low, and it kept falling. He remains a believer in buybacks as "another way of giving a dividend," but also flagged unease about the industry normalizing "insane levels" of stock-based compensation "a la Snapchat." A useful reminder that even disciplined operators can't time their own stock. He also confirmed the wild economics of the AI M&A market: Wix bought Base44 as a one-person company for $80 million, and it's roughly doubled in revenue since, and, tellingly, his board never even questioned paying that much for one person.
Teams get bigger, not smaller. Cutting against the fashionable "AI means tiny teams" narrative, Arvind Jain argued the opposite: per-person productivity will soar, but so will expectations. "To make the same amount of revenue, you have to produce a 10X better product in the future." At Glean, AI usage already follows a power law, some employees burn $10,000 to $15,000 a month in tokens while others spend $20, and genuinely advanced use is still limited to about 5% of staff. Wix's Abrahami was skeptical too: "We all give too much credit for AI and what it can do. And then when you start looking at the details, it's not really doing that well for a lot of the things." His customer-support team (Wix's largest) is still built on humans plus their own homegrown AI, because, in his experience, off-the-shelf AI support agents simply "don't work, we tried, it didn't work, we tried again, it didn't work."
Quote of the Week
"Microsoft owns 30% of OpenAI. If ownership stake resulted in besties, they'd be besties. They're not besties. They're in a stale marriage looking for a divorce, but can't quite pay the tax."
Rory O'Driscoll, Scale Venture Partners, on why Sam Altman handing the US government 5% of OpenAI won't buy the alignment he's hoping for (20VC, July 9, 2026)