We’re back with another deep dive into the shifting landscape of venture capital, and this week we’re joined by a legend who’s seen it all—Roger Ehrenberg. Roger founded IA Ventures in 2008, building one of the most respected early-stage portfolios in the industry with investments in companies like The Trade Desk, Wise, DataDog, and Digital Ocean. After a successful run, he’s now back in the game with a new fund, bringing 25+ years of institutional investing experience (including stints at Deutsche Bank and CSFB) to bear on today’s radically different market. Joining him are Rory O’Driscoll from Scale Venture Partners, Jason Lemkin from SaaStr, and Harry Stebbings from 20VC for a no-holds-barred conversation about what’s really changing in venture—and what investors are getting dangerously wrong.
Bottom Line Up Front: The venture capital playbook is being rewritten in real-time. Three seismic shifts are reshaping how investors think about returns: (1) AI companies are compressing the time from zero to billion-dollar valuations into months, not years, forcing investors to rethink traditional portfolio diversification strategies; (2) the concentration vs. diversification debate has become stage-dependent, with early-stage investors needing 20-25 initial bets before concentrating 75% of capital into 3-5 winners; and (3) the old social contract between founders and investors is breaking down when unprecedented wealth—like $3.5 billion offers—turns multi-period games into single-turn decisions. Meanwhile, the exit bar has doubled from $200M to $400M in ARR, extending the journey to liquidity by 2-3 years and fundamentally changing how investors should think about follow-on capital allocation.
When $665M Actually Makes Sense: Industry Ventures Goes to Goldman
Goldman Sachs’ acquisition of Industry Ventures for $665 million (with potential earnouts reaching $970M) provided a masterclass in asset management valuations. The deal traded at roughly 10% of Industry Ventures’ $7 billion in assets under management—a valuation that reflects the economics of secondary markets and fund-of-funds structures.
“If you manage 100 billion, the entity is worth 20 billion for someone like Carlyle or KKR where they own all the economics,” explained Rory O’Driscoll of Scale Venture Partners. “Here, which makes sense as a secondary because the economics aren’t typically quite as good as primary investors, it’s at 10%.”
The transaction highlighted a critical distinction in venture: not all GP businesses can be sold. “You couldn’t do that with a venture firm, especially a small venture firm, because in the end, all you have is the three people, and if you cash them out 100%, then you don’t have anything,” O’Driscoll noted.
For Goldman, the strategic rationale is clear: access to high-fee private assets at a time when public market management fees have compressed to single-digit basis points. The firm’s Apex platform for ultra-high-net-worth clients provides immediate distribution for Industry Ventures’ products.
Roger Ehrenberg, founder of IA Ventures and now back in the market with a new fund, emphasized the business model distinction: “Hans built an asset management business. The rest of us here do not run asset management businesses.”
The $3.5 Billion Question: When Founders Become Free Agents
Andrew Tullik’s departure from Thinking Machines—a company he co-founded and helped raise $2 billion for at a $10 billion post-money valuation—to join Meta for a reported $3.5 billion sent shockwaves through the venture community. Not because founders leaving isn’t common, but because of what it signals about the new rules of engagement.
“In the face of unprecedented wealth, I’m shocked to discover that most people behave badly,” said Jason Lemkin, founder of SaaStr. “Something’s broken in the way that we’re evolving as humans if everything ultimately reduces to what’s in it for me.”
But O’Driscoll offered a more clinical perspective: “Once you’re not playing a multi-period game, and when someone’s offering you $3.5 billion, you’re no longer playing a multi-period game. You’re playing a one-and-done. You’re going to get bad human behavior.”
The financial calculus is stark. “Would you prefer $2 billion in Thinking Machines unlisted stock with a chance to be amazing or a chance to go bust, or $3.5 billion of liquid Facebook stock over the next 5 years?” O’Driscoll asked. The answer, as Ehrenberg acknowledged, was obvious.
This raises profound questions about how investors should adapt. Extended founder vesting, cliff vesting, and repurchase rights are becoming standard protective measures. But the deeper issue remains: when the core asset of these AI investments is “a group of seven engineers,” as O’Driscoll noted, the traditional founder-investor relationship may simply not apply.
“The unusual thing as a career trajectory for an engineer—spending 14 years at Meta, bouncing to OpenAI for less than a year, then to Thinking Machines for less than a year, then back to Meta—that’s not crazy,” O’Driscoll observed. “What is unusual is because of the technical nature of these bets, how much reliance we’re putting on the behavior of an engineering/academic talent pool, which probably responds fairly differently than the ‘I’m a founder, I want to be the CEO’ talent pool.”
Masa’s $5B Margin Loan: Low-Octane Fire Management
SoftBank’s Masayoshi Son securing a $5 billion margin loan against ARM shares to invest more in OpenAI barely registered as news to veteran observers. “Masa rules,” said Ehrenberg simply. “This is Masa being Masa 100%.”
The move represents smart use of leverage—Son still owns 90% of ARM’s roughly $90 billion market cap, giving him $80 billion in equity value to borrow against. A $5 billion margin loan at reasonable rates is “a relatively low octane Masa move,” as Ehrenberg characterized it, noting Son could easily lever up to $25 billion against the ARM position.
“This man is full risk on all the time and just wants to get the bet on the table,” Ehrenberg continued. “He’s been spectacularly right at times and spectacularly wrong at times, but spectacularly willing to play, which on behalf of the audience we should be eternally grateful.”
The broader context matters: where will the capital come from to fund the massive compute infrastructure AI requires? As Lemkin pointed out, “We’re now building more data centers than office buildings,” with demand for tokens growing exponentially as companies deploy AI agents and applications.
The Token Economy: Why 100x Current Supply Won’t Be Enough
The conversation revealed just how far we are from meeting current AI compute demand, let alone future needs. Lemkin shared firsthand experience: “I’ve vibe coded eight apps in 100 days. We have 12 AI agents working at SaaStr now. What I can tell you is today, just what we’re doing today with 12 agents and eight apps, I could use 100x the tokens.”
O’Driscoll pointed to recent scaling research that takes a remarkably matter-of-fact view of the investment required: “A whole bunch of these folks are like, ‘Yep, this is the task we’ve embarked on ourselves for the next 5 years and it’s going to take around 1% of GDP to build a compute cluster big enough to get the flops to get the outcome we want.'”
The scaling laws have held for six to seven years at high degrees of accuracy, creating confidence that continued investment will yield continued progress. The question isn’t whether the technology will advance—it’s whether the economic returns will justify the capital deployment.
“If it’s slowed down, it won’t be because the technology direction is incorrect, it won’t be because the demand isn’t insatiable,” O’Driscoll argued. “It will be purely and simply that the marginal capital provider says, ‘Oh my god, even though the scaling law is holding, the economic return from that investment isn’t holding.'”
Kingmaking vs. Market Forces: The Polymarket and Kalshi Case Study
When Polymarket raised at a $9 billion valuation and Kalshi immediately followed with a $5 billion round from Andreessen Horowitz and Excel, it sparked renewed debate about whether capital can “kingmake” winners in venture.
The reality, as Ehrenberg explained, is pure regulatory arbitrage: “This is the purest regulatory arbitrage play of all time. You can look at the cumulative market cap of regulated sports betting and look at how it has dropped in response to the rise of Polymarket and Kalshi, who are not subject to the same rules and regulations.”
But O’Driscoll pushed back on the kingmaking narrative: “I don’t think anyone betting on Polly or Kalshi gives a damn how much money they have, provided they can pay their bet, and gives a damn who that money came from.”
The distinction matters: kingmaking requires either (1) giving one company so much capital they can overwhelm competitors, or (2) brand-name VCs creating customer preference through their backing. Neither applies to prediction markets, where user loyalty is driven by liquidity and payout reliability, not VC pedigree.
“What’s going on here is two non-sports betting companies who are doing ‘prediction markets’ where all we talk about is the 10% of the revenue that’s political and 90% of their business is sports betting, but we’re not calling it that,” O’Driscoll observed. “They’re just killing it because we all love to sports bet.”
The New Portfolio Mathematics: From Diversification to Concentration
The discussion revealed sophisticated thinking about how portfolio construction must evolve as the venture landscape changes. Ehrenberg laid out his approach: 20-25 initial portfolio companies, then concentrating 75% of total capital into 3-5 winners through aggressive follow-on investments.
“Where I’ve tended to get very concentrated is on the second and third checks, where we’ve gotten deep conviction in a team, their execution, in the market,” Ehrenberg explained. “Three to five companies out of the 20 to 25 constituting 75% of the capital deployed.”
The strategy requires treating each follow-on check as an independent decision. When Trade Desk raised its Series A at $280 post-money, Ehrenberg wrote a $3 million check out of a $50 million fund—after having already invested over $2 million across four earlier rounds. That single check turned into $40 million.
“We look at every check independent of the check prior, period,” Ehrenberg emphasized. “If I look at that and say the information I’ve got about this being a $300 million or $3 billion—I think this could be a $100 billion company—then I will write a very meaningful check into that company, up to about 10% of the fund.”
O’Driscoll pointed out the temporal dynamics: “You have to start off with a significant element of diversification and then concentrate down. We’ve moved from a world where an exit is $200 million in ARR to a world where an exit is $400 million in ARR at an IPO. The finish line has receded another two or three years, which means logically you’ve got more risk and more upside.”
The Founders Fund Playbook: When Concentration Makes Sense
Founders Fund’s evolution from 31 investments in Fund I to targeting just 10 in Fund III sparked debate about optimal concentration levels. The key insight: concentration strategy depends entirely on stage and conviction.
“The more certain you are that you can call the shots, the more focused you should be,” O’Driscoll noted. “Founders Fund both has the evidence that they can call the shots because they’ve done so, and frankly the confidence to call the shots because they got it right.”
But even for growth-stage funds, pure concentration has limits. O’Driscoll was surprised Founders Fund I had 31 investments: “It just didn’t feel in sync with what we’ve seen from these guys in general. I mean, if you look at their SpaceX non-diversification, these guys strike me as the most likely to be most concentrated.”
The distinction between growth and early-stage investing matters profoundly. “When you’re still at the ‘will this thing even work’ stage or ‘will it scale,’ you probably need some significant diversification and then concentrate when you’re effectively investing in what should be public companies but are just private,” O’Driscoll argued. “Then the growth fund strategy should be 10 or 11 or 12 deal concentration.”
Lemkin offered a contrarian view from his concentrated seed approach: “I’m doing 8% of my fund into almost every deal. I have to have such a high hit rate at seed. Half of them have to work. So it’s a stupid model because you have to turn away Clubhouse and Hop In and maybe even DataDog, but you got to just find the Wises and go all in.”
The Early-Stage Illusion: Why You Can’t Pick Winners at Day One
Harry Stebbings challenged the group with a critical observation from his Fund I experience: “When I look back over six years, the best performers—your Linear of the world—were not obvious early, and the early outperformers did not signify enterprise value in the long term. Clubhouse, Hopin, BeReal. If you think you can pick your winners early, I think you are wrong.”
Ehrenberg pushed back with nuance, citing dramatically different paths to success across his portfolio. Trade Desk experienced “multiple near-death experiences, multiple exit opportunities, bridging multiple times, didn’t have a product in market for more than a year and a half.” Wise, by contrast, was “as close to an up-and-to-the-right company as I’ve ever been involved with.”
The key is recognizing you’re playing “a multi-turn game,” as Ehrenberg emphasized. His first check into Wise was $750K at $5.5M post. “Then AR came in and we piled in with AR to $20M, then we piled in with AR to $160M, and we just kept going. We were 17% of TTD at IPO and we were 13% of Wise at IPO out of a little shitty seed fund.”
O’Driscoll validated the uncertainty while offering a path forward: “You don’t always know, but you know more than someone coming in from the outside. So at the margin, you can tilt it your way. And that’s all you can do.”
The group agreed that once companies have 12-24 months of revenue data, conviction increases dramatically. O’Driscoll noted Scale’s analysis showed that “if you get the first two years that we underwrote in terms of revenue from the moment of our investment, your probability of getting greater than a 5x goes from 30% to mid-70s.”
The Follow-On Dilemma: When $300M Post Becomes the New Normal
Lemkin raised the practical challenge facing seed investors today: “What are you going to do when you got in at $8 million post, but the next round’s at $300M or $500M because the AI kids come in? How much of your fund can you really put in? Is it even worth writing that second check?”
Ehrenberg’s answer was unequivocal: “We are hyper-disciplined. We look at every check independent of the check prior, period. If I look at that and say I think this could be a $100 billion company, then I will write a very meaningful check into that company, up to about 10% of the fund.”
He cited the Trade Desk Series A as example: after writing four small checks totaling just over $2M, the company raised at $280M post. “We wrote a $3 million check out of a $50 million fund into that at a $280 post. That $3 million ended up turning into $40 million, and that was a great investment.”
O’Driscoll reframed the concern: “What you’re saying is if all the following rounds are priced incorrectly relative to ultimate exit value, does your strategy work? It’s a fair question. But if the worst thing that happens is my initial check gets marked up and on the follow-on I don’t need to chase the money, then I’m money good on the check I’ve written.”
Ehrenberg summarized it perfectly: “It’s a cash-on-cash business. If the rounds trade up to the point where the follow-on round doesn’t have the return, so be it. The worst thing that happens is my initial check is money good, and provided every check you write is money good, in the end you’ll die rich.”
The Structural Solution: Cross-Fund Investing
One tactical approach emerged as particularly valuable: maintaining parallel LP bases across funds to enable cross-fund investing without conflicts. Ehrenberg explained how this transformed IA Ventures’ flexibility: “All of a sudden we weren’t investing out of a $100 million fund. We were investing out of a $260 million fund.”
The strategy requires careful LP base construction from the start. Ehrenberg recalled an early situation where Fund I and Fund II didn’t have matching LPs: “The LPAC was like, ‘Yes, if you want to do it, go do it. Just understand, if this doesn’t work, you got some explaining to do.’ The risk of that check was not simply financial, it was reputational. Ultimately, we ended up deciding not to write that check.”
With parallel LPs across later funds, the constraint disappeared, enabling aggressive follow-on capital deployment into breakout companies without creating cross-fund conflicts.
O’Driscoll validated the approach while noting the limitation: “Any cross-fund you do is going to be a good deal, otherwise you’re an idiot. So I think good deals you can find follow-on checks where you can cross-fund, even if the LPs aren’t fully aligned between funds, provided you run a process.”
The Peter Thiel Principle: When Outsiders Validate at High Prices
The discussion returned repeatedly to one of Peter Thiel’s key lessons about follow-on investing: when a big, reputable outside investor does a follow-on round at what feels like a high price, do everything you can to get into it.
“It’s not always true, and I could cite examples where it’s not true,” O’Driscoll acknowledged. “But risk-adjusted, information-adjusted, that next round in the deal that you’ve been in that’s performing well, provided the follow-on price is contemplatable, the advice would be: adjust your scales upward, or don’t be guilty of anchoring on what you did.”
The key is recognizing the information asymmetry has shifted. External investors seeing enough signal to lead at elevated prices possess different data than those who invested earlier. As O’Driscoll put it: “You got to find a way to take into account the information since then, both operational and the outside round.”
This connects directly to Ehrenberg’s framework of treating each check as an independent decision based on current information, not past pricing. “Everything in life you can price as an option,” Ehrenberg noted. “I walk through life. Everything looks like the Greeks. Everything looks like options theory, because that’s life.”
Key Takeaways
On Portfolio Construction:
- Early-stage investors should start with 20-25 positions, then concentrate 75% of capital into 3-5 winners through follow-on rounds
- Treat each follow-on check as an independent investment decision based on current information, not past pricing
- The exit bar has moved from $200M to $400M in ARR, extending the journey by 2-3 years and requiring more initial diversification
- Growth-stage concentration (10-12 companies) only works when investing in essentially public companies that are still private
On Follow-On Capital:
- Once companies show 12-24 months of revenue matching initial underwriting, success probability increases from 30% to mid-70s
- Be willing to deploy up to 10% of fund size into breakout companies at subsequent rounds, even at seemingly high valuations
- Build parallel LP bases across funds to enable cross-fund investing without creating conflicts
- When reputable outside investors lead rounds at elevated prices in your portfolio companies, that’s a strong buy signal
On Founder Retention:
- Extended vesting, cliff vesting, and repurchase rights are becoming necessary protective measures
- When offers reach multiple billions, founders shift from multi-period to single-turn games—expect rational economic behavior
- The “emotional commitment” model breaks down when core assets are engineering teams rather than founder-CEOs
- Academic/engineering talent responds to different incentives than traditional founder/CEO profiles
On Market Dynamics:
- AI has compressed the timeline from zero to billion-dollar valuations to months, not years
- Current token demand could absorb 100x more compute capacity—efficiency gains will be consumed by building more, not resting
- Regulatory arbitrage (prediction markets) represents pure value transfer rather than kingmaking through capital
- The “sweet spot” where value is clear but not obvious has compressed to weeks or months
On Capital Deployment:
- Asset management businesses (fund-of-funds, secondaries) can be sold for 10% of AUM; primary venture firms cannot
- Masayoshi Son can still lever ARM’s $80B equity value significantly—expect more aggressive capital deployment
- Goldman’s Apex platform for ultra-high-net-worth clients provides distribution for Industry Ventures products
- Regulated markets with legal/IP complexity offer better protection against rapid AI-driven disruption
Quotable Moments
Roger Ehrenberg on follow-on discipline:
“We look at every check independent of the check prior, period. If I look at that and say I think this could be a $100 billion company, then I will write a very meaningful check into that company, up to about 10% of the fund.”
Rory O’Driscoll on founder departures:
“Once you’re not playing a multi-period game, and when someone’s offering you $3.5 billion, you’re no longer playing a multi-period game. You’re playing a one-and-done. You’re going to get bad human behavior.”
Harry Stebbings on portfolio strategy shifts:
“I’ve realized even at our stage, you got to be trying to do that more and more because the journey from our stage is still 10 years now, whereas before it was six or seven.”
Jason Lemkin on token demand:
“I’ve vibe coded eight apps in 100 days. We have 12 AI agents working at SaaStr now. What I can tell you is today, just what we’re doing today with 12 agents and eight apps, I could use 100x the tokens.”
Harry Stebbings on early-stage prediction:
“When I look back over six years, the best performers—your Linear of the world—were not obvious early, and the early outperformers did not signify enterprise value in the long term. If you think you can pick your winners early, I think you are wrong.”
Roger Ehrenberg on investment outcomes:
“If the worst thing that happens is my initial check gets marked up and on the follow-on I don’t need to chase the money, then I’m money good on the check I’ve written. Provided every check you write is money good, in the end you’ll die rich.”
Rory O’Driscoll on early-stage uncertainty:
“You don’t always know, but you know more than someone coming in from the outside. So at the margin, you can tilt it your way. And that’s all you can do.”
Jason Lemkin on seed concentration:
“I’m doing 5%-8% of my fund into almost every deal. I have to have such a high hit rate at seed for the math to work. So it’s a “stupid” model because you have to turn away stuff with no traction, and maybe even a Datadog, but you got to just find the ones in a tight box and go all in.”
