Summary

Jason Lemkin’s central call is that AI is opening a “golden age of software” while destroying the old assumption that scaled revenue is durable. If software expands from 2% to 4% of GDP, he sees 100 or more decacorns. Harry’s framing was that every VC has five or ten unicorns that are no longer unicorns, now growing in the single digits or teens; Lemkin expects late-stage slowdowns to become commonplace by 2026 and those stranded assets to become less attractive to acquire.
Hot AI rounds can look cheap at roughly 20x forward ARR only because investors extrapolate extraordinary slopes while remaining uncertain whether the “R” is recurring and durable. Lemkin says markups still grade VCs and some LPs, while Harry notes that entering at $200M ARR can shorten duration; the counterweights are dilution that Lemkin estimates might approach 10% annually and a venture cash cycle stretching toward 20 years.
“Triple-triple, double-double” remains elite operating performance but is currently “silver in a gold rush” to perhaps 80% of B2B investors. Lemkin’s prescription is a year-long, relationship-led process with roughly 50 investors and honest monthly updates: “It only takes one.” Founders without Cursor-like heat should avoid artificial Friday deadlines and may be better off accepting a valuation 30% lower if that gets the round closed.
AI demand is producing real revenue faster than technical substance, while investors remain “addicted to top-line growth.” Lemkin has seen companies reach $2M in 60 days or eight figures inside a year with humans running prompts behind thin wrappers; yet AI efficiency does not necessarily reduce capital needs, because RevenueCat took a 2x productivity gain and “plowed it all into new hiring.”
Moveworks accepted ServiceNow’s $2.85B offer after years of deep integration and accelerating demand made ServiceNow’s distribution especially compelling. Bhavin Shah said 250 of 350 customers already used ServiceNow, while Moveworks had five million users against ServiceNow’s 150M-plus; he concluded that Moveworks could not match the market’s scale and speed independently. The strongest ROI was not saving an employee three hours, but automating core workflows across systems such as SAP, Workday, Salesforce, Concur and Jira.
Lemkin expects an IPO and M&A “gold rush” within roughly 18 months, but sees little evidence that PE will rescue slow-growth unicorns. He expects large private companies such as Stripe, Figma, Chime and Canva may list, yet says cash-flow-positive assets at $20M, $50M or even $100M are not attracting the “tire kicking” he once saw; PE instead appears to be combining holdings into “Frankensteins” such as SalesLoft with Drift and Gainsight with Skilljar.
CoreWeave getting public was a major entrepreneurial achievement, but its real scorecard begins on days 180, 365 and 450. Andrew Feldman called an IPO “the beginning of adulthood”; Lemkin nevertheless worries that, as he understands the structure, last-round investors can put back almost $2B of stock if shares fail to trade 70% above the IPO price within two years. That deadline could invite shorts and sacrifice long-term decision-making to a fixed date.
Feldman’s warning on fashionable hardware and defense investing is that accumulated experience still matters. A chip tape-out can consume $20M-$30M in non-recurring engineering costs—and a bug means paying again—while defense requires trusted relationships, cleared personnel, specialized facilities and patience with “Bible-sized contracts.” Commercial technology can still reshape warfare, but procurement and cost-plus incentives remain the constraint.

Deep dive

1. Tariff volatility does not break the software gold-rush thesis

Stebbings opened with tariffs ranging from 10% to 49% on imported goods and Apple falling more than 6%. Lemkin’s deliberately modest trading advice: “Go a little bit long,” ignore the noise and “check your stocks in six months.”
The Thoma Bravo thesis that software spending could rise from 2% to 4% of GDP leads Lemkin to “100 decacorns, 100 or more.” This is the real golden age, albeit one containing “a lot of stress.”
His contrast was 2021, when 27 public B2B “moldy oldies” were growing around 70% and GoToMeeting and Webex were reborn. That revenue was real, but the pandemic created a bubble in which investors mistook a temporary environment for permanent durability.
The closing callback was that news itself now decays almost instantly. Stebbings thought tariffs could dominate one week and become stale the next; in a separate gentleman’s bet, Lemkin predicted three executives would leave within 90 days, versus Stebbings’s one.

2. Scale no longer guarantees durable revenue

Lemkin sees portfolios full of supposedly mature unicorns now growing in the single digits or teens. Two recurring causes are failure to become multiproduct and moving too slowly into AI, but he considers the phenomenon “endemic.”
Through 2022, reaching $50M or $100M in B2B revenue with a decent team meant “you were set.” Zoom now growing around 0% is his shorthand for why that assumption was wrong.
Stebbings’s unresolved cohort is more troubling: companies growing in the mid-teens, barely profitable or just profitable, too weak to go public and insufficiently attractive to PE. “I’m not sure what happens to that.”
Lemkin expects the reset to become explicit in 2026: radical late-stage deceleration will be commonplace, joining a scaled company will no longer look safe, and stalled unicorns will become less attractive acquisition targets.

3. Hot AI rounds optimize for markups and shorter duration

On Cursor—recently valued at what Stebbings thought was $9.6B—Lemkin argued that many hot AI financings cluster near 20x revenue. The ambiguity is whether that means 2025 or 2026 revenue and whether the revenue is genuinely recurring and durable.
Those projections can amount to dragging “99% monthly growth” across a spreadsheet. Paradoxically, once growth arrives, the headline AI round may look cheaper on revenue than seed or Series A deals carrying extreme multiples on tiny bases.
The behavioral mechanism matters more: “VCs are still graded on markups,” as are some LPs. When putting $100M into a hot AI company can quickly produce a 5x markup, “hysteria sets into venture.”
Later-stage investing also compresses duration. Lemkin now frames seed liquidity as roughly 15 years to IPO, followed by about three years to distribute—closer to 20 years—and says essentially all his own liquidity came through the exceptional 2021 window.

4. “Triple-triple, double-double” is silver in a gold rush

Lemkin’s warning to founders is that roughly 80% of B2B investors currently will not touch a company merely following triple-triple, double-double. The performance is still top 0.1%; it simply appeals to perhaps 20% of the market.
His metaphor carries the incentive: “No one’s mining silver” when gold is easy to pull from the river. Investors fear waiting 10 or 20 years for a $2B exit when they can instead chase an Anthropic-style markup.
Dilution complicates those paper gains. Stebbings cited an LLM investor whose entry price had risen 12x but whose real return was only 3.8x; Lemkin estimated hot AI companies paying engineers $600,000-$1M might dilute around 10% annually, versus 5%-6% at a typical startup.

5. A loose fundraising process beats manufactured urgency for most founders

For strong companies below AI-growth velocity, Lemkin recommends knowing perhaps 50 investors and sending credible monthly updates for a year. Repeated movement through 6%, 8%, 10% and 12% growth builds conviction and reduces fears that the numbers are fraudulent.
His pushback on “run a tight process” is unusually categorical: it is excellent advice only for the tiny minority with genuine leverage. A seed founder growing 82% at $4M revenue may weaken the pitch by presenting a data room and an invented Friday deadline.
In a recent $500M round, an investor who had followed the company since $30 of monthly revenue declined because one hour was insufficient. Lemkin himself opts out of exploding processes “99% of the time.”
The practical posture is “confident but humble.” Investor appetite changes after one deal, from “hungry to full,” so a founder outside the hottest cohort might rationally accept a valuation 30% lower if the financing actually happens.

6. AI top-line can outrun the product underneath it

Lemkin has inspected AI startups moving from zero to several million dollars in months, or to eight figures inside a year, where humans were still running prompts or producing BI reports behind the scenes for customers unable to do the analysis themselves.
One team reached $2M in 60 days, then admitted during the demo: “There’s really nothing to demo.” They had wrapped ChatGPT and manually produced content for large, unsophisticated customers; Lemkin still called the revenue earned, but the AI definition had been “stretched” to implausibility.
His diligence verdict is blunt: “No.” Investors ignore gross margins and profitability, citing his claim that OpenAI would not become profitable until $127B of revenue; even skeptics rediscover their tolerance when preserving super pro rata is at stake.

7. AI productivity gains are being reinvested into a harsher race

RevenueCat, used by roughly 40% of mobile apps, measured about 2x productivity from tools including Cursor and Codeium. Yet its response was not to freeze a lean, cash-flow-positive team: “We plowed it all into new hiring.”
Lemkin’s uncertainty is whether twice as many people at twice the productivity yields 4x, 6x-8x or 16x output. The clear implication is that a two-person competitor cannot assume efficiency lets it stand still.
At another portfolio company near $20M revenue, a competitive slide had expanded from two known rivals to 11 columns, with the company’s competitor count rising from three to 100. The board split between selling immediately and “rebooting” the company for 2025; cash-flow positivity gave the founders room to choose aggression.

8. Moveworks sold distribution into accelerating enterprise demand

Shah’s framing of the ServiceNow agreement begins with “great companies are not sold, they’re bought.” Moveworks had spent eight years applying AI to workplace transformation and had been an integration partner for seven of those eight years.
The overlap was decisive but not coercive: 250 of Moveworks’ 350 customers also used ServiceNow. Moveworks still interoperated with Jira, Freshservice, Microsoft and other systems, and Shah said that cross-platform reach would continue after the transaction closes.
Demand accelerated sharply in the preceding nine months. Against Moveworks’ 350 customers and five million users, ServiceNow offered access to 8,400 large-enterprise customers and 150M-plus users; Shah concluded that even 100%-200% startup growth could not match the market’s appetite quickly enough.
The technical asset was an agentic employee layer translating ambiguous language into precise APIs through separate reasoners, slot-filling models and a “manifest generator.” That enabled deep workflows across SAP, Workday, SuccessFactors, Salesforce, Concur and Jira—not merely conversational search.

9. Enterprise AI adoption is a people problem on a six-to-nine-month clock

Shah distinguishes “weak ROI”—saving Bob three hours while still paying the same salary—from strong ROI that transforms core business processes. The latter requires integration, security readiness and change management, not just a capable model.
After roughly two years of ChatGPT, large enterprises are finally saying, “We gotta do something.” Yet Shah expects enterprise panels two years from now to discuss “exactly what we’re talking about right now”: adoption curves, implementation and organizational rollout.
Even an approved buyer can need six to nine months to implement, begin with one subgroup and expand. “It’s a people problem”; the technology may be ready while internal processes remain slow.
Moveworks had only recently received FedRAMP authorization, which Shah described as the first for an agentic platform. Building security controls and GovCloud infrastructure illustrates why enterprise readiness takes more time than a product demo suggests.

10. AI has become an incremental CIO budget line

Shah began seeing solid, explicit budget allocations for AI and agentic systems at the start of 2024. Spending was no longer coming only from a slush fund or one-off innovation experiment.
Broadcom was his concrete leverage example: it grew from 10,000 to 50,000 employees after acquiring CA, Symantec and VMware while keeping the same-sized support organization across IT, HR and other functions with Moveworks.
Lemkin’s synthesis was that CIOs were budgeting agentic AI as a line item and that this could represent growth in overall spend, whether funded by slower headcount growth or reduced support labor. Shah’s example emphasized doing more with existing budgets and lowering the cost of growth; whether the budget is truly incremental is the central question. If it is, Lemkin calls it a gold rush; otherwise, vendors are merely “moving chess pieces around the board.”

11. IPO and M&A liquidity may reopen, but PE is not a backstop

Lemkin expects a pro-M&A environment to create a “golden age of M&A,” citing ServiceNow’s largest acquisition in Moveworks and its subsequent purchase of Logic.io. Realized proceeds can also recycle through LPs into new venture funds.
IPOs remain structurally difficult: there are few dedicated buyers, anyone can purchase the next day, and Lemkin says most tech IPOs are flat two years later. Still, scale could lead Stripe, Figma, Chime, Canva and other $500M-plus companies to list within roughly 18 months.
An unnamed late-stage company pursuing an acquisition told its target it intended to IPO in the coming “teens of months.” Lemkin also stressed that private tender offers often address RSUs vesting and becoming taxable for employees before the IPO, not permanent investor liquidity.
His darker observation is that PE is barely kicking tires on cash-flow-positive software assets at $20M, $50M or $100M. SalesLoft’s $2.5B deal and Pipedrive’s $1.4B sale exist, but current sponsors are often merging holdings into “Frankensteins” rather than buying another slow-growth unicorn.

12. Chip startups reward domain experience over founder youth

Asked about a 19-year-old raising $20M pre-seed to challenge NVIDIA, Feldman advised: “If you don’t know a lot about hardware, I wouldn’t invest in hardware.” The field historically rewards experienced investors and entrepreneurs.
Chip design extends far beyond front-end logic: tool licenses can cost millions annually, geometry and foundry relationships matter, and non-recurring engineering can cost $20M-$30M. A bug can force the company to pay the entire amount again; back-end design and timing closure are separate specialties.
Young founders have an edge when they resemble the customer—social products for their friends are Feldman’s canonical example. Chips, databases, enterprise infrastructure and sales automation instead reward accumulated knowledge of the buyer and the system being built.

13. CoreWeave’s IPO begins the execution test, not ends it

Feldman praised CoreWeave’s creative use of GPU-backed debt and access to scarce compute, calling getting public an enormous achievement. The first-day performance was strong, but the real test begins on days 180, 365 and 450.
An IPO is “the beginning of adulthood”: lower-cost capital is exchanged for market discipline, regulation and relentless execution. Employees must return to work despite stock moves that can change an engineer’s net worth by more than his father earned in a lifetime.
Lemkin’s concern is structural. As he understands it, if CoreWeave does not trade 70% above the IPO price within two years, last-round investors can require the company to repurchase almost $2B of stock—capital it may need Microsoft, OpenAI or another source to supply.
Feldman avoided judging sophisticated investors’ work but shared the incentive concern: a two-year put can erase the long term and create a date the company must hit. Buying Weights & Biases was, in his view, a useful move toward a broader software layer; now the message is simply, “Godspeed” and execute.

14. Defense tech’s constraint is procurement, not technical possibility

Feldman described US defense procurement as designed to buy “without trust” from giant primes such as Lockheed Martin. Innovative startups struggle to find the responsible buyer, then encounter “Bible-sized contracts” attempting to specify every possible failure.
Experience and relationships are transformational: founders need a co-founder, sales leader or COO who understands how the military, intelligence agencies and nuclear-security organizations buy. Cleared staff, secure-site support and sometimes cleared manufacturing facilities add another layer.
Cost-plus contracting creates “dead wrong” incentives. Feldman’s analogy is a home contractor earning 17% on every window: there is no reason to negotiate the input cost, while procurement rewards avoiding being burned instead of taking risks that could produce exceptional outcomes.
The counterexample is commercial innovation: Anduril and inexpensive drones used in Ukraine showed that technology outside traditional primes can affect warfare. Yet even the supposedly simplified OTA mechanism failed Feldman’s company when a procurement agent copied the standard contract into it verbatim—evidence that the required change is cultural.

The Future of AI Avatars and DeepFake Technology w/ Joshua Xu & Steve Brown | EP #161 The Creators of Model Context Protocol