💥When to Sell, Next Wave in AI… | Data Driven VC Summary

1. Key Themes

Theme 1: Agent infrastructure is a new, distinct wave of AI infrastructure, and durability is the core engineering problem

Agents are the third infrastructure workload

Balderton frames agents as following model training and chatbots, each demanding different infrastructure.

"Model training needed GPU clusters and training data, and chatbots moved the load to inference and live web search. Agents are the third workload, running for hours, hundreds or thousands at once, and acting inside company systems."

Agents need a purpose-built, six-layer stack

Earlier infrastructure was designed for single-request chatbots, so agents need their own layers.

"Balderton maps agent infra into six layers: where agents run (including sandboxes for code), context, tool use, orchestration of long tasks, self-improvement, and security and observability."

Adoption does not guarantee revenue; four forces decide which layers win

The team identifies what caps the commercial value of infra layers.

"The team names four forces that decide which layers become large businesses: (1) customers building parts in-house, (2) AI labs bundling what is on their roadmap, (3) data platforms selling automation on top of customer data and (4) clouds moving up the stack."

Value migrates toward execution layers

"Balderton sees value move from compute toward the layers closest to the work, so after inference the next winners sit in agent execution, such as sandboxes and keeping long tasks running. For investors, the best bets are layers that none of the four forces can easily absorb."

Durable state is the practical engineering requirement

Avi Chawla's walkthrough shows why production agents must survive restarts and duplicate actions.

"A production agent must outlast the program that started it. An agent waiting hours for a human sign-off may be shut down, and simply rerunning the request can change its evidence and draft, or publish twice."

"Chawla tags each action with a unique key, so a repeated approval simply returns the first result."

Theme 2: When intelligence gets cheap, accountability and ownership capture the value

Accountability for outcomes is the scarce asset

Long Lake's Prathik Naidu argues the expensive part of AI-era services is owning results.

"As intelligence gets cheaper, Naidu argues the costly part is being accountable for results. Long Lake targets services firms that clients trust end to end, where the work needs human judgment and regulation sets who is responsible."

AI vendors leave most of the gain with the client

"Even then, a $1M-a-year contract in Naidu's example adds $5M of yearly EBITDA for the client, worth $50M at a 10x multiple, while the vendor must keep winning renewals."

Buying beats building for AI-native services

"Building an AI-native firm from scratch means starting without customers, brand or expertise, so Long Lake buys established businesses."

The upside is realized only when freed capacity is redeployed

"Its AI rollouts cost first, then free up staff time, which Naidu says reaches full value only once it is put to work: higher revenue per employee and better service."

Theme 3: Behavioral bias and incentive misalignment erode VC returns

Escalation of commitment distorts follow-on decisions

Stanford's Ilya Strebulaev shows the partner closest to the company is the most biased.

"The lead partner usually sits on the board and knows the company best. Strebulaev argues that closeness, plus the wish to prove the first call right, makes them too upbeat to judge the next round."

Sunk-cost bias shows up even in controlled experiments

"In his executive education test, a project has absorbed $500M with no profit. Often 30% or more choose to add $100M for a 5% chance at $1B, a bet worth far less than its cost."

Structural fixes exist and correlate with returns

"Almost nine in ten firms use the same rules for follow-ons as for new deals. The piece suggests a 'red team' that excludes the deal lead, and new investors leading rounds, which his research links to higher fund returns."

Early selling after IPO is a credibility problem

Dan Gray's data shows VCs selling into strength while criticizing public-market short-termism.

"Gray's point is credibility: VCs who sell into hot markets while calling public markets short-sighted make IPO buyers warier, which he suggests may be why going public now feels harder."

Theme 4: Fundraising is judged on trend and comparability, since investors lack shared benchmarks

Metrics fluency is table stakes

"Harris calls not knowing your metrics disqualifying. Founders should avoid flattering tricks such as cumulative charts or GMV reported as revenue, know how each number is calculated, and when stuck, promise an answer the same day."

Investors benchmark you against their pipeline, and AI has raised the bar

"With no shared benchmarks, investors judge you against the other companies in their pipeline, and AI's revenue speed has raised the bar."

A trend needs enough clean data

"Six months of 40% monthly growth reads as an early trend. Harris flags three weak spots: (1) too little data, (2) months that swing between growth and decline, which need an honest structural explanation, and (3) a long flat history before a recent spike."

2. Contrarian Perspectives

VCs' own post-IPO selling may be damaging the IPO market

The conventional view blames macro conditions or public-market short-termism for a weak IPO window. Gray points the finger at VCs themselves.

Evidence: "Of the 50 largest US VC-backed IPOs since 2014, only 8 have outperformed the Nasdaq-100 (QQQ) since listing, against 14 of 50 for 1994-2006 and 22 of 50 for 2004-2016."
Evidence: "Across 386 venture stakes, investors in 2014 IPOs made their first sale a median 14.7 months after listing. Since 2017, first sales have come at 5.9 to 6.9 months."
Evidence: "In Gray's sample, they rose from 0% of VC-backed IPOs in 2016 to 83% in 2021 and still appear in about 80%."
Implication: "GPs who want a healthier IPO window may have to show conviction by holding after listing."

Cheaper AI means VCs should buy services businesses rather than fund AI vendors

The consensus view is that AI vendors capture the value of AI. Naidu argues the economics favor owners of the operating business.

Evidence: A $1M contract produces "$5M of yearly EBITDA for the client, worth $50M at a 10x multiple, while the vendor must keep winning renewals."
Implication: "If this framework is right, the biggest AI returns in services will go to whoever owns the business."

More conviction and closeness can make an investor worse, not better, on follow-ons

Conventional wisdom says the partner who knows the company best should drive follow-on decisions. Strebulaev argues the reverse.

Evidence: Closeness plus "the wish to prove the first call right" makes the lead "too upbeat to judge the next round."
Evidence: "Almost nine in ten firms use the same rules for follow-ons as for new deals," and "new investors leading rounds" is linked "to higher fund returns."
Implication for founders: "Treat an insider keen to lead their next round alone as a signal worth questioning."

3. Companies Identified

Long Lake

Description: Firm that buys established services businesses and rolls out AI inside them.
Why mentioned: Case study for the thesis that ownership captures AI value.
Quotes: "Long Lake targets services firms that clients trust end to end, where the work needs human judgment and regulation sets who is responsible." / "Long Lake buys established businesses."

Balderton

Description: VC firm publishing a framework on agent infrastructure; co-hosting the Oct 28 'The Agentic Fund' webinar.
Why mentioned: Source of the "third wave" agent infrastructure thesis.
Quotes: "Niklas Falk and Tobias Winczer at Balderton argue that long-running agents have created a third wave of AI infrastructure."

Odin

Description: Research and analytics outfit where Dan Gray tracks VC exit behavior.
Why mentioned: Source of the post-IPO selling data.
Quotes: "Dan Gray at Odin tracks how fast VCs sell after IPO."

Stanford GSB

Description: Business school where Ilya Strebulaev researches VC decision-making.
Why mentioned: Source of the follow-on bias research.
Quotes: "Ilya Strebulaev at Stanford GSB explains how escalation of commitment... distorts follow-on decisions inside VC partnerships."

Magid

Description: Company founded by Aaron Harris.
Why mentioned: Harris's affiliation as he explains fundraising metrics.
Quotes: "Aaron Harris, founder of Magid and former YC partner, explains how to use metrics in a fundraise."

Y Combinator (YC)

Description: Startup accelerator.
Why mentioned: Harris's prior role lends credibility on fundraising advice.
Quotes: "Aaron Harris, founder of Magid and former YC partner."

Daily Dose of Data Science

Description: Educational data science publication co-founded by Avi Chawla.
Why mentioned: Chawla's affiliation as he demonstrates agent durability.
Quotes: "Avi Chawla, co-founder of Daily Dose of Data Science, builds an AI agent that pauses for human approval and resumes hours later."

Google

Description: Provider of the agent tools used in Chawla's build.
Why mentioned: Tooling for the durable-state agent demo.
Quotes: "...resumes hours later, using Google's agent tools."

Vestberry

Description: Portfolio management platform for data-driven VCs (sponsor).
Why mentioned: Sponsored case study with Giant Ventures.
Quotes: "Quarterly valuations and LP reports became much faster. Audits got simpler too, now that the team can pull the portfolio position for any past date."

Giant Ventures

Description: VC firm and Vestberry customer.
Why mentioned: Case study of replacing Excel-based valuation workflows.
Quotes: "Valuations were updated by hand in Excel, and the information behind them was scattered across files."

Harmonic

Description: Co-host of the Oct 28 webinar and the Nov 18 Helsinki breakfast.
Why mentioned: Event partner.
Quotes: "'The Agentic Fund' webinar on Oct 28 with Balderton & Harmonic."

Goodwin

Description: Co-host of the Slush breakfast in Helsinki.
Why mentioned: Event partner.
Quotes: "DDVC x Slush Breakfast in Helsinki Nov 18 with Goodwin & Harmonic."

4. People Identified

Dan Gray

Description: Researcher at Odin.
Why mentioned: Tracks VC post-IPO selling behavior and argues it harms the IPO market.
Quotes: "Dan Gray at Odin tracks how fast VCs sell after IPO and argues that early-release clauses let them cash out while the market is still hot."

Prathik Naidu

Description: Founding partner at Long Lake.
Why mentioned: Argues that owning services businesses is the way to capture AI value.
Quotes: "Prathik Naidu, founding partner at Long Lake, argues that buying services businesses outright is the path to putting AI to work."

Ilya Strebulaev

Description: Professor at Stanford GSB.
Why mentioned: Research on escalation of commitment in VC follow-ons.
Quotes: "Ilya Strebulaev at Stanford GSB explains how escalation of commitment, the urge to keep investing in a failing project, distorts follow-on decisions."

Niklas Falk and Tobias Winczer

Description: Investors at Balderton.
Why mentioned: Authors of the agent infrastructure "third wave" framework.
Quotes: "Niklas Falk and Tobias Winczer at Balderton argue that long-running agents have created a third wave of AI infrastructure."

Aaron Harris

Description: Founder of Magid, former YC partner.
Why mentioned: Guidance on using metrics in a fundraise.
Quotes: "Aaron Harris, founder of Magid and former YC partner, explains how to use metrics in a fundraise when investors share no common benchmarks."

Avi Chawla

Description: Co-founder of Daily Dose of Data Science.
Why mentioned: Demonstrates durable state and idempotency for long-running agents.
Quotes: "Avi Chawla... builds an AI agent that pauses for human approval and resumes hours later."

Andre Retterath

Description: Author of Data Driven VC.
Why mentioned: Newsletter author, running a live experiment automating VC work with AI.
Quotes: "Join 2,023+ investors in our free Slack group as we automate our VC job end-to-end with AI. Live experiment. Full transparency."

5. Operating Insights

Audit your follow-on process for lead-partner bias

Review past follow-ons and structurally limit the deal lead's influence on the next round.

"GPs should look back at past follow-ons and limit the deal lead's say in the next one, since the partner closest to the company is the most prone to overcommit."

Lead your fundraise with the metric that kills the biggest doubt

Pick the single number that addresses your biggest perceived risk; investors can use the same test in diligence.

"Founders should lead with the metric that removes the biggest doubt about the business, such as enterprise sales for a team that has never sold to enterprises. Investors can apply the same test in diligence: find the biggest risk, then check how the metric addressing it is calculated."

Make agents restart-safe and idempotent before shipping

Persist the full state of the paused task and guard side-effecting actions with unique keys.

"The fix is to store the paused task, the documents it found, its draft and the pending approval. In Chawla's test, a new program picked the task up later, published from the original draft and ran the search only once."

"A resumed agent can repeat a step, which is harmless for a search but costly for a payment or email."

6. Overlooked Insights

Consider the choice of who leads your next round as a signal

For founders, the Stanford research implies a negotiating and signaling tactic that most would overlook: cultivating outside leads is not just about price, it is about decision quality.

"Founders should treat an insider keen to lead their next round alone as a signal worth questioning, and keep courting outside investors who could lead it."

Pre-seed through growth investors may be judged against AI-speed comparables, not their own sector

Harris's tip to ask investors what they passed on or missed is a cheap way to calibrate against a moving bar.

"Harris advises timing the raise carefully and asking investors which companies they passed on or missed."