An empirical agenda for AI🔬, How…

1. Key Themes

Theme 1: AI is absorbing nearly all venture capital, and the data shows how concentrated it has become

AI dominates US deal value, with capital concentrating in megafunds

The PitchBook data shows both demand-side and supply-side concentration. Capital is flowing overwhelmingly to AI, and fund-raising is skewed to the largest vehicles.

"AI accounted for 82.7% of US venture deal value, while funds raised $108.5B across 699 vehicles and megafunds captured 78.1% of that capital. Active unicorns reached a record 992 with a combined valuation of $5.7T, and 18 companies went public in Q3, including 12 healthcare businesses."

Europe is following the same pattern

AI's share of European funding hit a record, and a single mega-round set a regional record.

"European startups raised $25B in Q3 2026, with AI companies attracting $18.8B, or 75% of the total, the highest share on record... while Mistral's $3.5B round set a record for a Europe-based company."

Public markets reflect the AI build-out, but consumer monetization is still thin

Tech's earnings contribution is lopsided, while end-user payment remains small.

"Technology contributed approximately 76% of S&P 500 earnings growth as of late August 2026, reflecting demand for semiconductors, power, and networking infrastructure tied to AI expansion. A100 GPU prices remained at or above their levels at the start of the year, while roughly 2% of US households paid for an AI service as of April."

Theme 2: AI-native software is hitting unit-economics and pricing problems that must be redesigned

Flat seat-based pricing breaks when inference costs scale with usage

Harvey's reported margins illustrate the risk of pricing AI products like traditional SaaS.

"Harvey's gross margins reportedly reached -50% in June as usage costs rose under flat seat-based pricing, prompting it to adopt a fine-tuned open-weight model that can cost up to 90% less than frontier alternatives."

The fix is consumption pricing, model routing, and honest COGS accounting

"Legora is moving toward consumption-based pricing and model routing, highlighting how AI software companies must align pricing with inference costs and account for AI expenses in COGS."

Incumbent SaaS platforms are also monetizing agent access

Agents hitting APIs at scale creates a new cost line for companies building on top of systems of record.

"New agent-access charges from Salesforce, Atlassian, and HubSpot could make automated API usage expensive, with one estimate putting SaaStr's annual bill at $240,000 for 35,000 to 40,000 daily calls."

Theme 3: Finance and commerce are being re-architected around AI, tokens, and agents as customers

The next financial giant may be AI-first and token-first

"As mobile and cloud become standard financial infrastructure, competitive advantage may shift toward AI models making decisions, tokens maintaining shared records, and agents operating as customers with wallets. Stripe, Revolut, Ramp, Robinhood, and Nubank already offer products across these areas, setting the stage for an AI-first, token-first financial giant."

Payments data is becoming a lens on AI's economic impact

"Stripe Economics will use payments, business formation, and sentiment data to study AI's effects on productivity, labor markets, firm structure, and agent-driven commerce."

Generative AI raises the stakes for fraud, favoring network-level intelligence

"Generative AI makes fake identities, forged documents, and impersonation easier to scale, with global fraud losses projected to reach $40 billion within the next few years. Static identity checks and some biometric defenses are increasingly vulnerable, while behavioral patterns across institutions and platforms offer additional signals for detecting fraud."

Theme 4: Differentiation is migrating from the AI stack to orchestration, domain expertise, and brand

The AI technical stack is converging

"AI-native startups increasingly rely on similar technical foundations, including agent loops, connectors, memory, and sandboxes, making their products and pitches harder to distinguish. Deep domain expertise, customer empathy, and a clear value proposition offer stronger differentiation than another standard AI stack."

Orchestration and agent infrastructure can matter more than the model

"Changing agent infrastructure while keeping the underlying model fixed can improve performance by up to 6x, making orchestration across video, image, voice, and sound models a key production advantage."

Reputation and storytelling are themselves competitive assets

"Thrive Capital cultivates its reputation through high-profile deals, founder endorsements, design leadership, merchandise, and personalized gifts rather than a conventional publishing strategy."

2. Contrarian Perspectives

"Fintech is dead": the winner is an AI/token-native player, not another app layer

The piece argues that mobile and cloud are now commodity infrastructure, so the next source of advantage lies elsewhere. This runs against the idea that fintech is a mature category being incrementally improved.

"As mobile and cloud become standard financial infrastructure, competitive advantage may shift toward AI models making decisions, tokens maintaining shared records, and agents operating as customers with wallets."

In AI booms, hard assets may capture the value, not software

Citing the historical pattern of past technology booms, the argument is that value accrues to physical inputs rather than the application layer.

"Past technology booms, from railroads to radio and the dot-com era, directed capital toward physical inputs such as electricity, metals, land, and food. Brazil's renewable power, quartz, copper, graphite, rare earths, and farmland position the country to supply resources demanded by AI infrastructure."

There is no magic fundraising threshold; "readiness" is about alignment, not metrics

This challenges the common founder heuristic of hitting a revenue or growth number before raising.

"No single revenue threshold or growth rate determines fundraising readiness, particularly as AI accelerates growth and software competitors can emerge within months. The Decisive Moment framework focuses on aligning the business opportunity, the reason for raising, and an investor prepared to recognize the opportunity."

3. Companies Identified

Stripe

Description: Payments infrastructure company with an economics research arm.
Why mentioned: Launching an empirical research agenda on AI's economic effects, and cited as a fintech offering products for the AI/token era.
Quotes: "Stripe Economics will use payments, business formation, and sentiment data to study AI's effects on productivity, labor markets, firm structure, and agent-driven commerce." / "Stripe, Revolut, Ramp, Robinhood, and Nubank already offer products across these areas."

Thrive Capital / Thrive Holdings

Description: Venture firm and its holdings company acquiring services businesses.
Why mentioned: Case study in brand-building and storytelling beyond conventional publishing.
Quotes: "Thrive Holdings takes a different approach by hiring a storyteller to document the accounting and IT services businesses it acquires."

Harvey

Description: Legal AI software company.
Why mentioned: Cautionary case on negative gross margins under flat seat pricing.
Quotes: "Harvey's gross margins reportedly reached -50% in June as usage costs rose under flat seat-based pricing."

Legora

Description: Legal AI software company.
Why mentioned: Moving to consumption-based pricing and model routing.
Quotes: "Legora is moving toward consumption-based pricing and model routing."

Salesforce, Atlassian, HubSpot

Description: Major SaaS platforms.
Why mentioned: Introducing agent-access charges for automated API usage.
Quotes: "New agent-access charges from Salesforce, Atlassian, and HubSpot could make automated API usage expensive."

SaaStr

Description: SaaS community and events business led by Jason Lemkin.
Why mentioned: Case study in agent API costs and a low-cost workaround.
Quotes: "Its AI VP of Revenue proposed mirroring the system of record into a $5 Postgres instance."

Mistral

Description: European AI model company.
Why mentioned: Record-setting European round.
Quotes: "Mistral's $3.5B round set a record for a Europe-based company."

Plaid

Description: Financial data network.
Why mentioned: Source of the report on financial network intelligence for fraud prevention.
Quotes: "Behavioral patterns across institutions and platforms offer additional signals for detecting fraud."

Nous Research

Description: Open AI models developer.
Why mentioned: Notable deal.
Quotes: "Nous Research raised $90M at a $1.5B valuation to develop open AI models and expand its AI research capabilities."

Type One Energy

Description: Stellarator-based fusion developer.
Why mentioned: Largest deal in the list.
Quotes: "Type One Energy raised $200M in Series B funding to advance its stellarator-based fusion technology."

VEIR

Description: Superconducting power delivery company.
Why mentioned: Infrastructure play tied to AI data center power demand.
Quotes: "VEIR raised $110M in Series C funding... to scale superconducting power delivery technology for AI data centers."

Spiko

Description: Tokenized money-market fund provider in Europe.
Why mentioned: Example of the tokenization trend in finance.
Quotes: "Spiko raised $90M in Series B funding to expand its regulated tokenized money-market funds across Europe."

Universal Quantum

Description: Modular quantum-computing company.
Why mentioned: Notable deal.
Quotes: "Universal Quantum raised over $100M in Series A funding to advance its modular quantum-computing technology."

Parallel Systems

Description: Autonomous, battery-electric freight rail developer.
Why mentioned: Notable deal.
Quotes: "Parallel Systems raised $100M in Series C funding led by AVP."

Stuut

Description: Enterprise order-to-cash automation.
Why mentioned: Notable deal.
Quotes: "Stuut raised $52.5M in Series B funding led by Insight Partners to automate enterprise order-to-cash workflows."

Namespace Labs

Description: AI-focused developer infrastructure.
Why mentioned: Notable deal backed by Scale, NEA, and 20VC.
Quotes: "Namespace Labs raised $42M in Series B funding... to expand its AI-focused developer infrastructure."

Other notable raises (healthcare/biotech)

General Medicine ($120M Series B), Multiply Labs ($75M Series B, "robotic manufacturing systems for cell therapies"), Sensible Biotechnologies ($47M Series A, "cell-based mRNA manufacturing platform"), Yazen (€50M, digital obesity treatment).

4. People Identified

Ernie Tedeschi and Marisa Rama

Description: Authors of the Stripe Economics research agenda.
Why mentioned: Leading Stripe's AI empirical research.
Quotes: "Its research will also examine economic risks and the market for intelligence, with opportunities for external researchers to collaborate."

Tomas Pueyo

Description: Author of "When Will We Make God?"
Why mentioned: Forecasting superintelligence timelines.
Quotes: "Scaling laws, longer task horizons, benchmark gains, AI-generated code, and prediction markets inform forecasts ranging from one to ten years."

Simon Taylor

Description: Fintech writer.
Why mentioned: Author of "Fintech is Dead."
Quotes: "setting the stage for an AI-first, token-first financial giant."

Jason M. Lemkin

Description: Founder of SaaStr.
Why mentioned: Documenting agent API cost risk and executive-turnover data.
Quotes: "an analysis of 22,000 attendees found that 49.7% of executives had changed jobs within 16 months."

Aaron Harris

Description: Author on fundraising timing.
Why mentioned: Created the "Decisive Moment" framework.
Quotes: "The Decisive Moment framework focuses on aligning the business opportunity, the reason for raising, and an investor prepared to recognize the opportunity."

Laurie Owen

Description: Author of the Thrive analysis.
Why mentioned: Dissecting Thrive's brand-building playbook.
Quotes: "Thrive Capital cultivates its reputation through high-profile deals, founder endorsements, design leadership, merchandise, and personalized gifts."

Drew Crawford

Description: Author of the Brazil hard-assets piece.
Why mentioned: Thesis on physical inputs for AI.
Quotes: "Brazil's renewable power, quartz, copper, graphite, rare earths, and farmland position the country to supply resources demanded by AI infrastructure."

Inaki Bererguer

Description: Author on AI video creation.
Why mentioned: Orchestration thesis.
Quotes: "orchestration across video, image, voice, and sound models a key production advantage."

Alex Kantrowitz

Description: Founder of Big Technology.
Why mentioned: Speaker at the sponsor's webinar on AI and the startup playbook.
Quotes: "Alex Kantrowitz, founder of Big Technology, and Vanta's SVP of Engineering share patterns before they're conventional wisdom."

5. Operating Insights

Price AI products against inference cost, and put AI spend in COGS

Flat per-seat pricing can produce negative margins when usage is heavy. Move to consumption-based pricing, route queries to cheaper models, and track AI costs as cost of goods sold.

"Legora is moving toward consumption-based pricing and model routing, highlighting how AI software companies must align pricing with inference costs and account for AI expenses in COGS."

Use cheaper, fine-tuned open-weight models where frontier quality isn't required

"...prompting it to adopt a fine-tuned open-weight model that can cost up to 90% less than frontier alternatives."

Differentiate through domain depth, not the standard AI stack

Since everyone shares agent loops, connectors, memory, and sandboxes, stand out through customer empathy, domain expertise, and a clear value proposition.

"Deep domain expertise, customer empathy, and a clear value proposition offer stronger differentiation than another standard AI stack."

6. Overlooked Insights

Executive churn is extreme

SaaStr's attendee analysis suggests rapid job turnover among senior operators, which matters for sales cycles, champion risk, and hiring.

"an analysis of 22,000 attendees found that 49.7% of executives had changed jobs within 16 months."

Agents may become metered customers and cost centers on legacy platforms

The idea of mirroring a system of record into a tiny database to avoid per-call fees hints at a broader architecture pattern as incumbents monetize agent access.

"Its AI VP of Revenue proposed mirroring the system of record into a $5 Postgres instance."