Public transcript

🔴 Alex Karp LIVE from AIPCon 10

By: TBPN
Length: 1 h 54 min
Language: English
Text from: the video's captions
Transcribed: 13 October 2026

Read the full transcript Open on YouTube

Notes in English.

Summary

This live episode of TBPN from Palantir's AIPCon 10 covers the AI industry's open letter on mandatory nucleic acid synthesis screening, Ramp's $750 million raise at a $44 billion valuation, and Benchmark's first growth fund. Alex Karp joins to discuss token overuse in enterprises, Palantir's ontology advantage, and the risk of nationalization, followed by AIG's Peter Zaffino on transforming underwriting with Palantir, forward-deployed architect Chad Wolquist on malleable software and dashboards, and USDA's Sam Barry on agriculture, SNAP fraud detection, and data centers.

Key points

AI and biotech leaders including Demis Hassabis, Sam Altman, Dario Amodei, and Alex Wang signed an open letter asking the US government to mandate nucleic acid synthesis screening, since only about 80% of commercial synthesis capacity is voluntarily covered by the International Gene Synthesis Consortium.
Ramp raised $750 million at a $44 billion valuation, making it worth more than PayPal despite PayPal's $32 billion in revenue, with the standout line that Ramp was 120th its current size the last time it grew this fast.
Benchmark raised $2 billion across two new funds, including its first ever dedicated growth fund, with Eric Glyman-style growth talent like Chrys Bulych having joined.
Alex Karp argues enterprises are tired of 'token maxing' and that business problems require 'taste plus money,' with Palantir's ontology and forward-deployed model providing structure that large language models alone cannot.
Karp warns that AI companies risk nationalization and regulation by people who don't understand the technology, and says CEOs bragging about firing two-thirds of their workforce with AI are inviting a political backlash.
AIG's Peter Zaffino says Palantir helped build a full ontology of AIG and then modeled the Everest acquisition's $2 billion premium portfolio on top of it in four days, with the partnership run on 90-day goals.
Chad Wolquist says Palantir's new Evolve tool helped customers eliminate 60% of token cost in two days by rearchitecting, picking different models, and prompt tuning.
USDA's Sam Barry highlights SNAP's roughly 12% improper payment rate (about $12 billion a year), a USDA-SpaceX partnership, and the need for automation as farm succession becomes a crisis.

Questions it answers

04:30The bio threat letter

Why did AI and biotech leaders sign an open letter on nucleic acid synthesis screening?

Because researchers have shown viruses like polio and the Spanish flu can be synthesized from published sequence data alone, and AI could make designing dangerous sequences easier. The letter asks the US government to mandate screening of synthesis orders, customer verification, and recordkeeping, since the existing voluntary consortium only covers a self-reported 80% of capacity.

Polio was synthesized from public data in 2002; Spanish flu in 2005.
About 80% of commercial synthesis capacity is voluntarily covered, leaving a 20% gap.
The hosts see it as a practical policy request rather than AI doom marketing.

14:14Biotech momentum

Is biotech recovering as an asset class?

There is notable momentum in early-stage biotech after the sector had been left for dead, with activity from Isomorphic Labs, New Limit, Retro Biosciences, Altos Labs, and Anthropic's acquisition of Coefficient Bio. However, the scale is much smaller than AI's trillions, and biotech has more of a culture of base hits and frequent company flips.

A biotech investor on the show ~14 months ago questioned why anyone would invest in the asset class.
Biotech exits tend to be smaller, more frequent flips rather than extreme power-law outcomes.

17:12Ramp's raise

What did Ramp announce and why does it matter?

Ramp raised $750 million at a $44 billion valuation, making it worth more than PayPal despite PayPal's $32 billion of revenue. The standout line is that Ramp was 120th its size the last time it grew this fast, and Eric Glyman published an essay framing tokens as the third pillar of finance after people and vendors.

$750 million raised at a $44 billion valuation.
Ramp is now worth more than PayPal.
Glyman calls token spend the 'quadrillion token blind spot.'

20:00Sabi, SpaceX, and Benchmark

What other fundraising news is discussed?

Sabi, a beanie BCI company, is reportedly being preempted at $35 million on a $500 million post with $200 million of B2B LOIs. Goldman expects SpaceX's AI revenue to surge 100x by 2030 with ~$360 billion of capex through 2028, and Benchmark raised $2 billion across two funds including its first dedicated growth fund.

Sabi has $200 million of LOI from B2B customers.
SpaceX forecasts ~$360 billion of capex through 2028.
Benchmark's first growth fund follows its hire of growth-focused talent.

28:30Karp on token maxing and taste

How does Alex Karp think enterprises should deploy AI?

Karp says business value comes from 'taste plus money' — knowing which problems are worth solving — and that enterprises are tired of 'token maxing' that looks like addiction. Palantir's ontology, FDEs, and on-prem security provide the structure LLMs lack, and frontier companies are charismatic with investors but not with enterprises.

Palantir's secret sales pitch: try the frontier company first, then come to us.
Free code (dashboards, one-off analysis) is magical but doesn't solve core enterprise problems.
Competition in defense tech expanded the market and set comparators rather than hurting Palantir.

47:16Karp on nationalization and layoffs

What political risks does Karp see for AI companies?

Karp has spent six months telling industry titans that nationalization is coming if they don't get their act together, and that regulation by people who don't understand the technology is the primary risk. He also argues that CEOs who say AI let them fire two-thirds of their workforce are effectively signing up for Bernie Sanders' manifesto, and that upskilled workers become more valuable, not less.

Karp predicts nationalization or heavy regulation if the industry doesn't self-police.
He clarifies he does not want a draft, but wants communal structures to remember being American.
Soldiers at the bottom have gotten more valuable using Palantir's products.

53:00Peter Zaffino of AIG

How is AIG using AI and Palantir?

Zaffino explains AIG underwrites complex global risks and has worked with Palantir for nine years, building a full ontology of AIG and then modeling the Everest acquisition's ~$2 billion premium portfolio on top in four days. The partnership runs on 90-day goals, embeds Palantir engineers with AIG teams, and focuses on growth and reskilling rather than job elimination.

AIG is 50% international, with Japan its second-largest country.
Palantir's data ingestion lets AIG use data it previously couldn't translate into workflows.
AIG's aspiration is growth and reskilling, not cutting jobs with AI.

1:06:00Chad Wolquist on Palantir architecture

What is Palantir's thesis for AI in the enterprise?

Wolquist says humans and AI work together, with the ontology providing the world model LLMs lack, and agents run combatively against each other with human feedback. He wants to 'kill all dashboards' — KPIs should be byproducts of operational applications — and Palantir's Evolve tool analyzes production logs to cut token costs, with one customer eliminating 60% of token cost in two days.

The ontology models the real world multimodally, not just rows and columns.
Companies often don't know their most profitable product because data is aggregated into KPIs.
Palantir has a free dev tier and a Shopify integration for small builders.

Notes

Bio Threat and Nucleic Acid Screening Letter

In 1981 researchers published the polio virus genome sequence; in 2002 infectious polio was synthesized from public data alone, and in 2005 the Spanish flu was reconstructed the same way.
An open letter signed by Demis Hassabis, Sam Altman, Dario Amodei, Alex Wang and dozens of others asks the US government to force nucleic acid synthesis companies to screen orders for sequences of concern, verify customers, and keep records.
The International Gene Synthesis Consortium (since 2009) covers a self-reported ~80% of commercial synthesis capacity; membership and reporting are voluntary, leaving a 20% gap.
The hosts found it refreshing that the letter is a practical policy request rather than another apocalyptic AI doom warning. Biotech momentum is picking up after being left for dead, with Isomorphic Labs, New Limit, Retro Biosciences, and Altos Labs active, though at smaller scale than AI's trillions.

Ramp and Other Fundraising

Ramp raised $750 million at a $44 billion valuation — now worth more than PayPal, which has $32 billion of revenue but negative momentum, while Ramp was 120th its size the last time it grew this fast.
Eric Glyman's essay frames tokens as the 'third pillar' of finance after people and vendors, boiling finance down to: who spent what, was it worth it, what's the bill next month.
Sabi, a beanie BCI company, is reportedly being preempted at $35 million on a $500 million post with $200 million of B2B LOIs.
Benchmark raised $2 billion across two funds, including its first dedicated growth fund.

Alex Karp

Karp compares unchecked enterprise LLM use to addiction and 'token maxing'; Palantir's value is 'taste plus money' — knowing which business problems are worth solving.
He distinguishes infrastructure code, managed code (FDEs), and 'free code' (dashboards, one-off analysis) — the last is magical but not the core enterprise problem.
Frontier labs are charismatic with investors but not with enterprises; Palantir's secret sales pitch is 'don't even call us' until you've tried the frontier companies.
He has warned industry titans for six months that nationalization is coming if the industry doesn't self-police, and criticizes CEOs who blame AI for layoffs.

AIG's Peter Zaffino

AIG is 50% international, 50% North America, with Japan its second-largest country, focused on complex risks like marine and energy.
Palantir built AIG's full ontology, then modeled the Everest acquisition (~$2 billion premium) on top in four days; the partnership runs on 90-day goals.
Zaffino's philosophy is growth and reskilling, not job elimination.

Chad Wolquist (Palantir FDE)

Decomping problems with humans plus AI, using the ontology as the world model; agents run combatively against each other with human feedback.
The Evolve tool analyzes production logs to swap models, cache, and prompt-tune — one customer cut 60% of token cost in two days.
He wants to 'kill all dashboards': KPIs should be byproducts of operational applications.

Sam Barry (USDA)

USDA includes SNAP (~$100B/year), Forest Service, rural development loans, and a large scientific arm; food self-sufficiency is a geopolitical edge.
SNAP has ~12% improper payments (~$12 billion/year); USDA is pushing states for data to deploy AI fraud detection.
USDA is pursuing a SpaceX partnership; Barry suggests data centers belong in space or in struggling industrial areas rather than farmland.

Text from the video's own captions. Summary and notes written by AI from the transcript, so check anything important against the recording.