The State of Adventure Capital

Author: Grant Gregory (Partner, Cantos) · Published: October 2025 · Format: 286-slide deck · URL: https://www.slideshare.net/slideshow/the-state-of-adventure-capital-grant-gregory-cantos/283913569

One-line: 286 slides arguing that the hard-tech category needs a vocabulary before it can be underwritten — supplying one (the Deep Tech Score, 0 to 5), then making the case that the timing failure that killed Clean Tech 1.0 has been fixed by companies, geography and capital.

Summary

The deck’s organizing complaint is a vocabulary problem. Gregory opens with the slide “Hard Tech / Deep Tech / Adventure Capital / Frontier Tech / Dynamism” — five names for one thing, none of which distinguishes a capital-light hardware company from a fusion reactor. He takes his preferred term from Sebastian Mallaby’s The Power Law: “The defection of the Traitorous Eight in 1957 was made possible by a new form of finance, originally dubbed adventure capital.”

Software ate the world, but not all of it. The S&P 500’s top ten went from IBM, AT&T and Exxon in 1980 to Nvidia, Microsoft and Apple in 2025; as of June 2024, 92 S&P companies (18.4%) were venture-backed pre-IPO, worth $20.5T or 43.6% of market cap. But the 2000–2025 price series splits cleanly: software, TVs, toys and computers got cheaper; hospital services, college tuition, childcare, food and housing got more expensive. The listed reasons are regulation, oligopolies, bureaucracy, monopsonies, demographics, capital flows, offshoring and zoning. Before 2022 only a handful of multistage firms took those categories seriously — Khosla, Lux, Eclipse, Obvious, DCVC, Founders Fund. After “It’s Time to Build,” everyone did, and the deck papers the wall with the evidence: General Catalyst’s Industrial Resilience, BCV’s Industrial Renaissance, Coatue’s America’s Industrial Reboot, Point72’s Defense Tech Revolution, YC’s Techno-Industrialist RFS.

The ADCU. Gregory’s frame for the opportunity is the “American Dynamism Cinematic Universe” — “basically anything that touches huge parts of our lives (and GDP), that hasn’t really been touched by technology yet.” An ADCU company is heavily enmeshed in the physical and digital world, solves the hardest problems itself, and is multi-product across multiple end markets; “the technology + business model are the vehicle that brings software to an industry in desperate need of change.” His most mature example is Amazon. It is explicitly not clean tech, science projects or moonshot vanity projects. The attraction list: huge opportunities, differentiated moats, you only invest in monopolies or oligopolies, M&A exits are actually feasible, and “if you make it through, you’re alone.” He runs the counterargument too, quoting David George: “You can create amazing venture scale outcomes that are even better than software-only businesses, but they’re probably going to be fewer and further between. They’re very, very, very hard to pull off.”

The Deep Tech Score. The deck’s most useful original contribution, built as a companion to NASA’s 9-level TRL scale and justified with a slide about surfers having twenty words for waves — “the greater your skill, the richer your vocabulary.” Six levels: 0 SaaS (a standard software company), 1 Slingshot (feasible and in range, capital-light hardware, $0–10M capex, under a year to revenue), 2 Snipershot (far out but within sight, moderate capex and engineering, $10–20M, 1–2 years), 3 Longshot (out of sight but feasible, $30–100M, 3–5 years), 4 Moonshot (the classic deep tech company, science plus capex risk, $100M+, 5+ years), 5 Shot in the Dark (“Let’s see if this works. The fat startup. Raise $1B & see what happens”). Worked examples run Notion/Airtable at 0, MoldCo at 1, Neros/Shinkei at 2, Anduril/Rivian at 3, Waymo/SpaceX at 4, OpenAI/CFS at 5. Plotted against capital intensity, SaaS and Shot-in-the-Dark are the crossed-out corners; 1 through 3 is the playable field.

Why now. Clean Tech 1.0’s DTS was “probably a 4.5.” The seven listed causes of its failure — capital needs, not enough talent, tech wasn’t ready, stringent regulation, well-capitalized incumbents, multi-year time to revenue, China’s state-led VC — reduce to one: “It really just boils down to: ‘Too early.’” Hence “Every failed idea from the dotcom bubble would work now” (Andreessen, 2020), and the pairings: Webvan/DoorDash, Pets.com/Amazon, Clippy/OpenAI, Friendster/Facebook. Gaurab Chakrabarti supplies the sharpest line: “If you’re one step early, you’re a genius. If you’re two steps early, you’re a martyr.”

What’s changed is the 3 C’s — Companies, SoCal, Capital. Trailblazers (SpaceX, Tesla, Palantir) paved the way and seeded mafias exactly as Fairchild and PayPal did, with a slide on the qualities of the parent company mattering and “the blessings of adversity” against “the curse of resources.” Hardware talent found a home: LA has been the centre of aviation for a century, VC-backed LA companies grew 57x, $80B invested in five years against 3x that in the prior decade combined, and El Segundo now maps as a dense cluster around SpaceX. And downstream financing is derisked. Layered on top are 3 deep currents: cheap ubiquitous compute (Starlink, AWS, smartphones), customer desperation and repatriation, and geopolitical conflict. Then the reshoring evidence — US dependence on China for electronics up 600% from 2014–2022, counts of Chinese semiconductors in critical US military platforms, and manufacturing construction spending going vertical since 2021.

Is it venture scale? Hardware investment through 1H25 exceeded all of 2024. Venture scale gets defined as high growth (exponentiality), high margins (power and leverage) and low dilution (capital efficiency), quoted through Don Valentine, Michael Moritz and Doug Leone, then compressed into a master formula: VC = [ idea * execution * capital efficiency ] * MVB * TTR. The market-size argument is that software is big — $5.4T of worldwide IT spend — but under 1% of a $113.8T global GDP, against space $800B, defense $2.4T, mining $2.3T, energy $7.8T, agriculture $5.3T, construction $9.4T, chemicals $6.3T, manufacturing $6.9T, healthcare $9.2T, with incumbents averaging 111 years old.

Exits. Median IPO age is now 14 years, up from 5 in 1999 and 8 in 2022. Top-1% exit outcomes roughly double every five years ($400m in 2005–09 to $10,234m in 2020–24), so David Clark’s implication is that a fund investing today should underwrite a $40B+ top exit over an 8–10 year hold. The headline statistic: ADCU companies have 2.4% odds of a $250m+ exit versus software’s 1.27% — roughly double. The history lesson is that every new category is questioned: Shopify’s own outcomes analysis said “we thought the biggest possible outcome was ~$50m” before a $1.27B IPO and a ~$200B market cap; Frank Rotman on fintech in 2008, “Nobody wanted to fund anything in fintech, there wasn’t a word for it by the way.” And every era of venture crowns new firms — 1980s Sequoia/Benchmark/Kleiner, 2000s USV/Founders Fund/YC/First Round, 2010s a16z/Thrive/Floodgate, 2015s Greenoaks/Paradigm, and for 2025 onwards a question mark.

Method. Search for maze historians — “it’s not enough to navigate the Idea Maze. You need to have encyclopedic knowledge.” Iterate or die. Beware the Field of Dreams fallacy: if you build it, they probably won’t show up. “The recipe is simple: start with a blank slate, design solutions from scratch, use mature existing technologies, then you layer in the new tech stacks.” The Great Tension is Buffett’s line about a management team with a reputation for excellence meeting an industry with bad economics. And the experts are wrong — the New York Times, October 9, 1903, gave the flying machine “one million to ten million years”; the Wright brothers flew two months later.

Predictions. The last thirty slides are roughly 25 named predictions, each signed by a different investor — commodities, Wright’s Law returning to American industry, AI scientists running the lab, service models beating machine sales, sub-vocalization unlocking ubiquitous AI, SpaceX’s playbooks remaking American industry, government becoming part of the venture stack, energy abundance, a nuclear ecosystem boom, data centers as the hardware testing sandbox, majority-robotic soft-tissue surgery, Wall Street welcoming defense tech, venture driving basic science, add-on hardware as an EBITDA lever, Israel as a critical-industries hub, industrial franchises, human superpowers, America’s private “shadow grid,” space tourism by 2030, creative capital stacks, pre-seed reweighting toward hardware, and the PE/venture line blurring in manufacturing.

Full text

The deck has no text layer. All 286 slides are archived at 2048px alongside an outline: ../attachments/the-state-of-adventure-capital/the-state-of-adventure-capital.md, slides in wiki/attachments/the-state-of-adventure-capital/slides/. The original DocSend link (cantos.docsend.com/view/sq7p7n49bxd9j8g7) is email-gated, so SlideShare was the open mirror; SlideShare rots, which is why the images were pulled locally.

Connections

• Deep Tech For Deep Minds Or Deep Pockets — Kyle’s own essay on the same question of who can actually underwrite hard tech; the Deep Tech Score is the taxonomy that essay was reaching for, and levels 1–3 are precisely the “deep minds, shallow pockets” band.
• Venture Capital Models — the master formula, the venture-scale definition, and the “every era of venture crowns new firms” slide, which is the deck’s real pitch.
• American Dynamism — the ADCU is an explicit reframing of it, and Gregory was a founding partner on a16z’s American Dynamism team before Cantos.
• The Mindset Shift We Need to Rebuild — Erin Price-Wright’s keynote from the team he came out of; same thesis, operator-facing rather than allocator-facing.
• Hard Tech and Power Law — the vocabulary problem, and the Mallaby source of “adventure capital.”
• Sebastian Mallaby — The Power Law supplies the term and the 1957 Traitorous Eight origin story.
• Founders Fund, Palantir, SpaceX, Tesla, Anduril — the trailblazer set and the mafia-genealogy argument.
• Cleantech — the failure case the whole “why now” section is built against, with “too early” as the single-cause diagnosis.
• Manufacturing, Supply Chain, Nuclear Energy, Robotics — the end markets the predictions section fans out into.
• Shopify and Fintech — the two “every new category is questioned” case studies.