The YC Ledger

Started as a Palo Alto restaurant-delivery app; now a broad local-commerce platform spanning restaurants, grocery, alcohol, and retail, plus DashPass subscriptions and a fast-growing advertising business. International operations run under Wolt and Deliveroo.

Founders

Tony Xu (CEO) — Stanford GSB MBA, UC Berkeley industrial engineering; prior McKinsey, eBay/PayPal strategy, VC at Matrix Partners. Stanley Tang — Stanford CS, briefly a Facebook engineer. Andy Fang — Stanford CS. Evan Moore — co-founded the original concept with Xu, left ~17 months in. All four met through Stanford entrepreneurship programs around 2010–2012.

Funding & outcome

Raised roughly $2.5B pre-IPO (seed from Khosla Ventures, Series A led by Sequoia, a $400M Series H at ~$15.6B pre-money in 2020). IPO'd on the NYSE Dec 2020 at $102/share (~$32.4B), popped ~85% on debut. Current market cap roughly $91–94B; FY2025 revenue ~$13.7B, up ~28% YoY.

Business model

Consumers pay delivery/service fees, discounted via DashPass (~$9.99/mo). Merchants pay tiered commissions (15–30% on delivery, 6% on pickup). Advertising is the fastest-growing, highest-margin line, crossing a $1B+ annualized run rate combined with Wolt Ads. This layering makes sense because restaurant delivery itself is thin-margin and merchants can't build last-mile logistics alone — monetizing the resulting first-party purchase data via ads (the Amazon retail-media playbook) is where the real profit lives.

Business type

Three-sided marketplace / B2B2C platform: consumers pay for convenience, merchants pay for demand access and ad placement, Dashers supply gig-labor capacity, and DoorDash sits in the middle of all three.

The AI angle

AI is structurally important beyond generic hype. Real-time dispatch/routing optimization directly cuts cost-per-delivery, and DoorDash's new Autonomous Delivery Platform decides order-by-order whether to use a human Dasher, its self-built delivery robot, a drone, or a sidewalk robot — the first credible lever against delivery's structurally thin margins. Demand forecasting and AI-driven ad targeting compound on top of that.

Two-sided marketplace connecting hosts with guests for short-term stays, expanded into Experiences and a new Services category (chefs, photographers, cleaning) as it evolves toward a broader travel & local-services super-app. One of YC's earliest and most iconic investments — $20,000 for 6%.

Founders

Brian Chesky (CEO) and Joe Gebbia — both RISD industrial/graphic design grads who met in school. Nathan Blecharczyk — Harvard CS, joined as technical co-founder shortly after the original concept, built the first site. Chesky and Gebbia were recently underemployed design grads, not serial entrepreneurs, when they founded the company in 2007–08.

Funding & outcome

Raised roughly $6.4B across its life (Series A from Greylock, a $112M Series B led by a16z, a pandemic-era $1B raise in 2020 at a reduced ~$18B valuation). IPO'd on Nasdaq Dec 2020 at $68/share (~$47B), popped ~113% on debut to ~$100B. Current market cap roughly $108B.

Business model

Pure take-rate marketplace with no owned inventory — transitioning to a flat 15.5% host-only service fee (guests pay listed price, no separate fee), with Experiences carrying a flat 20% commission. This fits Airbnb's market because supply (individual hosts) is fragmented and low-trust by default; a fee that only triggers on a completed paid booking aligns Airbnb's incentives with actually closing trustworthy transactions rather than charging for mere listing exposure.

Business type

Two-sided C2C/B2C hybrid marketplace: hosts (largely individual, increasingly professional) supply inventory, guests are the demand side, and Airbnb never takes title to inventory — it monetizes purely via take-rate.

The AI angle

Airbnb's AI support tool now autonomously resolves roughly 40% of guest/host inquiries, a meaningful cost lever in a disputes-heavy support model. Chesky says AI now writes ~60% of Airbnb's code and cut feature-shipping time ~60%. Most consequential near-term: testing AI-driven natural-language search to replace keyword/filter browsing in a famously browse-heavy funnel, alongside likely extensions into AI-assisted dynamic pricing and fraud/trust & safety detection — core to a marketplace built on strangers transacting in private homes.

Diversified crypto financial-services platform: retail and institutional trading, custody (Coinbase Prime), the Base Ethereum layer-2 chain, derivatives/prediction markets, staking, and a central commercial partnership with Circle on USDC. Formally headquarters-free since 2020.

Founders

Brian Armstrong (CEO) — prior developer at IBM, consultant at Deloitte, engineer at Airbnb on fraud prevention and cross-border payments; built Coinbase nights and weekends before founding it in 2012. Fred Ehrsam — Duke CS/Economics, FX trader at Goldman Sachs before co-founding; later left to co-found crypto investment firm Paradigm.

Funding & outcome

Sources disagree on the exact pre-listing total, roughly $525–850M (Union Square Ventures seed, a16z Series B, a $300M Series E led by Tiger Global in 2018). Went public via direct listing on Nasdaq, April 2021, at a $250 reference price (~$65B), closing day one near $328 (~$85–86B). Current market cap roughly $39–40B, added to the S&P 500 in 2025.

Business model

Transaction revenue (retail spread/fees, institutional volume fees) is highly crypto-cycle-sensitive, so Coinbase has deliberately built a larger recurring layer: USDC-related interest income, custody fees on $350B+ AUC, Coinbase One subscriptions, staking commissions, and Base L2 sequencer fees. This mix exists specifically to smooth earnings across crypto's boom-bust cycles — a strategic necessity unique to running a business on top of an asset class this volatile.

Business type

Fintech infrastructure with an embedded two-sided exchange at its core, spanning B2C (retail trading, Coinbase One) and B2B (institutional custody, Base as programmable settlement rails) — a deliberate shift from marketplace-transaction economics toward platform/infrastructure economics.

The AI angle

One of the more concretely AI-forward companies here. A domain-specific fraud model reportedly improved card-testing fraud detection from 59% to 97% for large merchants; AI chatbots handle roughly 65% of support contacts. Most structurally significant is x402, a payment protocol Coinbase introduced letting AI agents autonomously pay for resources over HTTP using stablecoins — Coinbase co-founded the x402 Foundation with Cloudflare, joined by Google, Visa, AWS, Circle, Anthropic, and Vercel. If machine-to-machine commerce materializes, this positions Coinbase's rails as a default settlement layer: a genuinely new revenue source, not just an efficiency play.

North America's leading online grocery marketplace — same-day delivery/pickup across 350+ retailers and 25,000+ stores, fulfilled by independent shopper gig workers — expanded into advertising and a B2B enterprise stack (Storefront, Caper Carts, FoodStorm) licensed to retailers.

Founders

Apoorva Mehta (co-founder) — University of Waterloo engineering; worked at BlackBerry, Qualcomm, and Amazon on supply-chain/fulfillment before founding Instacart in 2012 after roughly 20 failed startup ideas. Max Mullen and Brandon Leonardo — co-founders who joined around the YC batch; detailed prior background unverified.

Funding & outcome

Raised roughly $2.9B across ~19 rounds (Sequoia, a16z, D1 Capital). Valuation swung sharply: $13.8B (mid-2020) → a $39B peak (March 2021) → internal markdowns to $10B by late 2022, a ~75% drop from peak. IPO'd on Nasdaq Sept 2023 at $30/share (~$10B), popped ~40% on debut. Current market cap roughly $11.3B.

Business model

Consumers pay delivery/service fees (5–15% of order value) plus tips, discounted via an Instacart+ subscription. Retailers pay per-order fees for marketplace/fulfillment access — the lower-margin bucket. Advertising (Carrot Ads) is the actual profit engine at ~80% margins. Instacart Platform licenses the white-label tech stack, including NVIDIA-powered Caper Carts, back to retailers internationally. The layering exists because grocery itself is thin-margin; Instacart converts the resulting first-party shopping data into high-margin ads, then deepens retailer lock-in by licensing the tech back to them.

Business type

Multi-sided (three-to-four-sided) marketplace / B2B2C platform: retailers supply inventory and pay for ad placement, consumers are demand, shoppers fulfill orders, and CPG brands pay for placement — much of the consumer relationship is white-labeled through retailer branding, making the B2B2C label especially apt.

The AI angle

Several features are already shipped, not hypothetical: 'Ask Instacart' (natural-language search/recommendation), Caper Carts (computer vision identifying items in real time for personalized in-store promotions, now licensed internationally), and white-labeled enterprise AI tools for smaller grocers who can't build this in-house. The more strategic point: AI improvements compound across every revenue line at once — better ad targeting hits the highest-margin business directly, demand forecasting deepens retailer stickiness, and shopper-routing optimization improves core delivery unit economics simultaneously.

A single DevSecOps platform unifying source control, CI/CD, security scanning, and an integrated AI suite (GitLab Duo). Joined YC already at ~10 employees with paying customers, mainly to learn how to scale.

Founders

Dmitriy Zaporozhets — Ukrainian developer who wrote GitLab's first commit in 2011 while at Ukrainian firm Sphere Software; lead author of GitLab CE/CI, continued working from Ukraine. Sytse “Sid” Sijbrandij — Dutch co-founder/CEO who spent ~4 years building recreational submarines before discovering the GitLab open-source project and partnering with Zaporozhets on the business side.

Funding & outcome

Raised roughly $420–434M pre-IPO (Khosla Ventures and 500 Startups in the seed, later valuations reaching $6B in a 2021 secondary). IPO'd on Nasdaq Oct 2021 at $77/share (~$11B), closed day one near $103.89 (~$14.9B). Current market cap roughly $7.2B. FY2026 revenue $955M (+26% YoY), surpassing $1B ARR. Datadog has reportedly explored a takeover since mid-2024, unconfirmed as of this writing.

Business model

Tiered per-seat SaaS (Free, Premium, Ultimate) with GitLab Duo AI sold as a paid add-on, shifting toward hybrid seat-plus-usage pricing (agentic code review priced per review via credits). This fits DevOps tooling's natural bottom-up adoption: individual developers try the free Community Edition, and sales converts usage into enterprise contracts once security and scale requirements kick in — the open-core tier is the funnel.

Business type

B2B SaaS / platform. Buyers are enterprise engineering leadership, not consumers; the open-core Community Edition (50M+ registered users) creates a low-friction funnel that converts into paid enterprise contracts, a classic open-source-led B2B growth model.

The AI angle

AI is central to strategy, and cuts both ways sharply. On offense, GitLab Duo has expanded into a full agentic platform: autonomous code review, vulnerability remediation that opens merge requests unprompted, and a security-review flow catching business-logic issues static scanners miss — the bet is that enterprises pay for AI embedded across a governed lifecycle rather than a standalone coding tool. On defense, the threat is real: GitHub Copilot's share among professional developers fell from 67% to 51% in 2026 survey data as AI-native entrants (Cursor, Claude Code) gained ground fast, putting GitLab in a pincer between GitHub's bundling and faster-moving AI-native tools — making AI capability genuinely make-or-break for its enterprise position.

Cloud file storage and sync for individuals and teams, now expanding into AI-powered universal search and knowledge work via Dash — a pivot away from pure storage as that market commoditizes.

Founders

Drew Houston (CEO) — MIT EECS; founded an SAT-prep startup at MIT and worked as a startup engineer before Dropbox. Arash Ferdowsi — MIT CS student who dropped out to co-found after seeing Houston's demo, met through MIT. Houston applied to YC alone in 2007; Paul Graham told him to find a co-founder.

Funding & outcome

Raised roughly $607M (Sequoia seed 2007, Accel Series A 2008, a ~$250M round in 2011). IPO'd on Nasdaq March 2018 at an initial valuation of roughly $9.2–11.2B. Current market cap roughly $10B — flat-to-down from IPO, with trailing revenue around $2.5B and net income around $443M.

Business model

Freemium subscription SaaS: free tier, paid individual/family plans, per-seat team plans, plus add-ons (Dash AI search, Dropbox Sign). This makes sense because sync has near-zero marginal cost once infrastructure is built and the buying trigger (running out of storage) converts well — but storage is now a commodity utility bundled free by Google, Microsoft, and Apple, capping Dropbox's pricing power.

Business type

Primarily B2C with a B2B/SMB layer (Dropbox Business, team plans) — horizontal SaaS sold both directly to consumers and to companies as a per-seat tool, not a marketplace.

The AI angle

A genuine, if defensive, lever. Dropbox's core business is a commoditized utility given away free by hyperscalers, so its main path is layering AI-powered universal search (Dash) on top of files it already stores — repositioning from 'where your files live' to 'how you find and act on your knowledge.' This is a real moat opportunity because Dropbox has a privileged, permissioned view into a company's actual working documents, but it's racing Google Drive and Microsoft 365 Copilot, which have the same file access plus far deeper enterprise identity integration — AI likely slows the decline more than it reignites growth.

Social news and discussion platform organized into topic communities ('subreddits'). Part of YC's first-ever batch in 2005; acquired by Condé Nast in 2006, spun out independently in 2011.

Founders

Steve Huffman (CEO) and Alexis Ohanian — University of Virginia roommates, both 2005 grads, who pitched Reddit to YC after Paul Graham rejected their first idea. Aaron Swartz is often cited as a third co-founder via a merger shortly after founding — credited as such by Paul Graham, though he left soon after the Condé Nast acquisition and his status is sometimes disputed; he later became a well-known internet activist and died in 2013.

Funding & outcome

Raised roughly $1.3B over ~10 rounds (a16z, Sequoia, Tencent, Fidelity, plus individual backers including Sam Altman and Peter Thiel). IPO'd on the NYSE March 2024 at $34/share (~$6.4B), jumping ~48% on debut. First full-year GAAP profit in 2025 (~$530M net income). Current market cap roughly $29–35B, up 4–5x from IPO.

Business model

Primarily digital advertising, plus a fast-growing data-licensing business selling API/content access to AI labs for model training — $203M in aggregate licensing contract value disclosed in its S-1, including deals with Google (~$60M/year) and OpenAI. This works because Reddit's core asset — enormous volumes of authentic, vote-ranked, topic-structured human conversation — is valuable to advertisers for intent targeting and uniquely valuable to AI labs as high-signal training data that's hard to replicate.

Business type

B2C platform with a B2B2C dynamic layered on top: users generate and consume content for free, advertisers pay to reach them, and now AI companies pay for bulk data access built on that same content — three monetizations of one underlying asset.

The AI angle

One of the strongest, most concrete AI angles on this list, and genuinely two-sided. Reddit's user-generated content has become a direct, disclosed, material revenue line via AI training/licensing deals — a business model that essentially didn't exist five years ago. The risk cuts the other way: AI chatbots that answer questions directly instead of sending users to Reddit threads could cannibalize its own traffic and ad impressions, and Reddit's countermove (blocking non-paying crawlers, its own 'Reddit Answers' AI search feature) is itself an AI-native bet. Net: AI is simultaneously Reddit's newest revenue stream and its biggest traffic threat.

Live-streaming platform, primarily for video-game and esports content, with real-time chat interaction. Began as Justin.tv (a 24/7 lifecast) in 2007 and pivoted to game streaming in 2011.

Founders

Justin Kan and Emmett Shear — launched Justin.tv together through YC; Shear steered the pivot to Twitch and served as CEO through the Amazon era, later briefly interim CEO of OpenAI (Nov 2023). Kyle Vogt was also an early co-founder/engineer — he later founded Cruise (also in this ledger).

Funding & outcome

Early Justin.tv/Twitch raised roughly $8–15M+ across early rounds (Alsop Louie Partners, Bessemer, Draper Associates) — no single confirmed total. Acquired by Amazon in August 2014 for approximately $970M in cash (some estimates closer to $1.1B fully loaded with retention equity), at the time Amazon's largest acquisition. Operates as a wholly-owned subsidiary; Amazon doesn't break out Twitch financials separately.

Business model

Subscriptions ($4.99–$24.99/mo, split roughly 50/50 with streamers by default, up to 70/30 in the streamer's favor for top partners), Bits (a tip currency where Twitch keeps ~30%), and CPM advertising. Estimated $1.9B revenue in 2025. This fits Twitch because its value is a two-sided attention marketplace between streamers who need monetization to go full-time and viewers who pay to support specific personalities — direct fan-to-creator payment mirrors how parasocial fandom and live entertainment actually work.

Business type

Two-sided platform / marketplace connecting streamers (content supply) and viewers (demand), with Twitch taking a cut of subscriptions and virtual currency plus selling ads — the product itself is produced by the user base, making this C2C-with-platform-monetization rather than classic B2C.

The AI angle

A moderate, mostly operational lever rather than a business-model transformer. Real-time AI moderation at chat scale, AI-assisted discovery to surface niche streamers to the right viewers, and AI clipping for repurposing VODs into shorts are all concrete, already-in-motion improvements to retention and cost. What AI does not replace is the core product: live, unscripted human personality and real-time interaction is Twitch's actual differentiator versus on-demand or AI-generated video — a real ceiling on how far the AI story goes.

Self-driving vehicle technology company that developed and, until December 2024, operated a driverless robotaxi ride-hailing service in select US cities before General Motors ended its funding.

Founders

Kyle Vogt (CEO) — MIT CS/EE; previously co-founded Twitch (via Justin.tv) and Socialcam (acquired by Autodesk) — a serial YC founder with two prior exits before Cruise. Daniel Kan (COO) — previously built Exec, another YC company; brother of Twitch's Justin Kan.

Funding & outcome

Raised a $12.5M Series A (Spark Capital), then was acquired by General Motors in March 2016 for roughly $1B and continued raising outside capital (Honda, SoftBank) under GM, reaching a reported ~$30B peak valuation with $7B+ in disclosed external investment on top of $10B+ GM itself poured in by 2024. Status: wound down. After an October 2023 pedestrian-dragging incident triggered a nationwide fleet suspension, GM announced in December 2024 it would stop funding Cruise's robotaxi development entirely, with ~1,000 additional layoffs in February 2025 as the service formally shut down.

Business model

Operated a per-ride driverless taxi service (fares like Uber/Lyft/Waymo) while functioning as a long-term R&D bet subsidized by GM. This made sense in theory — a per-ride fee matches how consumers already pay for point-to-point transport — but required absorbing enormous fixed costs (sensors, safety operators, mapping, regulatory compliance, insurance) per vehicle long before ride volume could amortize them, and GM's willingness to fund roughly $2B/year in cash burn without a clear profitability path was ultimately the deciding factor, not the technology.

Business type

B2C consumer transportation service wrapped inside a wholly-owned corporate subsidiary — GM was a strategic financial backer rather than a customer, and Cruise owned/operated its own fleet rather than matching independent drivers to riders, so it was never a marketplace in the Uber/Lyft sense.

The AI angle

Cruise's core stack — perception, prediction, planning for autonomous driving — is fundamentally an AI/ML problem, making it the most AI-native company on this list in one sense. But its failure is a useful counter-case to 'AI solves everything': the shutdown wasn't an AI capability gap so much as the compounding non-AI costs of operating a real-world, safety-critical, heavily-regulated fleet — incident liability, regulatory trust, insurance, and the sheer capital intensity of scaling city by city. AI was necessary but not sufficient; GM's decision was strategic and financial (robotaxis 'not GM's core business'), and Waymo's continued expansion in the same period suggests the technology itself was viable — just not for this parent's risk tolerance.

Developer-first customer data platform (CDP) collecting, unifying, and routing customer event data from a company's apps to hundreds of downstream analytics and marketing tools via a single API. Now operates as 'Twilio Segment.'

Founders

Calvin French-Owen, Ilya Volodarsky, Peter Reinhardt, and Ian Storm Taylor — MIT classmates who founded the company in 2011. Reinhardt served as CEO through the Twilio acquisition and later founded carbon-removal company Charm Industrial.

Funding & outcome

Launched with $600,000 in early funding and raised further rounds totaling in the tens of millions (a single confirmed cumulative figure wasn't found). Acquired by Twilio in November 2020 for approximately $3.2B in an all-stock deal. Now fully integrated as Twilio Segment; Twilio doesn't break out its financials separately.

Business model

Usage-based/tiered pricing on Monthly Tracked Users (MTU) — a generous free/cheap entry tier drives bottom-up developer adoption via a single low-friction integration, while usage-based scaling captures more revenue as a customer's own product grows and generates more event volume, aligning revenue with customer success rather than seat count, which doesn't map well to a data-infrastructure product.

Business type

Pure B2B, horizontal infrastructure/API-as-a-service. Buyer and user are the same (a company's engineering/marketing/data teams); there's no consumer-facing layer, and Segment is connective infrastructure rather than an intermediary between two external parties.

The AI angle

A meaningful but supporting lever, not a core disruptor. Unified, identity-resolved customer data is exactly the substrate AI-powered personalization and predictive scoring need, so Twilio has layered AI predictions and audience-activation features directly on top as a natural upsell. The more interesting long-term question is whether LLM-native 'agentic' pipelines that auto-map events and self-heal broken tracking could commoditize the integration-and-plumbing work that was Segment's original pitch — a real structural threat to a category whose core value was 'we did the tedious integration work for you.'

Cloud-based digital operations and incident-management platform that alerts, escalates, and coordinates engineers when production systems break, now expanded into AIOps and operations automation.

Founders

Alex Solomon (original CEO) — University of Waterloo; prior software engineer at Amazon during its shift to microservices, which directly inspired the on-call model. Andrew Miklas — Waterloo, prior Amazon/ATI internships, later became a YC General Partner. Baskar Puvanathasan — Waterloo, prior Amazon and BlackBerry. Founded 2009 in Toronto, relocated to SF for YC.

Funding & outcome

Raised roughly $174–180M across 5 rounds (a16z, Bessemer), including a $90M Series D in 2018 at ~$1.3B valuation. IPO'd on the NYSE April 2019 at $24/share, popping ~59% on debut to ~$2.8B. Current market cap roughly $0.65–0.85B — a steep decline from IPO-day levels. FY2025 revenue $467.5M (+8.5% YoY), growth decelerating to high single digits.

Business model

Per-seat SaaS subscriptions in tiers, plus usage/volume-based add-ons (Analytics, Event Intelligence, AIOps) — a hybrid seat-plus-usage model. Seat count is a natural proxy for how many engineers need to be reachable during incidents, while usage-based add-ons let PagerDuty capture more value from high-alert-volume enterprise accounts; switching costs (escalation policies, integrations, muscle memory) support retention even as growth slows.

Business type

B2B SaaS with platform/ecosystem dynamics — the buyer is always an engineering/IT organization, and hundreds of integrations (Slack, Datadog, AWS, Jira) make it an integration hub other tools plug into, not just a point solution.

The AI angle

An active, already-shipping lever, not speculation. PagerDuty AIOps (claims up to 91% alert-noise reduction) and PagerDuty Advance (a generative-AI assistant that summarizes incidents, drafts updates, and recommends or executes remediation with approval) are live products, sitting on a decade-plus proprietary dataset of real incidents and resolutions — a genuine data moat. The tension: sufficiently good AI auto-remediation could shrink the on-call headcount its per-seat pricing depends on, and the market's subdued valuation suggests investors aren't yet convinced AI reaccelerates growth against Datadog, ServiceNow, and incident.io.

Corporate credit card and spend-management platform, originally built for cash-rich startups underserved by traditional issuers, later expanded toward enterprise. No longer independent — Capital One completed its acquisition April 7, 2026.

Founders

Henrique Dubugras and Pedro Franceschi — Brazilian, met as teenagers around 2012. Their first company, Brazilian payments startup Pagar.me, was sold to StoneCo in 2016. Both were admitted to Stanford but dropped out after one semester to pursue what became Brex, joining YC's W17 batch.

Funding & outcome

Raised roughly $1.2–1.5B across 11+ rounds (Ribbit Capital, DST Global, Kleiner Perkins, Tiger Global, Greenoaks, plus Max Levchin and Peter Thiel individually). Valuation ran from $7.4B (2021) to a $12.3B peak (Jan 2022). Reached ~$700M annualized revenue by mid-2025, cash-flow positive from October 2025. Acquired by Capital One for $5.15B (~$2.75B cash + shares) — a steep discount to the 2022 peak — deal closed April 2026.

Business model

Three intertwined streams: interchange fees on card spend (~2.7% gross take, ~65% gross margin on that flow); SaaS subscription fees via the Brex Empower platform and Brex Premium (reportedly ~40% of gross profit); and net interest margin on Brex Cash balances. Interchange alone stopped being sufficient once Brex pivoted toward enterprise, so bundled software fees captured value from customers evaluating it against Concur/Ramp, while idle treasury cash added a third, largely margin-free lever independent of card usage.

Business type

B2B, with a two-sided/platform dimension — direct payers are always businesses, but Brex also sits between card networks/issuing banks and merchants paying interchange, plus a software layer integrating into accounting/ERP systems. Not B2B2C: end users are business employees, never individual consumers paying directly.

The AI angle

A real, ongoing product investment, not superficial narrative. ML-driven receipt matching, expense categorization, and fraud/anomaly detection have been live since 2023 (85–92% accuracy on obvious fraud flags per third-party review); 'Agents on Brex' automates policy compliance and exception-flagging. Capital One's own deal rationale explicitly frames Brex as an 'AI-native software platform' — spend management is fundamentally a document-classification and policy-enforcement problem well suited to agent automation, suggesting this was a genuine driver of the acquisition.

Global payroll and HR platform enabling companies to hire, pay, and manage employees and contractors in 150+ countries, acting as Employer of Record where needed so clients never need their own local legal entity.

Founders

Alex Bouaziz (CEO) — MIT civil/environmental engineering, where he met Wang; previously co-founded Lifeslice, a bootstrapped mobile video app run with international contractors, widely cited as Deel's direct inspiration. Shuo Wang — MIT mechanical engineering/robotics; previously co-founded and was CTO of Aeris, an air-purifier company acquired by iRobot for roughly $100M in 2021.

Funding & outcome

Raised roughly $982M across 11 rounds (a16z Series A, ICONIQ Growth at a $1.25B valuation in 2021, Coatue at $5.5B later that year, most recently a $300M Series E in Oct 2025 co-led by Ribbit Capital at a $17.3B valuation). Claims to have crossed $1B ARR in 2025 and three consecutive profitable years, not independently audited.

Business model

Recurring per-worker fees: a flat monthly EOR fee per employee, contractor management fees, global payroll processing fees, and FX/payment fees on cross-border payouts. The core pain isn't payroll software — it's that legally hiring abroad requires a local entity and constant labor-law compliance. Deel monetizes by absorbing that legal/compliance burden as a rented layer, charging recurring per-head fees cheap relative to standing up a foreign subsidiary; land-and-expand (start with a contractor, later convert to an EOR employee) scales pricing naturally with customer growth.

Business type

B2B with a B2B2C/three-sided structure — the paying customer is always the employer, but Deel simultaneously services the individual worker (running payroll, holding the contract, sometimes literally being their legal employer of record), making it structurally different from pure SaaS.

The AI angle

A genuinely active lever. 'AI Workforce' (launched 2025) ships pre-built agents for HR, payroll, and compliance built on the knowledge of Deel's 2,000+ in-country specialists, plus 'Akai,' an agentic workflow platform introduced in 2026. This is a strong fit because Deel's core moat and cost center is navigating fragmented, constantly-changing labor and tax law across 100+ jurisdictions — exactly the kind of domain where LLMs can compress work that would otherwise require thousands of local specialists. The Series E explicitly earmarked proceeds for AI-powered payroll/HR products.

API-first platform automating employment background checks — criminal records, driving records, identity screening — for employers and gig-economy/marketplace platforms.

Founders

Daniel Yanisse (CEO) — MS Microengineering, EPFL; prior software engineer at on-demand delivery startup Deliv, earlier at Mogreet and Cisco. Jonathan Perichon (CTO) — studied in France. Both met as engineers at Deliv, where background-check delays were a hiring bottleneck — the direct inspiration for founding Checkr in 2014, apparently each founder's first company.

Funding & outcome

Reported totals range $680–812M. Key rounds include a $100M Series C (2018, T. Rowe Price) and a $250M Series E (2021) at a $4.6B valuation with Fidelity, Franklin Templeton, and YC among the backers; reported around $5B by 2022 and still cited near there in 2025. Gross revenue exceeded $700M in 2024 and $800M in 2025. As of an April 2026 interview, leadership has not committed to a near-term IPO.

Business model

Per-check transactional fees tiered by depth (roughly $30–80/check), self-serve under ~300 checks/year and custom volume contracts above that, plus API/platform embedding into marketplaces and ATS/payroll software. This fits Checkr's market because core customers — gig platforms like Uber, DoorDash, and Instacart, plus high-turnover employers — need hiring decisions made in minutes at massive scale; usage-based fees scale naturally with signup volume without large upfront contracts.

Business type

B2B2C, transactional platform/infrastructure — the paying customer is the business, but the subject of each transaction (whose records are pulled and adjudicated) is a third party, the applicant, who never pays. Checkr also sits between employers and a fragmented web of courts and registries, giving it infrastructure characteristics beyond pure SaaS.

The AI angle

A substantive, already-productized lever ('Checkr AI'). The core challenge in background checks isn't accessing records but making sense of messy court data — inconsistent name spellings, aliases, non-standardized charge language across thousands of jurisdictions — which causes both false positives (legal liability) and false negatives. ML for record matching and a proprietary charge classifier reduce manual review time and speed same-day gig-platform onboarding. What AI doesn't solve: integrating with thousands of disparate, often non-digitized government systems remains a legal/data-partnership problem, not a modeling one.

'The organism company' — a horizontal foundry platform for programming cells, serving pharma, agriculture, industrial biotech, and food customers. Founded 2008, well before joining YC in 2014 as its first-ever synthetic biology investment — not a pre-seed team formed inside YC.

Founders

Five MIT-tied co-founders from 2008: Jason Kelly (CEO, chemical engineering/biology BS and biological engineering PhD, all MIT), Reshma Shetty (President/COO, MIT synthetic-biology PhD), Austin Che and Barry Canton (MIT PhD lab work), and Tom Knight, their MIT professor and mentor — an early ARPANET/Lisp-machine computer scientist who moved into biology and created MIT's Registry of Standard Biological Parts.

Funding & outcome

Raised roughly $803M pre-SPAC (General Atlantic, Baillie Gifford, Bill Gates' Cascade Investment). Merged with SPAC Soaring Eagle Acquisition Corp in 2021 at a $15B pre-money valuation (~3x its last private mark), including a $775M PIPE, trading NYSE: DNA from September 2021. Current market cap roughly $0.5B — down ~97% from the SPAC deal. Revenue fell 53% YoY in Q1 2024 alone, triggering ~35% workforce cuts targeting EBITDA breakeven by end of 2026.

Business model

Foundry-services revenue (upfront/recurring fees for engineering organisms, ~65% of revenue) plus downstream value share (royalties, milestones, or equity across 100+ active programs) — high-margin but unpredictable, since most biology programs fail or take years to resolve. A biosecurity/biosurveillance line was divested in April 2026. The model has struggled because synthetic-biology R&D is capital-intensive and long-duration; fixed-cost foundry infrastructure runs expensive regardless of program success rate.

Business type

B2B R&D-as-a-service / shared infrastructure platform — customers pay for access to cell-programming capability rather than a fixed product, and payment contingent on uncertain multi-year biological outcomes makes it R&D-as-a-service rather than conventional SaaS.

The AI angle

Ginkgo has pivoted hard into 'AI x biology' positioning, with real substance alongside clear repositioning pressure from the stock collapse. Substance: Ginkgo Datapoints sells large-scale biological datasets for training AI foundation models, leveraging years of proprietary experimental data most AI labs lack; a proprietary protein language model and an autonomous lab platform partnered with OpenAI. Skepticism: this pivot arrives directly after a ~97% valuation collapse and major workforce cuts, with 2026 revenue still down sharply YoY — reading partly as genuine monetization of real assets, and partly as a survival pivot for a foundry-plus-royalty thesis that hasn't proven it can reach profitability alone.

Hosted, developer-friendly search-and-discovery API — indexing, ranking, typo-tolerance, and increasingly AI/RAG-based search — that businesses embed into their own products. Founded 2012 in Paris; ~800+ employees, 17,000+ paying customers.

Founders

Nicolas Dessaigne (CEO until 2020, now a YC Group Partner) — CS PhD; previously CTO at Arisem (acquired by Thales) and VP R&D at French search engine Exalead. Julien Lemoine (now an advisor) — engineering degree from EPITA; prior CTO of MASA Group. Both are French search-engine veterans from the same Arisem/search-engine lineage.

Funding & outcome

Sources disagree, roughly $221–335M lifetime (Accel-led Series A/B, a $150M Series D in 2021 with Fidelity, Salesforce Ventures, and Twilio). Private, last valuation $2.25B (2021), with estimated ARR around $210–250M by 2024–25 (unaudited third-party estimates).

Business model

Usage-based, tiered pricing on search requests and records indexed — free tier, metered pay-as-you-go, and enterprise (~$50K+/year) with SLAs and merchandising features. This fits Algolia because customer search traffic scales directly with the customer's own product success, so revenue grows in lockstep, and per-API-call billing matches how search is actually consumed — a stateless infrastructure call, not a per-seat tool.

Business type

B2B / vertical infrastructure platform with a B2B2C flavor — payers are businesses embedding search into their products, but the actual search experience is consumed by those businesses' own end users, with whom Algolia has no direct relationship.

The AI angle

A central, active battleground, not a bolt-on. Algolia has repositioned as an 'AI search and retrieval platform' — hybrid semantic+keyword ranking, generative shopping assistants, and an 'Agent Studio' aimed at agentic commerce (AI shopping agents querying catalogs on a user's behalf). Retrieval quality is now table stakes for RAG-based AI products, so owning fast, relevance-tuned retrieval infrastructure extends Algolia's moat into 'the API AI agents call to search your catalog' — but it's also a real existential risk, since vector-native competitors or foundation-model providers building native retrieval could disintermediate Algolia if it doesn't keep pace.

Digital/product analytics platform letting software teams track and analyze in-app user behavior — funnels, retention, cohorts, session replay, experimentation — to drive product decisions. Pivoted from a 2012 voice-messaging startup, Sonalight.

Founders

Spenser Skates (CEO) — MIT bioengineering/EECS coursework, MIT Battlecode team; worked as an algorithmic trader at DRW Trading before co-founding. Curtis Liu (CTO) — MIT EECS/Math, two-time MIT Battlecode winner; prior experience at Electronic Arts and Google. Both met at MIT, built Sonalight first, then pivoted to Amplitude in 2014 after struggling to find good analytics tools for their own app.

Funding & outcome

Raised roughly $311–336M (Battery Ventures, Sequoia, Benchmark, IVP). Last private round: a $150M Series F led by Sequoia (2021) at a ~$4B valuation. Went public via direct listing on Nasdaq Sept 2021, closing day one near a $7.1B implied valuation. Current market cap roughly $1.6B — down ~75–80% from IPO-day, reflecting broader SaaS/analytics multiple compression since 2021–22.

Business model

Usage-based B2B SaaS priced on Monthly Tracked Users and events per MTU, from a free Starter tier to enterprise contracts reaching $250K+/year. This fits the market because product-analytics value scales with the size and complexity of the product being analyzed, not with seats — a small startup's app costs little to serve and generates modest value, while a 10M-user consumer app generates massive data volume and correspondingly massive decision-making value, so revenue tracks both cost-to-serve and realized value and expands naturally as customers grow.

Business type

B2B SaaS / vertical software platform. Amplitude sells to product and growth teams at other companies; it has no relationship with the downstream consumers whose behavior is tracked, making it pure B2B infrastructure rather than B2B2C.

The AI angle

A genuinely active lever, not speculation. Amplitude has shipped agentic AI analytics since early 2026 — a Global Agent plus specialized agents that autonomously investigate metric changes and build dashboards, and an AI Plugin exposing 25+ skills to external tools via MCP. This attacks Amplitude's core historical bottleneck: most of its data has always gone unexplored by non-analyst product managers, and an AI layer that can query and explain data in natural language could expand the addressable buyer from 'the data team' to 'every PM.' The moat depends on proprietary access to a company's longitudinal behavioral data, but Mixpanel, PostHog, and BI tools are racing to ship similar features, so defensibility comes down to execution, not the AI capability alone.

Mobile-first Indian brokerage/investment app offering stock trading, F&O, mutual funds, US stocks, and digital gold. Founded 2016, launched 2017; became India's largest stockbroker by active NSE client count.

Founders

All four founders are ex-Flipkart: Lalit Keshre (CEO, IIT Bombay), Harsh Jain (COO, IIT Delhi, MBA UCLA Anderson), Neeraj Singh (CTO, built Flipkart's returns/refund system), and Ishan Bansal (CFO, CFA charterholder, earlier at Naspers).

Funding & outcome

Raised roughly $600M across ~12 rounds (Tiger Global, Peak XV, Ribbit Capital, GIC). Crossed unicorn status in 2021; a mid-2025 round valued it at ~$7B. IPO'd on the NSE/BSE November 2025, raising ~$748M, oversubscribed ~18x, closing day one up ~29%. Current market cap roughly $14B — roughly double the IPO price. FY25 revenue ~$440M, net profit ~$206M.

Business model

'Zero-commission' delivery trading as a low-cost acquisition hook, monetized on the back end via flat-fee brokerage on intraday/F&O trades, trail commissions from mutual-fund distribution, direct fund-manager economics via its own AMC, and Groww Pay (a licensed UPI operator). This fits India's retail market because delivery-equity trading is intensely price-sensitive, while trading volume is disproportionately concentrated in F&O speculation — which is where Groww actually earns revenue — and a large, growing SIP/mutual-fund culture provides an annuity-like stream that offsets that volatility.

Business type

B2C with platform/B2B2C elements — a consumer-facing app for millions of individual investors that's also an intermediary between AMCs (supply) and retail investors (demand) on the mutual-fund side, earning distribution commissions. Competes with Zerodha, Angel One, and Upstox in a market SEBI has been actively tightening around retail F&O speculation.

The AI angle

A real but double-edged lever. Upside: Groww sits on large proprietary trading and payment datasets well suited to AI-driven credit underwriting for its nascent lending vertical and fraud/market-manipulation detection, increasingly important for a regulated broker. The pushback: AI-personalized 'trading nudges' that increase trading frequency are a regulatory and reputational liability rather than a growth lever right now, given SEBI's explicit campaign against retail F&O overtrading — AI is strong for risk/credit/support cost, genuinely risky for engagement-maximizing personalization in the current policy climate.

Indian e-commerce marketplace connecting small sellers/manufacturers with value-conscious consumers, originally a WhatsApp/social reseller catalog platform, now primarily a direct-to-consumer app plus its own logistics arm, Valmo. Founded December 2015 as Fashnear Technologies.

Founders

Vidit Aatrey (CEO) and Sanjeev Barnwal (CTO) — both IIT Delhi (electrical engineering and computer science respectively), both worked at InMobi and Samsung Research Institute before quitting mid-2015 to apply to YC S16 with ~300 users.

Funding & outcome

Raised roughly $1.61B across 11 rounds, including a $25M corporate round from Meta and a $570M Series F at a $4.9B valuation (2021) backed by SoftBank and Prosus. A 2024 markdown implied ~$3.5B; the last private mark (Jan 2025) was $3.9B. IPO'd on the NSE/BSE December 2025, raising ~$606M, opening up ~46%. Current market cap roughly $10.2B, though several investors have been trimming stakes through 2026.

Business model

Evolved from a 0%-visible-commission reseller model to a commission-light marketplace monetizing mainly through seller logistics fees, advertising/sponsored listings, and Valmo, its own logistics-as-a-service network. A heavy percentage-of-GMV commission (Amazon/Flipkart-style) would push out the small unbranded manufacturers who give Meesho its cheap, long-tail catalog — monetizing logistics and ads instead keeps prices low for price-first shoppers while capturing margin from sellers who need reach, mirroring Pinduoduo's playbook in lower-tier, price-sensitive markets.

Business type

Two-sided marketplace/platform with a B2B2C logistics layer and vestigial C2C 'social reselling' roots — sellers (700,000+ annual transacting) are the paying supply side, while 234M+ transacting users shop largely for free, a classic pattern where sellers fund the platform. Deliberately targets Tier 2/3/4 India, distinct from Amazon/Flipkart's urban playbook.

The AI angle

A real, currently active lever. AI-powered customer support handles roughly 60,000 calls/day. The strongest opportunities map directly onto Meesho's structure: catalog intelligence across 700,000+ often unsophisticated sellers, vernacular commerce (LLM-driven search/chat/voice shopping across dozens of Indian languages for a Tier 2-4 base), recommendation/dynamic pricing given razor-thin average order values, and — arguably highest-leverage — logistics optimization via Valmo, attacking Meesho's largest cost line, which is also now a direct revenue line.

Rents and sells heavy construction equipment nationwide, and sells its own T3 telematics/fleet-management software both to its own rental customers and third parties who own equipment outright. Pitched as a peer-to-peer rental marketplace at a 2014 Startup Weekend before joining YC.

Founders

Jabbok Schlacks (CEO) and Willy Schlacks (President) — brothers with 25+ years in construction via the family firm Schlacks Construction. Additional co-founders Jeff Lowe, Matthew McDonald, and Brad Siegler are named in company records; their prior backgrounds weren't independently verifiable.

Funding & outcome

Raised roughly $806M in equity across ~11 rounds (BDT & MSD Partners led a ~$440M round at a $3.75B valuation), plus large debt facilities financing its owned fleet, including a $2.75B senior secured revolver with Wells Fargo. IPO'd on Nasdaq January 2026, raising ~$747M, closing day one near $32.56. Current market cap roughly $5B. FY2025 revenue $4.38B (+16% YoY), net income $40M.

Business model

Three streams: equipment rental (~62% of revenue), equipment sales/financing (~35%), and T3 telematics SaaS (~1.5%, but more than doubled YoY), sold bundled to renters and standalone to third-party fleet owners. This hybrid fits a fragmented, low-digitization industry where most contractors can't justify owning a full fleet, so rental dominates, but utilization and maintenance visibility across mixed-brand fleets is chronically poor. Owning both equipment and software lets EquipmentShare capture data exhaust from its own fleet to sell insights back to renters, while the rental business funds and de-risks the software build — a flywheel neither a pure-SaaS telematics startup nor a pure-rental competitor can easily replicate.

Business type

Asset-heavy B2B rental business with an embedded B2B SaaS layer — most revenue and nearly all of the capital structure is a traditional, capital-intensive rental business serving commercial contractors; T3 is a genuine but currently minority-revenue software layer on the same base.

The AI angle

A genuinely strong, non-generic case, precisely because EquipmentShare already owns the two ingredients AI needs most — physical assets at scale and the T3 telematics pipeline — rather than having to acquire either. Clearest levers: predictive maintenance across hundreds of thousands of tracked assets (hitting rental margins directly), fleet allocation/utilization optimization across 430+ locations (idle equipment is pure capital drag), demand forecasting tied to construction-starts data, and jobsite computer vision for safety and theft detection. Because ROI is easy to measure against a multi-billion-dollar owned asset base, this is one of the more credible AI opportunities among asset-heavy industrial businesses — though near-term impact will likely show up as better fleet capital efficiency rather than a new blockbuster software revenue line.

Builds a spatial-data platform that turns physical spaces into digital twins — customers capture buildings with Matterport's cameras or smartphones, and cloud software generates interactive 3D walkthroughs and floor plans, used mainly in real estate, plus construction, insurance, and retail.

Founders

Matt Bell — previously founded Reactrix (interactive floor-projection displays) and led its computer-vision team, earlier at Google research. David Gausebeck (CTO) — early PayPal fraud-detection engineer, co-created the first commercial CAPTCHA implementation. Michael Beebe (COO) — previously managed a group at SRI International building 3D models of outdoor environments for DARPA.

Funding & outcome

Best-supported figure is roughly $400M raised pre-IPO, including a $48M Series D in 2019. Went public via SPAC merger with Gores Holdings VI, trading Nasdaq: MTTR from July 2021. Acquired by CoStar Group, announced April 2024 at $5.50/share (~$1.6B enterprise value, a 212% premium), closed February 2025, delisted from Nasdaq.

Business model

Hybrid hardware plus SaaS: sells its own 3D cameras (~$6,000) plus tiered hosting subscriptions ($10–$300+/mo) priced by active spaces, plus AI-feature add-ons. FY2023 revenue was $157.7M, $99.6M of it subscription across 938,000 subscribers. Real estate, construction, and insurance customers capture an asset once but reuse the digital record repeatedly across a listing's life — that reuse is what justifies a recurring hosting fee rather than a one-time capture charge.

Business type

B2B2C with platform elements — paying customers are businesses (brokerages, contractors, insurers), but the digital twins are consumed by their end customers (home buyers, tenants, claimants); MLS/portal integrations give it a modest platform dimension without being a transactional marketplace.

The AI angle

AI is Matterport's clearest lever for turning a capture/hosting business into a data/insights business: automated floor-plan generation, measurement extraction, condition/damage detection for insurance, and generative redesign all monetize a proprietary 33B+ sq ft spatial-data corpus as premium upsells. The honest counter: auto floor plans and measurements are becoming table stakes, and as smartphone LiDAR and foundation models get better at 3D reconstruction from ordinary photos, the exclusivity of owning capture hardware erodes — the durable moat is the historical data corpus and workflow lock-in, not the AI features themselves.

Built cloud/mobile construction productivity software — blueprint storage, markup, RFIs, punch lists, field reports — replacing paper blueprints on job sites. Now part of Autodesk Construction Cloud/Build.

Founders

Tracy Young (CEO) — construction engineering management degree, CSU Sacramento; worked as a construction engineer on Bay Area hospitals before founding. Ralph Gootee (CTO) — previously a rendering engineer at Pixar (Cars 2, Toy Story 3, Brave), applied math master's from Johns Hopkins. Ryan Sutton-Gee, Antoine Hersen (ex-high-frequency-trading engineer), and Kenny Stone round out the founding team.

Funding & outcome

Rounds included a $1.1M seed (2012, with Box and Alexis Ohanian), an $18M Series A (2014, Sequoia), and a $40M Series B (2015) — total figures conflict across sources, roughly $69–109M. Acquired by Autodesk in November 2018 for $875M, net of cash, closed December 2018.

Business model

Per-seat, tiered SaaS priced by sheet volume (now roughly $140/user/month unlimited under Autodesk Build). Construction projects involve a general contractor plus dozens of subs, architects, and owners all needing the current drawing set — version chaos was the core costly failure PlanGrid solved, and sheet-volume-based pricing tracks project complexity, with the GC typically paying and subsidizing adoption across the project's stakeholder graph.

Business type

B2B with multi-sided platform dynamics — commercially a straightforward B2B SaaS sale, but within each project it functions as a coordination platform where the GC, subs, and architects must all onboard for the tool to deliver value, creating project-level network effects.

The AI angle

Real, substantive use cases beyond generic copilot framing: automated plan/change-detection between drawing revisions, computer vision on jobsite photos (progress, safety violations, installed-vs-specified discrepancies), automated code/compliance checking, and natural-language search over drawing sets — all high-leverage because rework costs the industry tens of billions annually and construction generates enormous unstructured visual and document data. Autodesk has shipped real applied-AI features since the acquisition (Construction IQ, Photo Autotags, automated spec handling), suggesting genuine workflow investment rather than a chat veneer.

Consumer app that finds and cancels unwanted subscriptions, negotiates bills, and tracks budgets by linking to bank accounts. Launched as Truebill, rebranded Rocket Money in August 2022.

Founders

Haroon Mokhtarzada (CEO) — University of Maryland economics, Harvard Law JD; previously co-founded Webs.com (2001), acquired by Vistaprint for $117.5M in 2011. Yahya Mokhtarzada — initial CEO, later CRO. Idris Mokhtarzada — University of Maryland CS, also a Webs.com co-founder. The brothers are sons of Afghan/Turkish refugee parents.

Funding & outcome

Raised roughly $85M pre-acquisition (an estimate; Bessemer- and Accel-led rounds valued the company at ~$530M pre-acquisition). Acquired by Rocket Companies in December 2021 for $1.275B in cash — implied roughly 12–13x its ~$100M 2021 ARR.

Business model

Freemium: a free tier (subscription detection, basic budgeting) funds a pay-what-you-think-is-fair Premium tier plus a fixed-price Premium+, and a bill-negotiation success fee (35–60% of first-year savings, $0 if unsuccessful). The value proposition is friction/dread-aversion, not information asymmetry — users could cancel subscriptions or negotiate bills themselves, but success-fee and pay-what-you-want pricing align price with already-realized value rather than an upfront speculative payment.

Business type

B2C, now functioning as B2B2C/platform-embedded within Rocket Companies — a high-frequency financial-engagement product layered onto Rocket's low-frequency, high-ticket products (mortgages, personal loans) as a funnel and cross-sell mechanism.

The AI angle

Core plumbing (subscription detection/categorization from transaction data, automated savings transfers) already relies on ML. The clearest AI leverage point is the labor-intensive bill-negotiation service, which still relies on human agents calling providers — voice-capable LLM agents that can navigate IVR trees and negotiate could cut cost-to-serve dramatically, letting Rocket Money lower its success fee or expand margin. Caveat: none of this needs frontier models, and negotiation calls carry compliance/authorization sensitivities that will slow full automation regardless of model capability.

Full-stack superconducting quantum computing company that designs and fabricates its own qubit chips at its Fremont facility, selling cloud quantum-computing access and direct hardware sales to government and national labs.

Founders

Chad Rigetti — sole verified founder (2013); Canadian physicist, BS Physics (University of Regina), Yale applied physics PhD under Michel Devoret, then ~3 years at IBM's quantum labs before founding the company. He stepped down as CEO in 2022 and later founded a separate company, Sygaldry Technologies, focused on quantum-accelerated AI.

Funding & outcome

Sources diverge sharply: named VC rounds total roughly $198M, while some aggregators cite ~$658M including later strategic rounds — an unreconciled range. SPAC merger with Supernova Partners Acquisition Company II closed March 2022, trading Nasdaq: RGTI at a pro forma ~$1.5B equity value. Current market cap roughly $6.28B, highly volatile ($4.7–8.6B across 2026). FY2025 revenue was just $7.1M against a ~$238.7M net loss — valuation driven almost entirely by sentiment, not fundamentals.

Business model

Revenue is currently dominated by government technology-development contracts (DARPA, AFRL, DOE, a $100M Commerce Department LOI in 2026), plus cloud quantum-computing access and direct hardware sales. This mix makes sense because quantum computing is still in the pre-commercial-advantage era — no proven quantum advantage exists yet, so agencies fund capability-development milestones rather than a product with demonstrated ROI, while cloud access mainly seeds a future research ecosystem rather than generating real recurring revenue.

Business type

B2G hard tech, with an emerging cloud-access platform layer — the dominant customer base is government/national-lab funding; fundamentally a capital-intensive chip manufacturer, not a software company.

The AI angle

AI/ML genuinely helps operationally — a verified partnership with Quantum Machines and NVIDIA on AI-assisted qubit calibration (reinforcement-learning tuning of control pulses) achieves real gate-fidelity gains in an area where ML outperforms manual tuning. But AI does not touch Rigetti's actual bottlenecks — materials science, fabrication, and cryogenic noise — and 'quantum plus AI convergence' is largely market narrative rather than current product reality. Tellingly, Chad Rigetti pursued that convergence in a separate new company rather than inside Rigetti Computing itself, suggesting even insiders treat it as speculative, not yet delivered by today's hardware.

Designs, builds, and plans to own and operate compact sodium-cooled, metal-fueled fast fission 'powerhouses' up to 75 MWe, plus adjacent fuel-recycling and medical-isotope businesses. Founded 2013.

Founders

Jacob DeWitte (CEO) and Caroline Cochran (COO) — married MIT nuclear-engineering graduates who met as teaching assistants in 2009. Sam Altman has chaired Oklo's board since 2015, recruited it into YC's first hard-tech batch, led its 2015 seed round, and later co-founded the SPAC that took it public.

Funding & outcome

Roughly $60M raised in pre-IPO private funding (larger aggregator totals blend in post-IPO capital raises). SPAC merger with Altman's AltC Acquisition Corp, announced 2023 at an $850M pre-money value, closed May 2024; stock fell 54% on debut. Since IPO, Oklo has raised roughly $1.88B gross via at-the-market equity programs. Current market cap roughly $8.3–8.6B, extremely volatile (52-week range $36.61–$193.84) — unprofitable, trading on an AI/power-demand narrative rather than fundamentals.

Business model

Build-own-operate independent power producer: Oklo doesn't sell reactors, it owns and operates each plant and sells electricity under 20-year power purchase agreements. This fits the environment because nuclear construction is extremely capital-intensive and regulated in ways most customers can't manage themselves — Oklo absorbs the licensing/construction risk, customers simply buy power, and Oklo's incentives align with actually finishing and licensing plants since it only earns once they run. Effectively pre-revenue: first plant broke ground September 2025, still under construction, with H1 2026 revenue of just $1.21M against an $81.6M net loss.

Business type

Merchant generator / capital-intensive industrial infrastructure, jointly B2B and B2G — the commercial pipeline (Meta, Switch, Equinix) is B2B, while first-reactor construction and licensing runs through a B2G channel with the Department of Energy as regulator, site landlord, and development partner simultaneously.

The AI angle

AI is Oklo's demand driver, not an internal capability — and this is well-documented, not just hype. Oklo's own investor materials explicitly tie its signed pipeline (which grew to 14+ GW by early 2025) to AI/data-center power growth, with concrete deals including Switch (a 12 GW master agreement tied to AI/cloud data centers) and Meta (a 2026 prepayment for a 1.2GW Ohio AI campus). On the supply side, however, AI does not meaningfully change reactor physics, fuel qualification, or regulatory timelines — Oklo's original 2020 NRC license application was denied without prejudice in 2022 for insufficient information, and its current path runs through paper-heavy authorization gates no software sophistication accelerates. AI created Oklo's customer base and stock premium; the company itself remains a slow, heavily-regulated hardware business whose fundamental constraints are untouched by the AI boom filling its sales pipeline.