Every year produces a few companies that make people stop scrolling. This year produced thirty. We went looking for startups that are not just well funded, but genuinely changing how their category works, and we pulled the real numbers behind each one: who founded it, how much money it has raised, what it actually earns, and why it matters. Every figure here is sourced and dated, because a startup list full of made up numbers is worse than no list at all.
We grouped them into seven categories so the scale of what is happening in each corner of the industry is easier to see at once.
AI Coding and Agent Tools
Lovable
Lovable lets someone describe an app in plain language and get real, working software back, a category people now call vibe coding. It was founded in Stockholm in 2023 by Anton Osika, a physicist turned entrepreneur, and Fabian Hedin, after an open source coding tool Osika built in his spare time went viral on GitHub. Lovable raised $8 million in seed funding in October 2023, then moved fast: $15 million in a pre-Series A in February 2025, $200 million in a Series A that July at a $1.8 billion valuation, and $330 million in a Series B that December at $6.6 billion. By February 2026 it was on a $400 million annual revenue run rate, and the round made both founders billionaires. What makes it mind blowing is the speed. Lovable became the fastest growing software startup in history, built on subscription revenue from people who have never written a line of code.
Cognition
Cognition built Devin, marketed as the first true AI software engineer, an agent that can read a codebase, plan the work, write the code, and open a pull request on its own. It was founded in November 2023 by Scott Wu, a three time gold medalist at the International Olympiad in Informatics, along with Steven Hao and Walden Yan. Funding moved from a $175 million round at a $2 billion valuation in April 2024, to $400 million at $10.2 billion in September 2025, to $1 billion at $26 billion in May 2026, and the company is reportedly already in talks for a $40 billion valuation. Cognition makes money by selling Devin as a subscription and enterprise product to engineering teams. Its own team has said around 90 percent of Cognition’s internal code is now written by its own AI.
Replit
Replit started as a browser based coding environment and became one of the fastest growing AI app builders on the internet. Founder and CEO Amjad Masad, born in Jordan, raised $400 million in 2026 at a $9 billion valuation, tripling the company’s worth in six months and making Masad a billionaire. Replit now has 50 million users and $240 million in annual recurring revenue, with a public goal of reaching $1 billion by the end of 2026. Its business model is straightforward subscription and usage based pricing for individuals and teams building software with AI assistance, and its investor list ranges from Andreessen Horowitz to athlete Shaquille O’Neal.
Genspark
Genspark builds an AI powered workspace, sometimes described as a super agent, that can research, write, and complete multi step tasks in one place. It was founded by Eric Jing, a former founding member of Microsoft Bing, and has raised $645 million across six rounds. Genspark crossed $100 million in annual recurring revenue within nine months of launch and reached $250 million in ARR at a $2.6 billion valuation in 2026. It makes money the same way most of this category does, through subscriptions to a product that replaces several other tools at once, which is exactly why growth this fast is possible.
The Race for AGI and Frontier Models
Safe Superintelligence
Safe Superintelligence, known as SSI, might be the strangest startup on this list, because it has no product, no revenue, and roughly 50 employees, yet it is valued at $32 billion. It was founded in June 2024 by Ilya Sutskever, OpenAI’s former chief scientist, along with Daniel Gross and Daniel Levy. SSI raised $1 billion at founding, $2 billion in February 2025 at $32 billion, and in July 2026 Nvidia put in a $5 billion equity investment, bringing total funding to around $8 billion. The company has publicly committed to building nothing commercial until its safety mission is complete, which makes it the rare startup being funded almost entirely on reputation and conviction rather than a business model at all.
Thinking Machines Lab
Founded in February 2025 by former OpenAI CTO Mira Murati and a group of OpenAI alumni, Thinking Machines Lab raised the largest seed round in history, $2 billion at a $12 billion valuation, in July 2025. Talks to raise another round at $50 billion collapsed in early 2026, but the company kept building, shipping its first product, Tinker, in October 2025, a tool that lets developers fine tune AI models without the usual cost of distributed training. By late 2026 it had two shipped products, infrastructure deals with the two largest AI cloud providers, and its team intact. Revenue comes from developers and companies paying to customize models through its platform.
Mistral AI
Mistral is Europe’s answer to the American AI labs, founded in 2023 in Paris by Arthur Mensch, Guillaume Lample, and Timothee Lacroix, all former researchers at Meta and Google DeepMind. The company has raised more than $6 billion in total, including a $3.5 billion round in mid 2026 that pushed its valuation as high as $20 billion depending on the source. Revenue reached roughly $400 million on an annualized basis in 2026, up from $100 million the year before. Mistral makes money licensing its language models to enterprises and governments who specifically want an alternative to US based AI providers, which has made it central to Europe’s push for what it calls AI sovereignty.
Cohere
Cohere was founded by Aidan Gomez, one of the co-authors of the original transformer research paper that made modern AI language models possible. The company has raised $1.6 billion, reaching a $7 billion valuation in 2025, and its April 2026 acquisition of German AI company Aleph Alpha created a combined business valued at $20 billion. Cohere’s 2026 revenue is estimated around $240 million, up from $100 million the prior year. Its business is entirely enterprise focused, selling secure, private language models to companies and governments that will not send their data to a public AI provider, a deliberately narrower and more defensible niche than chasing consumer users.
Robotics and the Physical World
Physical Intelligence
Physical Intelligence is trying to build one AI model that can control almost any robot for almost any task, rather than training a separate model for every machine. It was founded by Karol Hausman, formerly of Google DeepMind, alongside Sergey Levine, Chelsea Finn, and several others from the same research world. The company raised $600 million at a $5.6 billion valuation, and by mid 2026 was in talks to raise roughly $1 billion more at over $11 billion, doubling its valuation in about four months. Backers include Jeff Bezos personally. As a two year old company with no mass market product yet, its business model is still forming, but the pace of investor conviction is itself the story.
Skild AI
Skild AI is building what it calls the Skild Brain, a single foundation model meant to control any robot body without custom retraining for each one. It was founded by former Carnegie Mellon professors Deepak Pathak and Abhinav Gupta. In January 2026, Skild closed a $1.4 billion Series C led by SoftBank, with Nvidia, Jeff Bezos, Samsung, LG, and Salesforce all participating, valuing the Pittsburgh based company at more than $14 billion, over triple what it was worth seven months earlier. Skild went from zero to roughly $30 million in revenue in a matter of months in 2025 by deploying its models in security patrol, warehouse, delivery, and construction robots, an unusually fast path to real revenue for a foundation model company.
World Labs
World Labs is trying to build what its founder calls spatial intelligence, AI that understands and can generate real three dimensional environments rather than flat images or text. It was founded by Fei-Fei Li, the Stanford professor widely credited with helping launch the modern deep learning era through her ImageNet dataset. World Labs raised $1 billion in February 2026 at a $5 billion valuation, backed by Nvidia, AMD, and Autodesk, bringing its total raised to $1.23 billion since emerging from stealth in 2024. Its first product, Marble, generates editable 3D environments from text, images, or video, sold through free and paid tiers, with an eye toward future use in robotics and virtual and augmented reality.
Atomarine
Atomarine has one of the strangest pitches on this entire list: floating data centers, built on barges at sea, cooled directly by seawater to solve the AI industry’s growing compute and power shortage. It was co-founded by a team of MIT graduates spanning computer science, naval engineering, and nuclear engineering. The company plans a gas powered pilot by 2028 before transitioning to floating nuclear power ships by 2032, and it claims to have already secured more than $4 billion in customer interest through signed letters of intent, before a single barge has launched. Its eventual business model is renting out compute capacity, the same as any cloud provider, just with a genuinely unusual way of generating the power behind it.
Creative AI and New Hardware
HeyGen
HeyGen makes AI generated video avatars, letting a business turn a script into a realistic talking video without ever filming a person. It was founded in 2020 by Joshua Xu and Wayne Liang. The company has raised a comparatively modest $74 million total, including a $60 million round in 2024 at a $500 million valuation, but its revenue growth is the real headline. HeyGen surpassed $200 million in annual recurring revenue in June 2026, doubling in eight months, up from $57.5 million at the end of 2024, and it did this while staying close to cash flow break even, having burned only $25 million of the money it raised. It makes money through subscriptions for marketing teams, training departments, and content creators.
Suno
Suno lets anyone generate a full, produced song from a text description, and it has grown into a real business despite ongoing copyright lawsuits from major music labels. It was founded in Cambridge, Massachusetts in 2022 by Mikey Shulman, Georg Kucsko, and Martin Camacho, who met while working together at fintech company Kensho. Suno has raised $775 million total, including a $400 million round in June 2026 at a $5.4 billion valuation. It passed 2 million paid subscribers and $300 million in annual recurring revenue in February 2026. Its business model is a straightforward consumer subscription, which makes it one of the few AI companies on this list selling directly to individual hobbyists rather than businesses.
Black Forest Labs
Black Forest Labs builds the FLUX image generation models and was founded in Freiburg, Germany in August 2024 by Robin Rombach, Andreas Blattmann, Patrick Esser, and Dominik Lorenz, the researchers who originally built the latent diffusion technology behind Stable Diffusion while at Stability AI. After Stability AI ran into financial trouble, this entire research team left and started over. Black Forest Labs raised $300 million in a Series B at a $3.25 billion valuation, backed by Salesforce Ventures, Nvidia, and Andreessen Horowitz among others. It makes money licensing its image models to businesses and developers, and stands out as a case of an entire team rebuilding a company almost from scratch after their previous employer stumbled, and doing it successfully.
Pocket is a small AI hardware device that listens and takes notes for you, and it emerged from Y Combinator’s Winter 2026 batch as the clear revenue outlier of the entire cohort. The device shipped more than 30,000 units and reached $27 million in annual recurring revenue within five months of launch, an unusually fast path to real hardware revenue for a company this young. Pocket makes money the old fashioned way for a hardware company, selling the physical device plus likely a subscription layer for its AI features, proving that in a market obsessed with pure software AI, a genuinely useful physical product can still break out fast.
Industry-Specific AI That’s Quietly Taking Over
Legora
Legora is a legal AI platform built in Stockholm in 2023 by Max Junestrand, a 26 year old entrepreneur named to Forbes’ 30 Under 30 AI list, along with co-founder Sigge Labor. The platform automates document review, contract drafting, and legal research for law firms. Legora raised $550 million in a Series D in March 2026 at a $5.55 billion valuation, tripling its worth from just months before, with Nvidia making its first ever investment in a legal AI startup as part of the round. By August 2026, Legora was already in talks for a $10 billion valuation. It sells subscription access to law firms, directly competing with fellow legal AI unicorn Harvey in what has quickly become one of the most heavily funded professional services categories in AI.
Rogo
Rogo is building what it calls Wall Street’s first AI analyst, an agent platform that can generate investment memos, financial models, and diligence documents for investment banks. It was founded in New York in 2021 by Gabriel Stengel, John Willett, and Tumas Rackaitis, who developed the earliest version of the product while studying together at Princeton. Rogo raised $160 million in a Series D in April 2026 led by Kleiner Perkins, bringing its total funding to more than $300 million, with backing from Sequoia, Thrive Capital, and J.P. Morgan’s own growth equity arm. It sells directly to financial institutions, with more than 100 customers already including major names like Lazard, Moelis, and Nomura.
EliseAI
EliseAI builds conversational AI that automates communication for the housing and healthcare industries, handling everything from apartment leasing inquiries to healthcare scheduling. It was founded in 2017 by Minna Song and Tony Stoyanov, who met as undergraduates at Cambridge. The company has raised $392 million, most recently a $250 million Series E in 2025, and by August 2026 was in talks to raise $300 million more at a $3.7 billion valuation. EliseAI passed $100 million in annual recurring revenue in early 2025 after expanding beyond real estate into healthcare, and its software now runs on top of a straightforward SaaS model serving over 350 enterprise customers, including 70 percent of the fifty largest residential rental operators in the United States.
Applied Intuition
Applied Intuition sells software that automakers, robotics companies, and defense contractors use to develop and test autonomous vehicle systems. It was founded in 2017 by CEO Qasar Younis and CTO Peter Ludwig. The company raised $250 million at a $6 billion valuation in March 2024, then $600 million at a $15 billion valuation the following year, more than doubling its worth in twelve months, with growing expansion into defense contracts alongside its original automotive business. With 1,571 employees as of mid 2026, Applied Intuition makes money selling simulation and testing software licenses directly to the automakers and government agencies building the next generation of autonomous systems, a business model that profits from the industry’s caution rather than its recklessness.
Isomorphic Labs
Isomorphic Labs is trying to use AI to design new medicines, spun out of Google DeepMind in 2021 by Demis Hassabis, the DeepMind co-founder who won the 2024 Nobel Prize in Chemistry for AlphaFold, the AI system that solved the decades old problem of predicting protein structures. Isomorphic raised $600 million in March 2025 and $2.1 billion in May 2026, bringing total outside funding to $2.7 billion, with backers including Thrive Capital, Alphabet, and sovereign wealth funds from Abu Dhabi and Singapore. It has already partnered with Eli Lilly and Novartis on oncology and immunology drug design, and is targeting human clinical trials for its first AI designed medicine by the end of 2026. Its business model is licensing and co-development deals with major pharmaceutical companies, betting that AI can meaningfully shorten how long it takes a new drug to reach a patient.
Energy and Manufacturing for the AI Boom
Hadrian
Hadrian builds highly automated factories that manufacture precision machined parts for the aerospace and defense industries. It was founded in 2021 by CEO Chris Power. Hadrian’s valuation has moved with startling speed, closing a funding round at $7.9 billion in August 2026, more than four times where it stood just seven months earlier, backed by Baillie Gifford and J.P. Morgan’s Strategic Investment Group among others. It makes money the traditional way, manufacturing and selling physical hardware to aerospace and defense customers, but built on a software platform called Opus that automates factory operations, which is exactly the combination investors are currently paying a premium for in the broader push to rebuild American manufacturing capacity.
Crusoe
Crusoe builds AI data center infrastructure with an energy first approach, developing its own power generation, including nuclear, thermal, and renewable partnerships, to feed the enormous electricity demands of AI computing. It was founded by CEO Chase Lochmiller and President Cully Cavness. In September 2026, Crusoe announced the initial closing of a $3.9 billion Series F round at a $30.9 billion valuation, backed by Nvidia, Founders Fund, and Qatar’s sovereign wealth fund among others, and the company reports more than $140 billion in total contracted value across its infrastructure platform. Its business model is renting out compute capacity from its own AI factories, differentiated specifically by controlling its own power supply rather than depending entirely on the existing grid.
Base Power
Base Power installs home battery systems that back up a house during outages and can sell electricity back to the grid during periods of high demand. It was founded in Austin, Texas in 2023 by Zach Dell and Justin Lopas. The company’s valuation climbed from $4 billion in October 2025 to $13 billion by August 2026 after a $1 billion Series D round, with total funding of $2.28 billion from investors including Andreessen Horowitz, Lightspeed, and Thrive Capital. Base Power makes money selling and installing physical battery hardware to homeowners, then earning ongoing revenue by using the combined network of home batteries to sell power back into the grid during peak demand, turning individual houses into a distributed power plant.
Valar Atomics
Valar Atomics is building small, factory built nuclear reactors designed to be cheaper and faster to deploy than traditional nuclear plants, aimed squarely at the power hungry AI data center industry. It was founded in 2023 by CEO Isaiah P. Taylor. The company raised $1 billion in equity plus a $200 million credit line in 2026, with Sequoia’s Shaun Maguire joining its board, at a reported $6 billion valuation, up from $2 billion in an earlier round. In June 2026, Valar demonstrated its Ward 250 reactor successfully powering an Nvidia Blackwell AI system and announced a deal with Nvidia to build a waterless 30 megawatt AI factory. Its business model is selling and operating small nuclear reactors directly to the data center operators who need guaranteed, dedicated power.
The Ones You Probably Haven’t Heard Of Yet
Board
Board is a genuine counter trend on this list, a startup built specifically to get people off their phones and into in-person games and social experiences together. It was founded by Brynn Putnam, who previously founded the connected fitness company Mirror. In an environment where almost every other startup on this list is trying to build more powerful software, Board is betting that the actual scarce resource left is real human connection, and that people will pay for structured reasons to gather in person again.
Ndea
Ndea is an AGI research lab founded by François Chollet, the researcher behind the ARC benchmark, one of the most respected tests of genuine reasoning ability in AI systems rather than simple pattern matching. Ndea entered Y Combinator’s Winter 2026 batch already funded at a scale no other company in that cohort came close to matching, a signal of how seriously investors are taking Chollet’s specific approach to building AI that can reason, rather than simply predict the next word based on everything it has already seen.
Synthetic Sciences
Synthetic Sciences might be the most startling founder story on this entire list. It was co-founded by Aayam Bansal and Ishaan Gangwani, both around 18 years old, who raised $1.4 million before even joining Y Combinator, from backers including the a16z Scout Fund. The company sits at the intersection of biology and AI research. Whatever the company becomes, the fact that two teenagers can raise real institutional money for a scientific research startup before finishing their first year of college says something significant about how far down the age curve serious AI funding has started to reach.
Farther
Farther is a wealth management platform that combines AI with human financial advisors, and it reached a $1.25 billion valuation as one of the roughly 90 new unicorns minted globally in 2026, a year in which most new unicorns were AI related but a handful, like Farther, came from outside pure AI infrastructure. Its business model blends traditional advisory fees with AI powered tools meant to make each individual advisor able to serve far more clients than a traditional wealth management practice, a quiet but potentially significant efficiency play inside one of finance’s most relationship dependent industries.
Surge AI
Surge AI might be the single most surprising company on this entire list, because almost nobody outside the AI industry has heard of it. Founded by Edwin Chen and built entirely without venture capital until a small 2025 raise, Surge AI provides the human data labeling and reinforcement learning feedback that major AI labs, including Anthropic and Google, actually use to train their models. The company generated an estimated $1.2 billion in revenue in 2024 alone and around $1.4 billion since, all while Chen retained roughly 75 percent ownership. That ownership stake reportedly makes him worth close to $18 billion, making him one of the richest and least publicly known people in the entire AI industry, proof that the most mind blowing startup on a list like this one is not always the one with the flashiest product.
What These 30 Companies Have in Common
Look across all thirty and a few patterns repeat. Nearly every founder here built deep, specific expertise, often as a researcher or engineer at a major lab, before starting their own company, rather than being a generalist entrepreneur chasing a trend. Almost none of them are trying to build one AI model to do everything. They are each betting on a specific, narrow claim: this model can control any robot, this platform can replace a paralegal’s first draft, this reactor can power one data center. And the money moving into all of this is now large enough, and moving fast enough, that a company’s valuation can double in four months, as it did for more than one name on this list, which is either the clearest sign yet of how much real value is being created, or a bubble large enough that a correction would be felt well beyond the tech industry. Probably, if history is any guide, it is both at once.
Why We Wrote This
Wolvra covers business and technology stories with real numbers attached, not just hype, which is exactly what this list tries to do across thirty very different companies at once. If you want to understand more about how we approach stories like this one, visit our Brand Guidelines page, or learn more about what Wolvra stands for on our About Us page. If you know a startup that deserves a closer look, or spot a number here that has changed, our Contact Us page is open.