The Hottest AI Startups: November 2026

October's biggest AI rounds went to research labs. TypeSafe AI, which made our October list after a $40 million seed, raised $870 million at a $7.5 billion valuation in a round led by Andreessen Horowitz, and Flapping Airplanes raised a $1 billion Series A co-led by Index Ventures. A level below those headlines, robotics drew the most seed money: three of the eleven companies here build models or training data for robots, and two of those three raised $40 million or more.

Each company on this list had a funding round, a product launch, or a significant customer milestone between October 4 and October 31, 2026. Most are at seed or Series A, where a company's trajectory is easiest to change and hardest to predict. The list is refreshed every month, and the October 2026 edition remains available.

The Hottest AI Startups - November 2026

#	Company	Headquarters	What it does	Latest news	Lead investor(s)
1	Different	San Francisco	AI system that runs B2B demand generation	$48M seed	General Catalyst
2	Dream Labs	San Francisco	Robot foundation model sold through an API	$90M seed	B Capital, Kindred Ventures
3	Ghost	San Francisco	Personal AI computer that runs agents on the device	$11M seed	Andreessen Horowitz
4	Healthleap	Founded in South Africa	AI that flags hospital patients with overlooked conditions	$38M seed and Series A	Hummingbird Ventures
5	Hone	San Francisco	AI "engines" that own business outcomes	$60M seed	Benchmark, Index Ventures
6	Machenta	London	Robots that learn a task from one demonstration	$40M seed	Creandum, Union Square Ventures
7	Maxinsights	Santa Clara, Calif.	First-person video and touch data for training robots	$30M seed	Peak XV
8	Phinity Labs	San Francisco	AI agents that design and tape out custom chips	$5.2M seed	Uncork Capital
9	Phonic	San Francisco	Speech-to-speech model for voice agents	$27.5M Series A	First Round Capital
10	Rivercell	Paris	AI model of how human cells respond to drugs	$25M seed	HV Capital
11	Watchtower Labs	Seattle	AI planning tools for military logistics	$35M seed	Shield Capital, Conviction

$48M

Seed, October 20

Different

AI system that runs B2B demand generation

Headquarters: San Francisco
Backers: General Catalyst (lead) and Lightspeed

Different is the third company from Ajeet Singh, who co-founded Nutanix and ThoughtSpot. Its software plans, launches, and tunes B2B demand-generation campaigns end to end, including answer-engine optimization, for marketing teams too small to staff every channel. Singh's co-founders come from Adobe and Flipkart, and early customers include Motif and Descope.

Why it's hot

Hemant Taneja led the round for General Catalyst, and a $48 million seed is unusually large for a marketing product. Lean B2B teams are under pressure to produce more demand with fewer people, which makes software that runs the whole program across every channel an easy pitch.

$90M

Seed, October 19

Dream Labs

Robot foundation model sold through an API

Headquarters: San Francisco
Backers: B Capital and Kindred Ventures (co-leads), Wing, SV Angel, Unusual Ventures, and others

Dream Labs was started by former Nvidia researchers and is led by chief executive Joel Jang. Its Dream-1 model learns from video of people doing tasks and can control two-armed robots with parallel grippers without training on each new machine. The Robot Report says the Dream API supports robots that cost about $2,000 to $10,000 and that the company already has more than 20 hardware partners.

Why it's hot

This was one of the largest seed rounds in AI in October, and most of the money is going to compute. Selling robot intelligence by the hour to makers of cheap hardware is a different business from building a humanoid, and it lets Dream Labs grow with every robot maker that plugs in.

$11M

Seed, October 5

Ghost

Personal AI computer that runs agents on the device

Headquarters: San Francisco
Backers: Andreessen Horowitz (lead), Abstract, Audacious Ventures, SV Angel, and Nova

Ghost's first product is Core, a $3,499 screenless computer with an Nvidia RTX Pro 4000 Blackwell GPU that runs AI agents around the clock on a person's files, apps, and smart-home devices. TechCrunch reports that models and personal memory stay on the device, encrypted so that Ghost itself cannot read them. Founder and chief executive Zain Javaid is 19.

Why it's hot

Personal-AI hardware has a rough record, but Ghost is aimed at people who already run agents on a Mac Mini and want something built for the job. Preorders opened on October 5 and the first units were scheduled to ship in the last week of October, so buyer reviews arrive quickly.

$38M

Seed and Series A, October 7

Healthleap

AI that flags hospital patients with overlooked conditions

Headquarters: Founded in South Africa
Backers: Hummingbird Ventures (Series A lead), Sequoia Capital and First Round Capital (seed co-leads)

Siblings Jemima and Josiah Meyer founded Healthleap in 2022. Its software reads electronic health records, including clinical notes, lab results, and vital signs, and scores patients for conditions such as malnutrition and delirium that care teams often miss. TechCrunch reports it now runs in more than 50 hospitals, including Penn Medicine, Cedars-Sinai, Houston Methodist, and Emory.

Why it's hot

Healthleap went from three hospitals to more than 50 and grew revenue more than tenfold, which is the kind of traction health systems rarely give a young vendor. The risk scores sit inside the existing record system, so nurses and dietitians do not have to learn a new tool.

$60M

Seed, October 8

Hone

AI "engines" that own business outcomes

Headquarters: San Francisco
Backers: Benchmark and Index Ventures (co-leads)

Hone builds what it calls engines: persistent agents that are given an outcome, such as running procurement, then onboard themselves, work alongside the team, and improve over time. Index Ventures describes the founders as Moritz Stephan, who helped build Cognition's Devin, Oliver Brady, who worked at Harvey and Mercor, and Thiel Fellow Carlo Kobe.

Why it's hot

Two top-tier firms co-leading a $60 million seed, at a reported $285 million valuation, is a strong bet on agents that are paid for results. Most enterprise agents still handle single tasks, and Hone is trying to own the whole job.

$40M

Seed, October 27

Machenta

Robots that learn a task from one demonstration

Headquarters: London
Backers: Creandum and Union Square Ventures (co-leads)

Machenta is led by Ed Johns, director of the Robot Learning Lab at Imperial College London and former head of robot learning at Dyson. Its model is trained entirely on synthetic data, so a robot can watch a person do a task once and then repeat it with objects it has never seen. Union Square Ventures calls the approach "physical vibe coding".

Why it's hot

Most robotics money is going into collecting more real-world data. Machenta is the contrarian bet that robots need far less of it, and if one demonstration is enough, the cost of teaching a robot a new job drops sharply.

$30M

Seed, October 22

Maxinsights

First-person video and touch data for training robots

Headquarters: Santa Clara, Calif.
Backers: Peak XV (lead), Tola Capital, South Park Commons, INCE Capital, and Alumni Ventures

Maxinsights pays more than 5,000 contractors in six countries to wear head, wrist, and chest cameras while they cook, clean, and work, then turns the footage into robot training data. The Next Web reports it has captured more than 2 million hours and supplied all 200,000 hours of first-person data behind Genesis AI's latest robot model. It is building MaxGlove, a sensor glove for grip force, with Genesis.

Why it's hot

The company has more than 30 customers, including frontier AI labs, and eight are on recurring contracts. Investors put $47.4 billion into physical AI in the first half of 2026, according to Crunchbase, and the training data those robots need is still scarce.

$5.2M

Seed, October 8

Phinity Labs

AI agents that design and tape out custom chips

Headquarters: San Francisco
Backers: Uncork Capital (lead), Moxxie, and angels including Jeff Dean

Phinity is building agents that design, verify, and revise chips using physical feedback, from architecture through tapeout. The company says it already works with some of the largest foundation model labs and has reached an eight-figure annualized run rate. Its founders came to chip design from post-training work, including building synthetic training data for Nvidia's chip-design models.

Why it's hot

A small seed round paired with eight-figure revenue is rare, and Google's Jeff Dean is among the angels. Custom silicon today takes nine to 24 months, so any credible way to shorten that cycle has a large market waiting.

$27.5M

Series A, October 29

Phonic

Speech-to-speech model for voice agents

Headquarters: San Francisco
Backers: First Round Capital (lead), Lux Capital, Felicis, Definition, and Essence VC

Phonic released Phonic-1, a speech-to-speech model that can call tools while a caller is still talking, alongside its Series A. The company says the model took first place on the Artificial Analysis Speech to Speech Index and works in more than 50 languages. Customers include Ashby, which uses Phonic for AI interviews in 25 countries, Maven AGI, and Rho.

Why it's hot

Voice agents are moving from stitched-together stacks of transcription, language, and speech models to single models that hear tone and timing. Phonic already handles calls on six continents, and the round brings its total funding to $31.5 million.

$25M

Seed, October 7

Rivercell

AI model of how human cells respond to drugs

Headquarters: Paris
Backers: HV Capital (lead), HCVC, Alven, and Bpifrance Digital Venture

Rivercell was started in 2025 by Yann Fleureau, who co-founded Cardiologs and sold it to Philips, and Eric Durand, who led oncology data science at Novartis and data science at Owkin. The Next Web reports that it films living cells through each experiment and then reads their gene activity, with the model choosing which experiment the lab runs next.

Why it's hot

The race to build an AI "virtual cell" depends on data that does not exist yet, and Rivercell is generating its own in a Paris wet lab. A $25 million seed for a team with an exit and deep pharma experience makes it one of Europe's better-funded entrants.

$35M

Seed, October 21

Watchtower Labs

AI planning tools for military logistics

Headquarters: Seattle
Backers: Shield Capital and Conviction (co-leads) and Founders Fund

Watchtower is led by chief executive Akhil Iyer, a former Marine special operations officer, with co-founders Souren Aghajanyan and John Morman, both formerly of Microsoft. The company says the Defense Innovation Unit selected it in January 2026 for its Joint Sustainment Decision Tool program, and it is starting with logistics planning.

Why it's hot

Watchtower was incubated at Shield Capital, whose co-founder Raj Shah is its chairman, and it already has a defense program in hand. Defense buyers are moving money toward software that can plan faster than a staff can, and logistics is where those delays cost the most.

What to watch in December

Three threads will shape next month's list. First, robot training data: Machenta's one-demonstration approach and Maxinsights' footage-at-scale approach are opposite bets on how much real-world data robots need, and new customer deals on either side would show which one buyers trust. Second, voice: Phonic-1 is now competing with Deepslate, from our October list, and the big labs' own speech models for the same contact-center work. Third, personal AI hardware: Ghost's first Core units were scheduled to ship at the end of October, so the first buyer reviews should appear in November.