San Francisco still gets the headlines, but New York is running its own AI economy and it is bigger than most people know. More than 2,000 AI startups now call the city home, backed by over $27 billion in funding since 2019 and a workforce of 40,000-plus AI professionals.
That is an industry, plain and simple.
What makes New York different is the lack of a single obsession. San Francisco chases foundation models and developer tooling. New York builds AI for the industries it already dominates: finance, media, healthcare, and enterprise software.
Less “we trained a model,” more “we automated the thing your compliance team dreaded.”
A few things set this ecosystem apart:
- Density at the early stage. Manhattan now produces more seed and Series A companies in a single year than the Bay Area, 543 versus 486 by one count.
- Vertical depth over horizontal hype. Where Silicon Valley optimizes for general-purpose model capability, New York optimizes for fitting AI into finance, healthcare, and media workflows that already have entrenched compliance and regulatory demands.
Here are the 25 companies proving that thesis right now, ranked loosely by relevance and momentum rather than valuation.
1. Hugging Face
Hugging Face anchors much of the open-source AI world from its New York base, hosting the model hub that practically every ML team has bookmarked at some point. It is the closest thing the AI community has to a town square, with hundreds of thousands of models, datasets, and Spaces hosted for free.
The company’s bet is that open infrastructure beats closed gardens long term, and so far the betting odds look decent.
2. Runway
Runway builds multimodal generative video tools that have shifted what “AI filmmaking” even means, with its open research lineage now powering creative pipelines well beyond its own product. It is one of a small number of companies globally training its own foundation models rather than wrapping someone else’s API. That puts Runway in direct competition with labs many times its size, which is either bold or slightly unhinged, depending on the day.
3. Clarifai
Clarifai has spent over a decade in computer vision, long enough to watch the rest of the field catch up to problems it was already solving. The platform handles image, video, and text recognition at enterprise scale, serving government and commercial clients who needed production-grade vision models before “AI” was a board-level agenda item.
Longevity in this industry is rare, and Clarifai has it.
4. Two Sigma
Two Sigma applies machine learning to systematic trading at a scale most quant shops only dream about, running statistical models across enormous datasets to find signals in market noise. Founded by former technologists rather than traditional Wall Street traders, the firm treats trading as a data science problem first and a finance problem second.
It remains one of the clearest examples of New York’s edge: deep capital markets expertise paired with deep ML talent.
5. Kalshi
Kalshi uses AI to power federally regulated prediction markets, and its funding trajectory tells its own story: a $300 million Series D in October 2025, then a $1 billion Series E two months later, then another $1 billion Series F in May 2026 that pushed its valuation to $22 billion. Prediction markets let traders bet directly on real-world outcomes, from economic data releases to cultural events, and Kalshi’s models help price and surface those markets in real time.
Regulators are still figuring out exactly how to think about this category, which makes Kalshi worth watching closely.
6. Ramp
Ramp embeds AI into corporate spend management, so finance teams stop reconciling receipts by hand and start catching wasteful spend automatically. Its AI agents flag duplicate charges, unused software subscriptions, and policy violations before a human even opens the expense report.
For an industry built on tedium, automating the tedium is the entire pitch, and it is working.
7. AlphaSense
AlphaSense mines market intelligence from earnings calls, filings, and expert transcripts, turning a research process that used to take analysts days into a search query that takes minutes. Investment banks, hedge funds, and corporate strategy teams treat it as a primary research layer now, ahead of the analysts who used to own that job.
It is a good reminder that the most valuable AI products are often the boring ones people end up relying on daily.
8. F2
F2 accelerates private markets diligence for private credit, private equity, and commercial banking teams, and the company says its platform evaluates deals over 60% faster by turning unstructured deal data into investment-grade analysis.
Backed by NFX, Left Lane Capital, and Y Combinator, F2 is betting that the firms doing the most deals will pay the most for speed. Given how much of private equity still runs on PDFs and prayer, that bet looks sound.
9. OffDeal
OffDeal is an AI-native investment bank built for small business owners who want to sell their companies at premium prices for a fraction of typical Wall Street fees. Its AI handles the analyst-level grunt work, financial modeling, buyer outreach, and document preparation, freeing human advisors to focus on strategy and negotiation.
It is a niche many assumed AI would skip entirely, mostly because nobody thought to try.
10. Petal
Petal applies machine learning to credit underwriting for people that traditional credit scores routinely ignore, using cash flow data instead of credit history alone. The approach lets the company extend credit to people with thin or no credit files, a group standard scoring models tend to overlook entirely. It is a useful case study in AI expanding access rather than just optimizing margins.
11. Socure
Socure verifies identity at a scale that makes manual fraud checks feel almost quaint, combining device, document, and behavioral signals into a single trust score. Banks, fintechs, and government agencies use it to onboard customers while staying ahead of synthetic identity fraud, which has exploded alongside generative AI’s ability to fabricate convincing fake documents. The irony of fighting AI fraud with AI detection lands on everyone in the building eventually.
12. Forter
Forter catches e-commerce fraud in real time by analyzing transaction patterns across its network of merchant clients, approving good orders and blocking bad ones before checkout completes. The company built its model on the idea that fraud signals are stronger in aggregate across merchants than in isolation at any single retailer. That network effect is hard to replicate, which is exactly why it works.
13. BigID
BigID maps and protects sensitive data across the enterprise, which is a polite way of saying it finds the personal and regulated data your security team forgot existed. As privacy regulation expands globally, knowing where sensitive data actually lives has become a board-level liability question rather than an IT afterthought. BigID’s platform answers that question continuously instead of through an annual audit panic.
14. Deep Instinct
Deep Instinct applies deep learning to malware prevention out of its joint New York and Tel Aviv operations, predicting and blocking threats before they execute rather than detecting them after the fact. Traditional antivirus reacts to known signatures; Deep Instinct’s models are trained to recognize malicious intent in files they are encountering for the first time. In a threat landscape where zero-day exploits move faster than patch cycles, that distinction matters.
15. Adaptive Security
Adaptive Security has raised roughly $136 million across two rounds: a Series A led by Andreessen Horowitz and the OpenAI Startup Fund in April 2025, followed by a Series B led by Bain Capital Ventures that December, building AI-driven defenses against AI-generated threats like deepfake phishing and voice cloning scams. As generative AI makes social engineering attacks cheaper and more convincing, the company’s pitch is that you need AI watching the door. Two funding rounds in eight months from investors, which suggests they agree this problem is only getting worse.
16. Behavox
Behavox watches for compliance and conduct risk inside enterprise communications, scanning email, chat, and voice for signs of misconduct, collusion, or regulatory violations. Financial institutions use it to catch the kind of behavior that used to surface only after a regulator’s subpoena arrived. It is either deeply reassuring or mildly terrifying, depending on how your last Slack message reads out of context.
17. Flatiron Health
Flatiron Health turns oncology data from electronic health records into research that actually changes treatment decisions, partnering with cancer centers and pharmaceutical companies on real-world evidence studies. Its models help identify which treatments work best for which patient subgroups, a question clinical trials alone often answer too slowly to matter.
The company proves that healthcare AI’s biggest wins sometimes live in data plumbing rather than diagnosis itself.
18. Oscar Health
Oscar Health builds AI directly into the health insurance experience, from claims triage to member support, trying to fix an industry famous for opacity and frustration. Its tools help predict member risk earlier and route care recommendations before small problems become expensive ones.
Whether AI can actually fix health insurance remains an open question, but Oscar is one of the few insurers willing to test it at scale.
19. K Health
K Health offers AI-driven symptom guidance grounded in real, anonymized patient data rather than generic medical literature, comparing a user’s symptoms against millions of similar historical cases. It is a useful first stop before urgent care or the emergency room, and it might stop you from googling your symptoms at 2 am and assuming the worst.
20. Spring Health
Spring Health applies AI to mental health benefits at the employer level, matching employees to the right type of care, therapy, coaching, or medication management faster than traditional referral networks manage. Its matching algorithm is built to reduce the trial-and-error most people experience when finding the right therapist.
For an industry where the average wait time for care is measured in weeks, faster matching alone is a meaningful outcome.
21. Understood Care
Understood Care, fresh out of Y Combinator, matches Medicare patients with AI-assisted patient advocates who help navigate paperwork, find resources, and access care, a benefit that became Medicare-covered only in 2024.
The company cites a figure of 88% of US adults reporting difficulty navigating the healthcare system, and with 68 million people on Medicare, the addressable problem is enormous regardless of the exact number. It is a clean example of policy change creating room for a genuinely new AI-enabled service category.
22. Datadog
Datadog has built AI-assisted observability into the backbone of how engineering teams monitor production systems, using machine learning to detect anomalies and predict incidents before they cause outages.
As infrastructure has grown more distributed and harder to reason about manually, Datadog’s models take over the pattern recognition work that now exceeds human scale. It is infrastructure AI that stays invisible to most end users while every engineering team relies on it daily.
23. Bloomberg
Bloomberg trained its own domain-specific language model on decades of proprietary financial data, betting that a model built on its corpus would outperform general-purpose alternatives on financial tasks. The company has the kind of data moat that took 40 years to build, the sort of advantage money alone struggles to buy.
It is a reminder that incumbents with deep proprietary data still have a real structural advantage in the AI era.
24. Yext
Yext applies AI to search and brand answers across the web, helping companies control how their information appears when customers ask AI assistants and search engines questions about them. As generative AI search reshapes how people find businesses, Yext’s pitch has shifted from managing listings to managing how AI itself describes a brand.
That shift from SEO to what some are calling “answer engine optimization” is becoming its own discipline fast.
25. Dataminr
Dataminr scans the open web in real time, from social media to public sensors, to surface breaking events before most newsrooms or corporate security teams catch wind of them.
Its AI flags everything from natural disasters to security incidents within minutes of the first public signal appearing online. For organizations where minutes of advance warning translate into real safety and financial outcomes, that speed is the entire product.
Want to see this thesis in action?
A good chunk of this list lives at the intersection of finance and AI, and that intersection gets its own stage this October. The Agentic AI in Financial Services Summit lands in New York on October 1, 2026, bringing together engineers, risk leaders, and domain experts from firms including Citi, BNY Mellon, T. Rowe Price, and Fitch Group.
Benefit from the following:
- A grounded view of what’s working, from architecture to evaluation
- A benchmark against deployed systems, on governance and inference
- A direct line into the regulatory debate, where compliance meets engineering
- A sharper network of engineers and leaders building institution-grade AI
