What are the most valuable generative AI startups?

Top startups in the Generative AI market ranked by valuation

Here is an updated table that ranks the top startups in the Generative AI market based on their latest reported or estimated valuations.

If you need to dig deeper and get more detailed data, please check our report covering the Generative AI market.

OpenAI and Anthropic are by far the most valuable Generative AI startups today, followed at a huge distance by xAI.

Our dataset puts OpenAI at $1.2T–$1.3T based on its current fundraising discussions, Anthropic at $965B–$1.2T and xAI at $250B. After those three, the ranking drops abruptly to Anysphere, the company behind Cursor, at $60B and Cognition at $48B.

That gap is large enough to split Generative AI into several valuation tiers. OpenAI and Anthropic are approaching the scale of the world's largest public technology companies. xAI sits in its own middle tier. Then come AI coding companies, frontier-model challengers such as Mistral and Reflection AI, and fast-growing application or infrastructure businesses such as Fireworks AI, Sierra, Harvey and ElevenLabs.

The lower half of the top 20 is already highly diverse. It contains AI search, inference infrastructure, legal software, voice AI, enterprise search, clinical documentation and music generation. Generative AI has produced valuable companies across a lot of markets, even though most of the industry's value is still concentrated at the very top.

Rank	Startup	Current valuation	Main business
1	OpenAI	$1.2T–$1.3T	Frontier models and AI products
2	Anthropic	$965B–$1.2T	Frontier models and enterprise AI
3	xAI	$250B	Models and AI infrastructure
4	Anysphere / Cursor	$60B	AI coding
5	Cognition	$48B	AI software-engineering agents
6	Reflection AI	$27.5B–$45B	Frontier AI models
7	Perplexity	$30B–$35B	AI search and agents
8	Mistral AI	$24B	Foundation models
9	Cohere	$18B–$20B	Enterprise foundation models
10	Fireworks AI	$17.5B	AI inference infrastructure

Is OpenAI really worth more than $1 trillion now?

OpenAI can currently be described as a trillion-dollar Generative AI company only if we use the valuation being discussed for its next financing rather than its last completed round.

Bloomberg reported that OpenAI completed a $122B financing at an $852B valuation earlier this year. More recently, the company entered early discussions around another financing at more than $1.2T. That prospective figure is the best indication of where investors may price OpenAI now, but the transaction has not closed.

The difference is substantial. Moving from $852B to $1.2T would add roughly $348B of value, or about 41%, in less than a year.

OpenAI also has enough revenue scale to make the conversation less absurd than the headline initially sounds. The Financial Times recently reported that OpenAI's annualized revenue had moved above $40B. At $1.2T, that would still imply a valuation of roughly 30 times annualized revenue, an extremely rich multiple for a company with enormous infrastructure spending.

The financial risk is just as unusual as the growth. OpenAI's expansion requires huge commitments to chips, data centers and model development. Investors are betting that revenue can keep growing fast enough for today's massive compute bill to become much smaller relative to the future business.

So $1.2T is best treated as OpenAI's current fundraising value, while $852B remains the strongest completed transaction anchor.

Is Anthropic actually catching OpenAI?

Anthropic has caught OpenAI in private-market valuation much faster than most people expected.

Anthropic's latest completed financing is unusually easy to verify for a private AI company: it announced a $65B Series H at a $965B post-money valuation. Anthropic also said its run-rate revenue had crossed $47B.

That puts Anthropic above OpenAI's last completed $852B valuation, although OpenAI's newer fundraising discussions point to a value above $1.2T. Depending on which OpenAI reference we use, Anthropic is either slightly ahead or roughly 20% behind.

The commercial gap has narrowed too. Claude has become deeply embedded in professional workflows, and Claude Code has helped Anthropic capture unusually valuable software-development usage. Enterprise customers also appear to account for a much larger part of Anthropic's business than they did during the early ChatGPT-versus-Claude period.

OpenAI still has broader consumer distribution through ChatGPT and operates across consumer subscriptions, enterprise products, APIs and coding tools. Anthropic's position is more concentrated around professional and enterprise use.

Still, a $965B completed valuation makes Anthropic far more than a distant challenger. In valuation terms, Generative AI currently has two companies operating near the trillion-dollar line.

How big is the gap between OpenAI, Anthropic and every other Generative AI startup?

The Generative AI valuation gap is enormous: OpenAI, Anthropic and xAI together are worth roughly ten times the next seven companies combined.

Using the lower end of our valuation ranges, OpenAI, Anthropic and xAI together represent about $2.4T. Using the upper end, they approach $2.75T.

Companies ranked fourth through tenth add up to roughly $205B–$254B. Even that group includes Cursor at $60B, Cognition at $48B, Perplexity at more than $30B, Mistral at $24B and Fireworks at $17.5B. Where the next tier of generative AI companies sits relative to the giants is shown on the market grids in our generative AI market report.

Another way to see the gap is to compare entire application markets with one frontier lab. Harvey and Legora together are worth around $21B in legal AI. Runway, Synthesia and Luma AI together are worth roughly $13B–$14B in generative video. Sierra, Decagon and Parloa total about $23B across customer-service AI.

Each of those clusters represents only a small fraction of Anthropic's $965B valuation.

Investors are placing a much larger financial value on companies that could own a general-purpose intelligence platform than on companies dominating one specific workflow.

How many Generative AI unicorns are there?

Our dataset contains roughly 45 Generative AI companies that clearly exceed $1B or have a credible current valuation range extending above $1B.

The exact number changes depending on how strictly we apply the unicorn definition. Character.AI sits around $900M–$1.1B, AI21 Labs around $900M–$1.4B and Hebbia around $800M–$1.1B. Counting every company whose range crosses $1B produces a higher total than requiring the whole range to remain above it.

The more useful finding is how crowded the higher levels have become. Around 15 companies in our ranking are worth roughly $10B or more. Five reach at least $48B. Those valuation levels used to be extraordinary outcomes for mature private technology companies.

A $1B valuation therefore tells us much less about market leadership in Generative AI today. The companies really separating from the field are increasingly those breaking through $5B, $10B and, in a handful of cases, $50B.

Are foundation-model startups still worth more than Generative AI applications?

Foundation-model startups still dominate Generative AI valuations overall, but the strongest AI applications have now become tens-of-billions-of-dollars companies.

OpenAI, Anthropic, xAI, Mistral, Reflection AI and Cohere collectively represent an enormous share of the value in our ranking. Their advantage comes from selling intelligence across many different applications instead of depending on one workflow.

Application companies are getting much larger, though. Cursor reached a $60B acquisition value. Cognition is worth $48B. Harvey reached $15.5B in legal AI, Sierra $15.8B in customer-service agents and ElevenLabs $11B in voice AI.

Those companies also weaken the old idea that application startups are inevitably thin wrappers around somebody else's model. The leaders increasingly own workflow, customer relationships, distribution, proprietary context or specialized technology around the underlying model.

Gamma provides a smaller but useful example. The presentation and content platform announced that it had crossed $100M in ARR while remaining profitable, then raised at a $2.1B valuation. That's a fairly conventional software business with AI built deeply into the product rather than a company trying to finance frontier-model research.

Frontier models still capture the biggest valuations, but applications have clearly shown that they can create durable multi-billion-dollar companies.

Has AI coding become the most valuable Generative AI application?

AI coding is currently the most valuable pure application category in Generative AI.

Cursor's $60B acquisition value and Cognition's $48B valuation put both companies ahead of every other application-focused startup in our dataset. Coding is one of the problems buyers pay to solve, and we map which others they pay for in our market report about generative AI.

Cognition's recent numbers help explain the enthusiasm. The company said annualized run-rate revenue grew from $492M to nearly $900M in about four months. Its valuation moved from $26B to $48B over roughly the same period. Revenue rose around 83%, while valuation increased about 85%.

Cursor provides even stronger external validation because its $60B value came through a completed acquisition rather than another venture financing. Bloomberg reported that SpaceX completed the deal after an earlier agreement giving it the right to acquire the company.

Coding is unusually well suited to AI. Software engineers are expensive, almost all of their work is already digital, generated code can often be tested automatically, and even modest productivity improvements can justify expensive subscriptions.

Legal AI and customer-service agents are producing large companies too, but neither category has yet created two businesses approaching $50B.

Generative AI application	Leading company	Current valuation
AI coding	Cursor / Anysphere	$60B
Software-engineering agents	Cognition	$48B
Customer-service agents	Sierra	$15.8B
Legal AI	Harvey	$15.5B
Voice AI	ElevenLabs	$11B
Enterprise search	Glean	$8B–$11B
Legal AI	Legora	$5.6B
AI music	Suno	$5.4B
Generative video	Runway	$5.3B
AI avatar video	Synthesia	$4B

Are AI agents becoming more valuable than basic Generative AI tools?

AI agents are attracting much higher valuations than many of the writing, image and chatbot startups that dominated the first Generative AI boom.

Cognition is worth $48B, Sierra $15.8B, Harvey $15.5B, Decagon $4.5B and Parloa $3B. These companies sell systems that complete parts of a workflow rather than stopping after generating text.

The contrast with several early Generative AI names is sharp. Jasper is estimated around $700M–$1.1B. Stability AI sits around $450M–$550M. Character.AI is around $900M–$1.1B. Those companies were once some of the most visible names in the sector.

The economics explain a lot of the change. A writing tool has to justify a software subscription. An AI customer-service agent can be compared with the cost of thousands of human support interactions. Coding agents compete with expensive engineering labor. Legal AI can save hours billed at hundreds of dollars each.

Investors are now paying the largest application valuations for AI tied directly to expensive work.

Why is legal AI suddenly worth so much?

Legal AI has become one of the richest Generative AI verticals because law firms are paying for products that save expensive professional time.

Harvey recently raised $550M at a $15.5B valuation, only months after reaching $11B and less than a year after being valued around $8B. The company says 80% of the Am Law 100 now uses Harvey.

Legora has followed a similar path at smaller scale. Its latest financing valued the company at $5.6B after it crossed $100M in ARR. That valuation is more than three times the roughly $1.8B figure attached to its previous major round.

The customer economics are unusually attractive for AI. Large law firms bill substantial amounts for research, drafting, document review, due diligence and other text-heavy work. If software removes even a few hours from a matter, a high annual subscription can be easy to justify.

Legal work also gives specialist platforms room to differentiate. Firms need document permissions, matter-specific context, auditability, integrations and tools designed around legal workflows. A generic chatbot can perform some legal tasks, but deploying AI across an entire firm requires much more than a good answer in a chat window.

Harvey and Legora now have more than $20B of combined value. Legal AI has turned into one of the clearest examples of a vertical market producing genuinely large Generative AI companies.

How big has healthcare AI become?

Healthcare AI has already produced several billion-dollar startups, although its leaders are still much smaller than the biggest legal or coding companies.

Abridge sits around $6.2B–$7.2B in our dataset. Hippocratic AI has a confirmed $3.5B valuation after raising $126M in its latest round. Ambience Healthcare is estimated around $1.2B–$1.4B.

There is also a substantial group below $1B: Heidi Health around $900M, Nabla at roughly $550M–$800M, Corti around $450M–$700M, Regard around $400M–$550M and Suki around $430M–$460M.

The largest cluster is developing around clinical documentation and workflow automation. Abridge, Ambience, Nabla and Heidi all reduce the time clinicians spend writing notes or interacting with health-record systems. Hippocratic AI has taken a different route with patient-facing healthcare agents and said it was working with more than 50 large health systems, payers and pharmaceutical companies around the time of its $3.5B financing.

Healthcare has plenty of economic value to capture, but deployments are harder than in ordinary software. Hospitals deal with sensitive data, complex integrations, procurement processes and high consequences when systems fail.

That friction helps explain why healthcare has created many valuable AI companies without yet producing a $10B-plus application leader.

Which creative AI startups are worth the most now?

ElevenLabs, Suno and Runway currently lead the creative Generative AI market, while video has produced the deepest group of multi-billion-dollar companies.

ElevenLabs is worth $11B. The company closed the previous year with more than $330M in ARR and has expanded from synthetic voice into enterprise conversational agents.

Suno raised more than $400M at a $5.4B post-money valuation. Fortune later reported that the music-generation company had reached roughly 2 million paying subscribers and $300M in ARR.

Runway reached $5.3B after raising $315M. Synthesia is at $4B after a $200M Series E, and Luma AI sits around $3.5B–$4.5B in our dataset. HeyGen occupies a similar $3B–$4.5B range.

Image generation looks less valuable today. Stability AI is around $450M–$550M, while Krea and Ideogram are each roughly in the $450M–$650M zone.

The creative market has moved well beyond basic image generation. Voice, music and especially video are supporting much larger businesses, partly because they can be sold into professional workflows as well as consumer creation. Selling into professional workflows, not only to consumers, is the kind of growth driver we weigh in our generative AI market report.

Why is ElevenLabs worth $11 billion?

ElevenLabs reached an $11B valuation because it has turned voice generation into a large enterprise business rather than relying mainly on novelty text-to-speech usage.

The company raised $500M at $11B after finishing the previous year with more than $330M in ARR. Its valuation had more than tripled in one year.

ElevenLabs has also widened the product considerably. Its ElevenAgents platform lets companies deploy voice and chat agents for customer service, sales and internal workflows. Deutsche Telekom, Square and Revolut were among the enterprise customers the company highlighted around the financing.

That expands the addressable market quite a bit. A voice-generation tool competes for media and creator spending. A voice agent can compete with part of a call center, sales operation or customer-service budget.

PolyAI, valued around $750M–$850M, provides another indication that businesses are willing to fund voice-agent specialists. Hume AI and Speechify sit lower, at approximately $150M–$300M and $120M–$180M respectively.

ElevenLabs has separated from that group because its revenue scale is already much larger and its product now spans both generative audio and conversational automation.

Why are AI infrastructure startups worth billions if AI models keep getting cheaper?

AI infrastructure startups are worth billions because cheaper models are driving much more AI usage, and somebody still has to run, customize and serve those models efficiently.

Fireworks AI is the clearest example. The company raised $1.505B at a $17.5B valuation after surpassing $1B in annualized revenue run rate. It also said its infrastructure was processing more than 40 trillion tokens per day.

Together AI reached $8.3B after raising $800M. Reuters reported that its annual bookings had crossed $1.15B, with customers including Cursor, Cognition and Decagon. Inference infrastructure is a layer where we separate what already works at scale from what is still being tested in our generative AI market report.

Those numbers show how rapidly the open-model infrastructure market has grown. Companies increasingly use several models, customize them on proprietary data and move workloads depending on price or performance. Specialized infrastructure providers make that easier.

Falling inference prices can actually help these businesses when lower costs trigger much more usage. Cloud computing followed a similar pattern: cheaper compute increased the number of workloads companies were willing to run.

The long-term risk is strong competition from hyperscalers. AWS, Microsoft and Google can bundle compute, models and enterprise relationships. Fireworks and Together AI still have to prove that specialization is worth paying for.

For now, annualized revenue or bookings above $1B make both companies much more substantial than infrastructure startups built only around investor expectations.

Is Europe actually producing big Generative AI startups?

Europe is producing several major Generative AI companies, although it still has nothing close to OpenAI or Anthropic in valuation.

Mistral AI is the obvious leader. The French company recently raised €3B at a valuation of roughly €21B, or about $24B. Reuters reported that Mistral's annual recurring revenue was around $1B.

ElevenLabs follows at $11B. Legora is worth $5.6B, Synthesia $4B and DeepL around $2B–$3B in our dataset. Photoroom, Nabla and Dust add another group of European companies worth hundreds of millions of dollars.

Europe is also producing leaders in several different parts of the stack. Mistral competes in foundation models, ElevenLabs in voice, Legora in legal AI, Synthesia in enterprise video and DeepL in language tools.

The remaining gap is concentrated at the very top. Mistral's $24B valuation is only about 2.5% of Anthropic's $965B completed valuation.

Europe has proved that it can build global Generative AI companies. It has not yet produced a frontier lab with anything close to the financial scale of the two American leaders.

Which Generative AI startups have increased their valuations the fastest?

Cognition, Fireworks AI, ElevenLabs, Sierra, Legora and Decagon are among the clearest examples of Generative AI startups whose valuations have jumped several times over in a short period.

Cognition moved from $10B in an earlier financing to $26B and then $48B. More importantly, its revenue moved quickly at the same time: annualized run-rate revenue climbed from $492M to nearly $900M between its two latest rounds.

Fireworks AI went from roughly $4B to $17.5B as annualized revenue passed $1B. ElevenLabs more than tripled its valuation to $11B. Decagon said its $4.5B financing tripled its valuation in less than six months.

Sierra reached $15.8B after a $950M financing, around 60% above its previous $10B valuation. Harvey went from roughly $8B to $11B and then $15.5B in less than a year.

Legora has moved even faster from a smaller base. Its valuation rose from about $1.8B to $5.6B while ARR crossed $100M.

The common thread is unusually fast commercial growth in categories where customers can connect AI spending directly to labor or infrastructure costs.

Startup	Earlier valuation	Current valuation	Approximate change
Cognition	$10B	$48B	4.8×
Fireworks AI	~$4B	$17.5B	~4.4×
Sierra	$4.5B	$15.8B	~3.5×
ElevenLabs	~$3.3B	$11B	~3.3×
Legora	~$1.8B	$5.6B	~3.1×
Decagon	$1.5B	$4.5B	3.0×
Harvey	~$8B	$15.5B	~1.9×
Synthesia	$2.1B	$4B	~1.9×

Are Generative AI valuations rising faster than the actual businesses?

Some Generative AI valuations are extremely aggressive, but several of the biggest recent increases have come with revenue growth that is just as unusual.

Cognition is the cleanest example. Its annualized run-rate revenue increased from $492M to nearly $900M while its valuation rose from $26B to $48B. Both figures increased by roughly 80%.

Fireworks announced more than $1B in annualized revenue at the time of its $17.5B financing. That puts its valuation at roughly 17.5 times run-rate revenue. A multiple like this reads differently next to the market's size, which we estimate ourselves from the ground up in our report covering the generative AI market.

Mistral's latest $24B valuation came alongside annual recurring revenue of roughly $1B, giving a multiple in the mid-20s. Gamma looks somewhat less expensive by comparison: $2.1B against more than $100M in ARR, or about 21 times revenue, with the additional distinction that Gamma said it was already profitable.

The biggest multiples still require aggressive assumptions about future growth. Cognition at $48B against nearly $900M of run-rate revenue is worth more than 50 times that figure. Some agent startups with lower disclosed revenue appear richer still.

Those multiples would look extreme for mature public software businesses. Investors are accepting them because the fastest AI companies are adding revenue at rates mature software companies rarely achieve.

The real test over the next few years is whether those growth rates remain extraordinary after the easiest early adopters have already bought the products.

Do Generative AI acquisitions prove these startup valuations are real?

Completed acquisitions give us some of the strongest valuation evidence in Generative AI because a buyer actually agreed to pay the stated price.

Cursor is the biggest example in our dataset. Bloomberg reported that SpaceX completed its $60B acquisition of the AI coding company after an earlier agreement between the two businesses.

Several other companies in the ranking are also anchored to acquisition values rather than theoretical private-market estimates. Moveworks appears at roughly $2.9B, Sana at $1.1B and Lakera at $202M.

Acquisition prices still need interpretation. A strategic buyer can justify paying more than ordinary financial investors if the target brings scarce talent, technology, customers or products that fit a larger platform.

Cursor is especially useful because its transaction came after AI coding had already produced strong revenue growth across several companies. Cognition's $48B financing gives us a separate market reference in the same category.

A completed sale gives more confidence than a rumor or comparables-based estimate, although it does not mean every similar startup could command the same multiple.

Does raising more money mean a Generative AI startup is worth more?

Generative AI funding and valuation are closely linked among frontier-model labs, but the relationship becomes much weaker among application startups.

OpenAI has raised an extraordinary amount of capital because training models and securing compute requires huge upfront spending. Anthropic's latest round alone was $65B. xAI has followed the same capital-intensive pattern.

Applications can reach large valuations with much smaller amounts of funding. Gamma announced a $2.1B valuation after raising $68M in its Series B and roughly $90M overall. HeyGen is estimated at $3B–$4.5B after raising about $75M. Photoroom sits around $550M–$800M with roughly $64M raised.

The reverse happens too. Inflection AI raised around $1.5B historically but is now estimated at only $600M–$1.2B following major changes to the company. Adept raised more than $400M but sits around $300M–$600M in our dataset.

Funding tells us a lot about how capital-intensive the company's strategy has been, but much less about how valuable the remaining business is today.

For application startups, revenue and customer adoption usually tell us more than cumulative funding alone.

Are early Generative AI stars like Jasper and Stability AI being left behind?

Several of the companies that defined the first Generative AI boom are now worth far less than newer coding, agent and vertical-AI startups.

Jasper is currently estimated around $700M–$1.1B. Stability AI sits around $450M–$550M. Character.AI is around $900M–$1.1B.

Compare those figures with Cognition at $48B, Sierra at $15.8B, Harvey at $15.5B, ElevenLabs at $11B and Decagon at $4.5B. Early leaders that were later overtaken invite the question of what stalled startups had in common, which we study in our market report about generative AI.

The change reflects how quickly basic generation features became available everywhere. Writing assistance is now embedded in ChatGPT, Claude, Gemini, Microsoft products and countless SaaS applications. Image generation has followed a similar path as high-quality models became easier to access.

The companies climbing fastest today are generally attached to a more expensive workflow or a harder technical layer. Coding, legal work, customer support and inference infrastructure all give vendors a clearer path to capturing large budgets.

Early visibility helped create the first generation of AI unicorns. These days, investors appear much more interested in who controls valuable work after the novelty of generation wears off.

What does the Generative AI valuation ranking tell us about where the market is going?

The Generative AI ranking increasingly points toward three groups: trillion-dollar frontier labs, a small group of $10B–$60B infrastructure and application leaders, and a much wider market of specialized companies below them.

The second group is the biggest change. Cursor at $60B, Cognition at $48B, Fireworks at $17.5B, Sierra at $15.8B, Harvey at $15.5B and ElevenLabs at $11B show that enormous value can now be created without building the most powerful general-purpose model.

Those companies tend to control something concrete. Cursor and Cognition sit inside software development. Harvey owns legal workflows. Sierra automates customer interactions. ElevenLabs has become a major voice platform. Fireworks operates the infrastructure that other AI companies use to run models.

Creative AI has also produced real scale, but its valuations remain lower. Suno, Runway, Synthesia and Luma AI are worth billions rather than tens of billions.

The ranking suggests that the market is becoming less impressed by generation for its own sake. The biggest application valuations now tend to follow products that automate costly work or become infrastructure for other AI businesses.

What is the final verdict on the most valuable Generative AI startups?

Generative AI has already created an extraordinary amount of private-company value, but the market is much more concentrated than a list of 100 startups initially suggests.

OpenAI and Anthropic now occupy a near-trillion-dollar tier, depending on whether we use completed financings or current fundraising discussions. xAI is a distant third at $250B. The rest of the market begins another large step lower, with Cursor at $60B and Cognition at $48B.

The biggest surprise is how quickly the application layer has matured. AI coding has produced two companies around $50B or more. Legal AI has produced Harvey at $15.5B and Legora at $5.6B. Voice AI has ElevenLabs at $11B. Customer-service agents have Sierra at $15.8B and Decagon at $4.5B. AI infrastructure has Fireworks at $17.5B and Together AI at $8.3B.

There is also a clear split between the first Generative AI wave and the current one. Earlier stars built around writing, image generation or consumer chat are generally much lower in the ranking. The companies gaining value fastest now are tied to software engineering, professional work, enterprise automation and infrastructure.

Some valuations remain much firmer than others. Anthropic's $965B, Harvey's $15.5B, Cognition's $48B, Fireworks' $17.5B and ElevenLabs' $11B all come from recent completed financings. OpenAI's $1.2T-plus figure is still an active fundraising discussion. Other companies farther down the ranking depend on revenue multiples, comparable transactions or broader operating evidence.

Even after allowing for that uncertainty, the picture is clear. Frontier-model labs still capture the overwhelming share of Generative AI value, while the best applications have moved well beyond the old "AI wrapper" stage and become major software companies in their own right.