How is the funding in the agentic AI market now?

Are AI agents still getting funded, or is the hype already cracking?

Agentic AI funding is still accelerating today, but the average agent startup is not living in the same market as the winners.

The simplest number already shows the split. In Q1 2026, agentic AI startups raised about $2.66 billion across 44 rounds. In the comparable 2025 period, they raised about $1.09 billion across 71 rounds. So total capital grew about 2.4x, while the number of rounds fell by roughly 38%. The average disclosed round jumped from about $15 million to about $60 million, close to a 4x increase.

That is a very specific funding signal. Investors are concentrating bigger checks into fewer companies. This is why the market can feel hot from the outside and much colder if you are a generic agent founder raising a mid-stage round.

So, all things considered, agentic AI funding is not fading but rather becoming more concentrated, more expensive, and much less forgiving.

If you want more recent data on this point, please see our latest agentic AI market report.

Is “agentic AI” still enough to raise money?

No, “agentic AI” alone is now too vague to carry a funding round.

The recent numbers make that pretty clear. The biggest checks are attached to specific jobs: Sierra raised $950 million for customer experience agents, Parloa raised $350 million for enterprise customer operations, Cognition raised more than $1 billion for coding agents, Hippocratic AI raised $126 million for healthcare agents, Armadin raised about $190 million for cybersecurity agents, and Coram AI raised $35 million for physical security agents.

Across those six examples alone, we get about $2.65 billion of recent capital. The common point is that each company points to a known budget: support, engineering, healthcare operations, security, or physical-site protection.

So we can conclude that the fundable unit has changed.

Are vertical agents winning the funding race?

Yes, vertical agents are currently winning because they make the funding math easier.

Look at the concentration by workflow. Customer experience alone has Sierra at $950 million, Parloa at $350 million, and Wonderful at $150 million. That is $1.45 billion across just three recent rounds, with combined reported valuations above $20 billion. Coding agents have Cursor at $2.3 billion and Cognition at more than $1 billion, so two coding rounds alone represent more than $3.3 billion of capital and more than $55 billion of combined valuation.

That is a different funding universe from broad “agent platform” startups. A vertical agent can tell investors exactly what success looks like: fewer support tickets, more resolved calls, more code shipped, faster incident response, lower admin cost, or fewer human handoffs.

At the end of the day, the market is rewarding agents that can say, “we own this workflow.” The broader the product sounds, the harder the funding story gets.

If you want more recent data on this point, please see our latest agentic AI market report.

Is customer support the first real agentic AI funding category?

Yes, customer support is probably the first clean enterprise funding category for agentic AI.

The numbers are unusually strong. Sierra raised $950 million at more than $15 billion valuation. Parloa raised $350 million at a $3 billion valuation, bringing total capital raised to more than $560 million. Wonderful raised $150 million at a $2 billion valuation only four months after a $100 million Series A, and its total funding passed $280 million.

That means three customer-support or customer-experience agent companies raised about $1.45 billion recently. More importantly, the category has operating proof, not just pitch-deck proof. Nubank’s 2026 customer-support agent work reported a 37 percentage-point improvement in AI transactional NPS and a 29 percentage-point gain in self-service rate in one deployment.

That is why customer support is so fundable now. It combines huge volume, measurable outcomes, large labor cost, and clear buyer urgency. In a market full of agent demos, customer support gives investors something much easier to underwrite: if the agent resolves more issues without hurting satisfaction, the value is visible.

Are voice agents becoming their own funding market?

Yes, voice agents have clearly broken out as a separate funding lane inside agentic AI.

The aggregate is stronger than it first looks. ElevenLabs raised $500 million at an $11 billion valuation and crossed more than $500 million in ARR. Vapi raised $50 million after reaching 1 billion calls. Gnani.ai raised $10 million while processing more than 30 million voice interactions daily across 12+ languages and 200+ enterprise customers. Arrowhead raised $3 million for voice sales agents in financial services. If we include Parloa’s $350 million customer-operations round, voice-related agent funding gets close to $913 million across these recent signals.

The reason investors care is that voice sits on top of a huge messy channel. Businesses still run sales, support, collections, insurance, banking, logistics, and healthcare interactions through calls. Chatbots never fully solved that. Voice agents can attack a bigger labor pool because they plug into the channel where customers already are.

Are coding agents overfunded, or is the revenue really there?

Coding agents are definitely expensive, but the best companies have enough revenue and usage signals to explain why investors are still paying up.

The two headline numbers are extreme. Cursor raised $2.3 billion at a $29.3 billion valuation. Cognition raised more than $1 billion at a $26 billion post-money valuation. Together, those two companies alone represent more than $3.3 billion of recent capital and about $55.3 billion of valuation.

The valuation math is aggressive, but not empty. Cognition reportedly reached about $492 million in annualized revenue, which implies a valuation multiple of roughly 53x ARR at $26 billion. That is very high, but it is at least anchored to real revenue. Cursor also reported 100x enterprise revenue growth in 2025 year-to-date when it announced the $2.3 billion round.

The usage side also matters. Recent GitHub studies found 24,014 merged agentic pull requests with 440,295 commits in one dataset, 33,000 agent-authored PRs in another, and a separate dataset with more than 142,000 agentic PRs. The messy part is that failure is still very real: one study found a 27.67% merge-conflict rate in processed agentic PRs, and another found high message-code inconsistency PRs had 51.7% lower acceptance rates.

So, finally, coding agents are both real and overheated. The leaders are raising like infrastructure companies because the usage is intense, but the technical evidence also says the category is not mature enough for every startup to deserve a premium valuation.

If you want more recent data on this point, please see our latest agentic AI market report.

Are sales agents quietly coming back?

Yes, sales agents are coming back, but the money is going to workflow depth, not “AI writes emails.”

The funding pattern is more specific than the old AI SDR hype. Rox AI reportedly reached a $1.2 billion valuation. Actively raised $45 million, bringing total funding to $68 million. Mega raised $11.5 million to automate marketing work for SMBs. Gushwork raised $9 million, bringing total funding to $11 million. Add Kana’s $15 million round and we get more than $80 million of disclosed recent GTM-agent funding, excluding Rox’s undisclosed round size.

The interesting part is the size gap. Rox gets unicorn pricing because it targets autonomous sales productivity at enterprise scale. Actively gets a meaningful Series B because it sells persistent account-level agents. Mega and Gushwork are smaller because they are attacking SMB marketing and AI-search discovery. Same broad GTM theme, very different capital intensity.

So we can say sales agents are investable again, but only when they move beyond copywriting. The funded story today is: find the account, read the signals, trigger the workflow, update the system, and move revenue forward.

Is healthcare agent funding still serious?

Healthcare agent funding is still serious, but the market is paying for controlled autonomy, not free-form automation.

Hippocratic AI raised $126 million at a $3.5 billion valuation and brought total funding to $404 million. Abridge raised $300 million at a $5.3 billion valuation and has raised about $800 million total. Together, those two healthcare AI companies represent $426 million of recent large-round capital and about $8.8 billion of combined valuation.

But the research signals make the market more complicated. A recent benchmark for long-horizon healthcare workflows found the best agent resolved only 28% of tasks, and strict multi-pass performance stayed below 20%. Another healthcare administration benchmark found the best computer-use agent achieved only 36.3% task success, even though some subtask scores were much higher.

That tells us why healthcare agents are funded differently from coding agents. Investors like the size of the opportunity, but they also know the product cannot just “act autonomously” across clinical and administrative systems. The fundable version of healthcare agents is safer, narrower, auditable, and full of approval gates.

Everything considered together, healthcare remains a major agentic AI opportunity, but the winning pitch is not “replace clinicians.” It is “remove administrative load without breaking trust.”

If you want more recent data on this point, please see our latest agentic AI market report.

Is cybersecurity becoming the next agentic AI funding pocket?

Yes, cybersecurity is becoming one of the clearest new funding pockets in agentic AI.

The recent capital is already meaningful. Armadin raised $189.9 million across Seed and Series A to build autonomous cybersecurity agents. WitnessAI raised $58 million for enterprise AI security and agent governance. General Analysis raised $10 million for agentic AI security infrastructure. Sycamore raised a $65 million seed round for a trusted enterprise agent operating system. Together, those four rounds add up to about $323 million.

The reason this is a real pocket, not just a theme, is that agents create a new security object. A chatbot answers. An agent can use tools, touch files, retrieve data, call APIs, and act inside systems. Recent security research shows why that matters: one penetration-testing study found that more than half of malicious prompts succeeded across tested agentic configurations, and another 2026 survey argues that agent security problems are structurally different from traditional software security.

So it looks like the next wave of agent funding is not only about building agents. It is also about controlling them.

Are physical-world agents finally getting funded?

Yes, physical-world agents are starting to get funded because they attach AI to high-value real-world operations.

The two recent signals are very different in size but point in the same direction. Coram AI raised $35 million Series B, bringing total funding to $66 million, and says it is deployed across more than 1,500 sites. Prometheus raised $12 billion at a roughly $41 billion valuation to build an “artificial general engineer” for complex physical products.

That is a huge range: one company is turning cameras, access systems, and emergency workflows into agentic security operations; the other is trying to compress physical product engineering across aerospace, medical devices, robotics, and manufacturing. But both are about the same deeper shift: agents are moving from screens into physical decision loops.

At the end of the day, physical-world agents are still earlier than customer support or coding. But the funding signal is now too large to ignore. Investors are starting to price agentic AI as a way to affect factories, buildings, engineering teams, and physical risk.

Are agent infrastructure startups still fundable?

Yes, agent infrastructure is still fundable, but only if it solves the deployment mess.

The funding here is not huge compared with coding or customer support, but it is very telling. Sycamore raised $65 million seed to build an operating system for autonomous enterprise AI. WitnessAI raised $58 million for AI security and governance. General Analysis raised $10 million for agentic security infrastructure. Vapi raised $50 million for voice-agent infrastructure after reaching 1 billion calls. Nimble and Trace also sit in this broader “make agents usable in production” bucket.

This tells us where the pain is. Companies do not only need agents but also to monitor them, evaluate them, control their tool access, secure their data flows, understand their cost, and explain their actions. That is boring compared with a flashy autonomous demo, but it is exactly what enterprise buyers need before they scale.

Is the agentic AI market becoming a mega-round market?

Yes, the agentic AI market is becoming a mega-round market at the top and a proof-heavy market everywhere else.

The numbers are pretty brutal. In customer experience, three companies recently raised about $1.45 billion. In coding, two companies raised more than $3.3 billion. In healthcare, two companies raised $426 million in large rounds and have more than $1.2 billion of total funding between them. In agent security and enterprise control, four companies raised about $323 million. Then Prometheus alone raised $12 billion for physical-world engineering AI.

This means the market headline is being written by a small group of category-defining companies. But Q1 2026 agentic AI also had only 44 rounds, down from 71 in the comparable 2025 period. So fewer companies are getting funded, while the visible winners are raising much larger checks.

As pointed out above, this is why the market can look euphoric and selective at the same time. If a company looks like the future owner of a workflow, capital is abundant. If it looks like a thin agent wrapper, the market is much colder.

If you want more recent data on this point, please see our latest agentic AI market report.

So, how is funding in the agentic AI market now?

Agentic AI funding is very hot today, but the market is now separating winners from tourists.

The strongest recent pattern is concentration. Capital is flowing toward companies that can prove work done: calls handled, support issues resolved, code shipped, healthcare workflows supported, threats detected, sites protected, or engineering cycles shortened. That is a much higher bar than saying “we build autonomous agents.”

The second pattern is that each hot segment has a different funding logic. Customer support is funded on ROI clarity. Voice is funded on call volume and labor replacement. Coding is funded on revenue intensity and developer usage. Healthcare is funded on workflow size but constrained by safety. Security is funded because agents create new risk. Physical-world agents are funded because the value per decision can be massive.

So, finally, the agentic AI funding market is not broadly easy anymore. It is still accelerating, but only for companies that can turn autonomy into a measurable business result.

Trend	Signals proving it
Funding is hotter but more concentrated	Q1 2026 agentic AI funding reached about $2.66B across 44 rounds versus $1.09B across 71 rounds a year earlier; average round size rose from about $15M to about $60M
“Agentic AI” alone is no longer fundable	About $2.65B of recent capital across Sierra, Parloa, Cognition, Hippocratic AI, Armadin, and Coram is tied to specific workflows, not generic agent platforms
Customer support is the cleanest enterprise wedge	Sierra, Parloa, and Wonderful raised about $1.45B combined; Nubank’s production work showed +37pp AI transactional NPS and +29pp self-service improvement in one deployment
Voice agents are now a separate funding lane	ElevenLabs, Vapi, Gnani.ai, Arrowhead, and Parloa point to roughly $913M of recent voice-related agent capital, with signals like 1B calls at Vapi and 30M daily interactions at Gnani.ai
Coding agents are real but richly priced	Cursor and Cognition raised more than $3.3B combined at about $55.3B combined valuation; Cognition’s reported $492M ARR implies roughly 53x ARR
Sales agents are back, but only with workflow depth	Rox reportedly reached $1.2B valuation; Actively, Mega, Gushwork, and Kana show more than $80M of disclosed recent GTM-agent funding excluding Rox’s undisclosed round size
Healthcare agents are serious but safety-constrained	Hippocratic AI and Abridge represent $426M of recent large-round capital and about $8.8B of combined valuation; benchmarks still show low end-to-end healthcare task success
Agent security is becoming a real funding pocket	Armadin, WitnessAI, General Analysis, and Sycamore raised about $323M combined around securing, governing, or operating enterprise agents
Physical-world agents are moving into the funding map	Coram raised $35M and runs across 1,500+ sites; Prometheus raised $12B at about $41B valuation for AI that helps engineer physical products
The market is now barbell-shaped	Mega-rounds define the category at the top, while early rounds still happen for narrow wedges; the weak zone is generic mid-stage agent companies without proof of workflow ownership

Who is the author of this content?

NEW MARKET PITCH TEAM

We track new markets so founders and investors can move faster

We build living "market pitch" documents for emerging markets: AI, synthetic biology, new proteins, and more. Instead of outdated PDFs or hallucinated LLM answers, our clients get a clean, visual, always-updated view of what's really happening: key players, deals, regulations, and signals that matter. Learn more about us.