Find out how fundable you are before you send a deck.
Your market decides whether investors are looking. Your evidence decides whether they bite.
11 questions , 4 minutes , Your email only at the end
A score, one of five verdicts, and three fixes.
Step 1
Which market are you actually in?
Pick the closest one. Capital is not spread evenly across these, and the gap is the point.
Where the numbers come from, and who built this
86%
AI companies took 86% of US venture dollars in the first half of 2026. That is $355.9 billion out of $412.7 billion.
Roughly nine in ten US venture dollars went to rounds of $100 million or more.
That doesn't mean every AI startup is fundable. It means capital is loud at the top and quiet in the middle. This tells you which end you're at.
Built by Farzad Khosravi, who has worked with 500+ founders and was fractional COO at Humoniq through an $8.5M seed round.
Market ratings researched July 2026 from PitchBook data and published theses at Sequoia, a16z, Bessemer and General Catalyst.
The markets you are being scored against
Nine markets took the bulk of 2026 venture dollars. Four of them gate on access a founder cannot acquire between now and the next raise. Read the second column before the third.
| Market | Fundability now | What the fundable startup looks like | What investors will want to see | Founder bar and recent evidence |
|---|---|---|---|---|
| Vertical AI systems of action | Very high | Runs a central, expensive workflow rather than assisting around its edges. It dispatches, underwrites, reconciles, schedules, sells, files, procures or resolves cases. Ideally it can charge against labor, revenue or EBITDA rather than the existing software budget. | A live workflow, several serious design partners, and evidence that the product completes work rather than merely generating suggestions. A credible expansion from one wedge into the operating layer for the industry. | Strong domain access matters more than generic AI credentials. a16z's Probook investment centered on owning the dispatch decision in home services. Hilbert begins with data plumbing and expands into growth execution. Bessemer and General Catalyst are explicitly pursuing software that absorbs service and labor budgets. |
| AI infrastructure bottlenecks | Very high, elite-gated | Removes a bottleneck in inference, memory, agent orchestration, evaluation, reliability, multimodal data or networking. It must solve something that becomes more painful as model usage scales. | Real benchmarks, meaningful open-source adoption, usage by sophisticated technical teams, or credible hyperscaler and frontier-lab relationships. A generic "developer platform for agents" will not cut it. | The technical bar is as high as it gets. Inferact was built around the maintainers of vLLM. Protege was founded by repeat data-infrastructure entrepreneurs and focused on hard-to-access real-world training data. Bessemer argues value is moving from model training toward inference, continual learning, memory and reliability. |
| Defense autonomy and sovereign technology | Very high, access-gated | Replaces an existing military capability with something cheaper, faster, more autonomous or more resilient. Strong areas include command and control, electronic warfare, communications, navigation, autonomous systems and industrial capacity. | Operational demonstrations, a credible procurement pathway, manufacturability, and performance under adverse conditions. A defense-themed pitch deck without operator validation is worthless. | The ideal team combines military operators, technical depth and procurement knowledge. NODA's founders had direct joint-fires and defense-software experience before raising a $25 million Series A. General Catalyst backed Constelli after years of electronic-warfare development and operator use. |
| AI data-center power, networking and cooling | Very high, capital- and relationship-gated | Shortens time to energization, raises GPU utilization, improves networking throughput, manages distributed energy, or solves cooling and grid-interconnection constraints. The value should be measured in megawatts, deployment months or compute output. | Utility, hyperscaler or data-center partnerships. Technically credible hardware or infrastructure. Contracts or deployments, and a believable path through long enterprise sales cycles. | A difficult sector for outsiders. Bessemer identifies generation, grid equipment, energy orchestration, networking and cooling as the major constraints, and led both Verse's $54 million Series B and DriveNets' $410 million Series D. Lightspeed and a16z put $500 million into Nexthop at a $4.2 billion valuation. |
| Agentic cybersecurity | High | Provides identity, permissions, runtime policy, monitoring or autonomous remediation for fleets of AI agents. The stronger companies take corrective action rather than producing another dashboard full of alerts. | Deep enterprise integrations, low false-positive rates, reduced remediation time, and control over what agents can access or execute. Security credibility is mandatory. | Bessemer calls agent security a defining cybersecurity problem for 2026, while a16z is backing infrastructure for controlling enterprise agents. Bessemer also led QIZ's $17 million seed around post-quantum cryptographic posture and remediation. |
| Physical AI and industrial robotics | High, brutally elite-gated | A vertically integrated system that performs a valuable task in a messy real-world environment. The winning company usually owns the model, hardware, deployment process and data flywheel rather than selling a general-purpose robotics model. | Real deployments, reliability data, improving unit economics, and a privileged source of training data. A polished lab demo is not enough. | Requires rare hardware, ML and operational talent plus substantial capital. Bessemer expects near-term value to accrue primarily to full-stack companies. Mind Robotics' Rivian deployment gives it a live manufacturing data environment. Waymo's scale shows how operational and regulatory advantages compound. |
| Healthcare operations and data infrastructure | High | Automates medication access, reimbursement, care navigation, clinical administration or fragmented data exchange. The best wedge sits where delays and administrative labor produce measurable financial or clinical harm. | EHR, payer, provider or pharmaceutical integrations. Reimbursement clarity, compliance, and measurable improvements in turnaround time, workload or access. | Founder-market fit and industry relationships are unusually important. General Catalyst's Forus investment focused on medication access across the fragmented payer, provider and pharma ecosystem. Baba combines AI with human advocates and Medicare reimbursement. Protege targets proprietary, regulated healthcare and enterprise data. |
| Stablecoin and AI-native financial infrastructure | High, selective | Becomes a system of record or a transaction rail for cross-border payments, treasury, compliance, banking infrastructure or machine-to-machine commerce. It should remove structural friction rather than adding crypto to an existing product. | A regulatory pathway, banking partners, transaction activity, distribution, and evidence that stablecoins materially improve settlement cost or speed. | Deep financial and regulatory expertise is essential. Bessemer argues stablecoins have crossed into infrastructure, while a16z sees stablecoins plus AI automating global financial operations. Stitch was funded around a core banking API and strong regional regulatory relationships. |
| Consumer AI | Selective, traction-heavy | Creates a genuinely new habitual behavior, identity, relationship or entertainment experience. It needs organic distribution, retention or network effects rather than better generated content. | Exceptional cohort retention, frequency, virality or monetization. Consumer founders cannot substitute a compelling market narrative for actual user behavior. | Consumer and enterprise fundability are different games. AI lowered the cost of building consumer products. That makes differentiation and defensibility harder, so investors price the round off behavior. |
What a fundable company looks like across all those markets
The market decides whether you get the meeting. These ten decide whether you get the round. They are the ten questions the calculator above asks you.
| Dimension | Fundable version | Weak version |
|---|---|---|
| Market narrative | "This transition is inevitable, and this bottleneck must be solved." | "The market is large and AI is growing." |
| Product position | Owns the central decision, transaction or execution workflow. | A copilot, dashboard or peripheral feature. |
| Economic value | Replaces labor, raises revenue, frees capacity or controls serious risk. | Saves employees a few minutes. |
| Wedge | Narrow, painful and immediately valuable. | Broad platform before proving one job. |
| Expansion | Wedge → system of action → system of record or network. | Remains a narrow point solution. |
| Moat | Proprietary data loop, network effects, workflow depth, regulation, distribution, hardware or switching costs. | Prompt engineering and public data. |
| Founder credibility | The team has lived the problem, built the underlying technology, or already knows the buyers. | Smart generalists discovering the industry through customer interviews. |
| Pre-seed proof | Working deployment, serious design partner, open-source adoption, procurement interest, or an unusually strong technical result. | Waitlist registrations and nonbinding letters of intent. |
| Pitch legibility | One sentence that makes the company's importance obvious. | Seven categories and twelve buzzwords. |
| Timing | A recent technological, regulatory or economic change makes the company newly possible or necessary. | A long-standing problem with no convincing reason it gets solved now. |