What Investors Look For in an AI Startup at Seed and Series A

Short answer. At seed, investors can back a founder and a product wedge worth testing. At Series A, revenue, margin, retention and capital efficiency give them more evidence to assess. Use comparisons for the company you are building; foundational-model financing and application-company financing can differ sharply.

Headline AI funding figures can mix stages and business models with very different capital needs. A useful benchmark names the population, period and metric before a founder uses it to plan a raise.

The two-tier market: why AI money looks easy and mostly is not

In Q1 2026, more than 60 cents of every venture dollar tracked on Carta's platform went to AI companies, the highest share on record. That aggregate covers companies with very different capital needs.

An AI foundational model startup at Series A can raise at a median valuation around $300M, while a non-AI startup at the same stage sits closer to $55M.

That $300M comparison concerns foundational-model companies. An application company built on an existing model may depend on workflow, data or distribution advantages instead. Choose comparisons reflecting its product and delivery economics.

If you are building on an existing model, use stage- and product-relevant comparisons. The following figures label their populations and periods so you can judge which gives useful context for your raise.

What do investors look for in an AI startup at seed?

Carta's primary seed-round data gives a $24M median post-money valuation in Q4 2025, versus $18M a year earlier and $16M two years earlier. Separately, Peter Walker's December 1, 2025 Carta post gives 19.5 percent median seed dilution for software startups only. That dilution figure is not an all-sector Q4 2025 statistic. Do not combine separate medians into a supposed typical transaction.

Carta's Q1 2026 pre-seed report puts AI at around half of pre-seed dollars among US companies on its platform. That is not a Series A share. In his May 25, 2026 podcast interview, Walker described a narrow cohort of AI infrastructure and foundational-model seed rounds at $160M to $200M post-money. Use that as context for the split, not as an application-company valuation target.

In that dated interview, Walker reported median seed headcount of four employees excluding founders and a one-year seed-to-Series-A graduation rate around 10 to 11 percent for the 2025 cohort, versus 4 to 5 percent for 2022 and 2023. Those are his interview estimates. A seed round buys time to test a specific demand and delivery model; it does not promise a Series A on a fixed timeline.

What do investors look for in an AI startup at Series A?

Median primary-round Series A post-money valuation reached $78.7M in Q4 2025, up 37 percent year over year from $57.5M, in Carta's valuation report. Walker's separate May 2026 interview described Series A raises around $13M to $15M, versus $8M to $10M a few years earlier. Keep the dated chart and interview estimates separate.

Keep the revenue metric precise when choosing your comparison.

SVB's H1 2026 report, page 28 gives a $2.5M median trailing-four-quarter total revenue for US technology companies raising Series A in pooled 2024 to 2025, excluding extensions.

That includes recurring and one-time revenue, not ARR, and it is not AI-only. The middle half of this pooled Series A population raised with roughly $1M to $6.5M in revenue. Page 29 separately reports a 2025-only median of $3.3M versus $1.6M in 2021. These are different charts, not one current minimum.

The podcast's companion article cites $3.5M as ARR. That is not the metric shown in the SVB charts above. The source disagreement is clear; the precise origin of the circulating number is not established here.

Carta's Q2 2025 report put the median seed-to-Series-A interval at 616 days. Plan runway around a range of outcomes rather than assuming a one-year graduation. Walker's May 2026 interview estimated median Series A headcount at fifteen to seventeen employees, down from roughly forty-five a few years earlier. That is a dated staffing comparison, not a required team size.

Sources: Carta, Q4 2025 primary-round valuations; SVB, H1 2026 report, page 28; Peter Walker, Product Market Fit Show, May 25, 2026. The table labels each period and population; the interview estimates and SVB pooled revenue distribution are separate sources.

The four things investors actually look for

These four questions organize the evidence for diligence. Their weight changes with stage and investor strategy; answering them does not guarantee a round.

1. Revenue, correctly labelled. ARR and total revenue are different metrics. Keep recurring revenue, one-time sales and booked or collected amounts traceable to their records. Builder.ai's May 2025 insolvency, reported by the Financial Times, is a cautionary case in financing and revenue scrutiny. It does not establish that the company had no technical problems or that one revenue error alone caused the collapse.

2. Gross margin with a cost plan. ICONIQ's January 2026 executive survey reported average AI product margin of 41 percent for 2024 and 45 percent for 2025, and projected 52 percent for 2026. Its separate July 2026 edition projects 53 percent for 2026. Keep the editions separate. January's cost breakdown puts model inference at about 20 percent of product cost pre-launch and 23 percent at scale; those are shares of product cost, not revenue. Compute economics explain the exposure, while pre-raise pricing work prepares your account costs, allowances and rates. Explain your own trajectory rather than using a survey average as a minimum.

3. Retention that holds past the free trial. ChartMogul's December 2025 analysis classified 3,500 software companies, including roughly 200 AI-native businesses. Among companies with at least $250K ARR, AI-native median gross revenue retention was 40 percent and net revenue retention 48 percent. By monthly price, the AI-native figures were 23 percent GRR and 32 percent NRR below $50, versus 70 percent GRR and 85 percent NRR above $250. These are annualized revenue measures, not customer retention. AI-native GRR rose from 27 percent in January to 40 percent in September 2025. The price comparison is an association, not proof that raising prices alone fixes churn.

4. Capital efficiency with a stated growth strategy. Bessemer's research on roughly twenty selected high-growth AI companies described two archetypes. "Supernovas" reached about $40M ARR in year one and $125M in year two, with roughly $1.13M ARR per employee, but margins near 25 percent or sometimes negative. "Shooting Stars" reached nearer $3M ARR in year one, with margins near 60 percent. Bessemer proposed Q2T3 (quadruple, quadruple, triple, triple, triple) as a growth benchmark and noted that five-year data was not yet available. Those are selected success patterns, not minimum founder targets. Separately, High Alpha's 2025 SaaS benchmarks put upper-quartile ARR per employee at $350K in the $20M to $50M ARR band and roughly $397K above $50M. Its AI-native growth comparison reported 110 percent median growth for AI-native companies at $1M to $5M ARR versus 40 percent for companies adding AI features. That comparator is not every traditional SaaS company. Choose a growth and margin plan your funding can support, and name who owns whether the AI product earns revenue.

A demo tests product capability. Revenue quality, margin and retention answer additional questions about what it costs to turn that capability into a business.

Enterprise buyer context

BCG's October 2024 research reported that 74 percent of surveyed companies had not shown tangible AI value. McKinsey's November 2025 survey found 88 percent of respondents reporting AI use somewhere in their organizations but 39 percent reporting enterprise-level EBIT impact. MIT NANDA's preliminary 2025 report, hosted by AI News, described weak returns from many enterprise generative AI pilots. These are different buyer studies with different methods and outcomes. They do not prove why an investor raised a startup's funding bar. They do give a founder reason to ask which customer outcome will support renewal.

What actually de-risks an AI startup

Use data you have the right to use, and show why it improves the product for a specific buyer. Embed the product in a workflow where the customer can measure its contribution. Keep the team and delivery costs consistent with the growth plan, and report retention alongside acquisition. These are operating choices you can test; none makes an AI application impossible to compete with.

Frequently asked questions

What do investors look for in an AI startup?

At seed, investors can back a founder and a promising product wedge. At Series A, they have more evidence to assess: correctly labelled revenue, a margin and cost plan, customer retention and capital efficiency. The weight of each depends on the stage, business model and investor strategy.

How much revenue do you need to raise a Series A in 2026?

No single minimum follows from SVB's data. Page 28 of its H1 2026 report gives a $2.5M median trailing-four-quarter total revenue for US technology Series A raises in pooled 2024 to 2025, excluding extensions, with the middle half between roughly $1M and $6.5M. Page 29 separately gives a 2025 median of $3.3M. These are total revenue including recurring and one-time sales, not ARR or AI-only requirements.

What is a good gross margin for an AI startup?

Explain the margin your own product has and the costs behind its trajectory. ICONIQ's January 2026 executive survey reported 45 percent average AI product margin for 2025 and projected 52 percent for 2026. Its July edition projected 53 percent for 2026. These are survey figures and forecasts, not a universal investor cutoff. January's inference figures of 20 to 23 percent are shares of product cost, not revenue.

Why does AI have worse retention than SaaS?

ChartMogul found lower revenue retention in its AI-native cohort, especially low-priced products. Below $50 per month, gross and net revenue retention were 23 and 32 percent; above $250 they were 70 and 85 percent. These are annualized revenue measures, not customer retention. Test your own cohorts and workflow use before assigning a cause.

---

Sources

Valuation, round size, and headcount data: Carta, "State of Private Markets: Q1 2026" (published May 29, 2026); Carta, "Record-setting early-stage valuations"; Carta, "State of Pre-Seed: Q1 2026"; Carta, "Series A Funding Slides in Q2 2025" (the 616-day figure is dated to Q2 2025). Seed and Series A headcount, elite seed valuations, and graduation rate commentary: Peter Walker, Head of Insights at Carta, Product Market Fit Show, May 25, 2026, a podcast interview rather than a published Carta chart. Revenue benchmark: Silicon Valley Bank, "State of the Markets Report H1 2026," pages 28 and 29. Gross margin and inference cost: ICONIQ Growth, "State of AI: Bi-Annual Snapshot" (January 2026, survey of roughly 300 software executives). Growth archetypes: Bessemer Venture Partners, "The State of AI 2025" (August 2025). Retention: ChartMogul, "The SaaS Retention Report: The AI Churn Wave" (2025, data through September 2025). Capital efficiency benchmarks: High Alpha, 2025 SaaS ARR-per-employee benchmarks and High Alpha, AI-native growth comparison (800-plus founders and operators). Builder.ai insolvency: Financial Times reporting, May 2025. Enterprise adoption backdrop: BCG, "Where's the Value in AI?" (October 2024); McKinsey, "The State of AI in 2025" (November 2025); MIT NANDA, "The GenAI Divide" (AI News-hosted PDF) (2025, preliminary and non-peer-reviewed).

---

About the author

Jeff Brokaw co-founded the AI fintech FaaStrak, which went from zero to $1M ARR in nine months. He is a Certified Chief AI Officer and technical-commercial operator who builds AI systems and their commercial models. He was the first marketing hire twice at companies tied to $850M+ in associated exits, and reframed institutional Bitcoin mining as energy-backed digital infrastructure; a $114M institutional raise came together in under thirty days.