AI Startup Financial Model: Forecast Burn, Runway, and Funding Needs

1. Start with a baseline you can defend

Choose a recent completed month. Reconcile starting cash to the bank balance and explain restricted or unavailable cash separately. Use actual customer counts, collected cash, revenue, and supplier bills as distinct inputs. A signed contract, recognized revenue, and money received are different events; a forecast should not silently treat them as interchangeable.

Maintain an assumptions register with four columns: input, value, evidence, and owner. A price can come from your billing system; acquisition may come from a recent cohort; a future hire remains an assumption until agreed. Record the date and keep the model version used for each investor conversation.

Sequoia’s business-plan guidance includes both the business model and financials. The practical objective here is to make the connection between those two understandable, rather than to add more spreadsheet tabs.

For a pre-revenue startup, label the model as a launch plan. Do not invent recurring revenue or customer-retention history. Our tool’s percentage-based acquisition cannot grow from zero customers; use a clearly stated initial customer assumption, or maintain an explicit launch-sales schedule separately.

2. Model pricing-plan economics before projecting growth

A single average revenue figure can hide a plan whose customers consume more than they pay for. Model subscription revenue, included usage, expected usage, overage prices, and variable delivery costs separately for each plan. Count retries and unsuccessful work in the cost base.

Contribution at plan level excludes shared fixed costs. Company gross margin includes relevant fixed delivery costs and delivery staff; net profit requires the rest of the company’s expenses as well. Keep cost scope explicit when you present a percentage.

Bessemer’s AI pricing playbook discusses the relationship between AI delivery costs and monetization. For your model, test the actual allowance, overage, and usage pattern rather than borrowing a universal gross-margin target.

Stress-test customers who consume twice the expected units. If you do not charge overages, incremental usage produces cost without incremental subscription revenue. If you do charge, test whether collection, discounts, or caps prevent the full modeled amount being realized. Our AI subscription pricing guide goes deeper on these tradeoffs.

3. Tie the hiring timeline to a measurable milestone

Give each planned role a start month, headcount, fully loaded monthly cost, one-time setup cost, and accountable milestone. Include employer costs and benefits in the monthly figure. Put existing employees in baseline expenses so future hires are not counted twice.

For example, an engineer might own a reliability release, while a delivery specialist supports onboarding capacity. Specify the release acceptance criteria or onboarding capacity you need. A role alone is a spending category; a role attached to a measurable deliverable helps explain the use of funds.

Do not assume hiring automatically creates revenue. In this model, customer growth is a separate assumption. Moving a hire earlier increases cash outflow without inventing an acquisition benefit. If there is evidence of a growth benefit, adjust that assumption separately and explain why.

4. Worked example: two pricing plans and two hires

Assume the following fictional AI startup. All amounts are USD. Month 0 is the baseline; its expenses are not deducted again from opening cash. Revenue is collected and costs paid in the same month.

Starting cash is $200,000. Shared fixed delivery cost is $1,500 per month. Baseline operating expenses are $18,000 per month and grow 1% monthly. Usage per customer grows 3% monthly; inference unit costs fall 1% monthly. Prices and allowances stay fixed. The horizon is 24 months with a three-month reserve based on end-period net burn. These are chosen assumptions, not market forecasts.

• Engineer: starts in month 3; $8,000 monthly plus $2,000 setup, classified as operating expense.
• Delivery specialist: starts in month 6; $4,000 monthly plus $1,000 setup, classified as COGS.

The baseline produces $13,900 revenue, $5,300 COGS, and 61.9% gross margin. Baseline net burn is $9,400. The outcomes below are calculated using the same engine as the public tool.

“Slower acquisition” reduces each plan’s monthly acquisition from 12% to 6%, keeping churn at 2% and every other assumption unchanged. “Both hires start in month 1” changes only their start dates. These are custom sensitivity cases, not the calculator’s bundled upside/downside presets. Forecast customer counts can be fractional expected values; displayed dollars are rounded.

Notice what changes cash: plan mix and allowances determine revenue and variable cost; timing determines when hiring cash leaves the business. A future positive margin cannot pay a bill due before financing closes.

5. Turn the cash forecast into a funding ask

The tool uses end-period monthly net burn multiplied by the chosen buffer months as its reserve. It includes existing cash and assumes any additional funding is available before the cash shortfall. If the business is cash-generative at the end of the forecast, that particular reserve rule becomes zero; you may still need a separately chosen minimum cash balance.

Do not label the computed amount “the required round size” without qualification. Add omitted capex, debt service, taxes, financing fees, and collection timing to a full cash schedule. Compare alternative milestone dates and reserve policies before selecting an ask.

YC’s seed fundraising guide recommends tying the raise to a believable plan and considering different amounts raised. Its historical dollar figures are not used as current benchmarks in this article.

Test a delayed financing close separately

Keep a no-new-funding cash schedule. Then add the proposed round in the expected closing month and repeat with the close shifted later. For an illustrative timing-only check, if you have $40,000 at the end of month 4 and burn $15,000 in each following month, you finish months 5, 6, and 7 at $25,000, $10,000, and −$5,000 before financing. A round arriving in month 8 cannot resolve the earlier shortfall without another action.

The current public calculator does not schedule financing inflows. Run that timing check in a separate cash sheet; do not count a prospective round as opening cash. Decide in advance which discretionary hires or expenses can move if the close slips.

6. Prepare the numbers behind your investor slide

Use one financial summary slide supported by the model, not a screenshot of dozens of rows. Distinguish the last actual month from the forecast and specify currency and time horizon.

• Funding ask and use of funds: amount, planned hires, major delivery costs, and contingency.
• Milestone: target date and measurable evidence, such as retained paying customers, reliability, or repeatable onboarding capacity.
• Economics: revenue, gross margin, net burn, and cash runway, with cost scope stated.
• Assumptions: acquisition, churn, customer usage, unit cost, and hiring start dates.
• Downside response: what changes if growth slows, usage is heavier, or fundraising takes longer.

Keep a dated assumptions sheet and monthly actual-versus-plan comparison available. When a number changes, explain the driver: price, customer count, usage intensity, supplier cost, or headcount timing. Avoid implying that forecasts demonstrate retention, product-market fit, or guaranteed investor returns.

Use the AI Unit Economics Calculator to examine cost per successful outcome, and AI Invoice Reconciliation to investigate usage-versus-invoice gaps. Both support the input work; neither substitutes for complete financial records.