How does the forecast combine the three growth factors?

Each month it multiplies the previous month's cost by (1 + user growth) × (1 + tokens-per-user change) × (1 + price trend). Compounding all three captures how usage and price changes stack on top of growing user numbers.

Should I use positive or negative price trends?

LLM list prices have generally fallen as competition increases, so a small negative trend (e.g. -2%/month) is often realistic. If you expect to move to a more capable, pricier model, use a positive trend instead.

Does this account for caching or batching discounts?

Not directly. Model the net effect by lowering your tokens-per-user change rate or applying a one-off reduction to your current monthly cost baseline before forecasting.

Is my data sent anywhere?

No. The forecast runs entirely in your browser. Nothing you enter is uploaded, stored, or logged.

Why does cumulative cost matter more than month 12?

Your annual budget is the sum of all twelve months, not just the final month. A workload that grows steadily can cost far more in total than its end-of-year run rate suggests.

What is the LLM Cost Forecasting Tool?

Build a 12-month LLM cost forecast that combines monthly user growth, changing tokens-per-user, and provider price trends — see projected monthly and cumulative spend so you can budget your AI bill before it surprises you. It runs free in your browser on Gera Tools, with nothing uploaded.

LLM Cost Forecasting Tool

Name: LLM Cost Forecasting Tool
Creator: Gera Tools
License: https://creativecommons.org/licenses/by/4.0/

Forecast your LLM bill 12 months out

AI costs rarely stay flat. Your user base grows, each user sends more (or fewer) tokens as features mature, and provider list prices drift over time. This tool compounds all three forces into a clear 12-month projection so you can budget realistically instead of extrapolating a single flat line.

How the forecast works

Starting from your current monthly cost, each month is computed by applying three multipliers:

next_month = this_month
           × (1 + user_growth_rate)
           × (1 + tokens_per_user_change)
           × (1 + price_trend)

User growth scales the number of calls, tokens-per-user scales the size of each call, and the price trend captures provider pricing changes. Because the factors multiply, even modest monthly rates produce large swings over a year — 10% monthly user growth alone compounds to roughly a 2.9× increase by month 12 (eleven months of compounding from the month-one baseline).

Worked example — a growing AI writing assistant

Suppose you are running a writing assistant that costs $800/month now, with 2,000 monthly active users. You expect 8% monthly user growth, tokens per user to grow 3% per month as users engage more deeply with the feature, and provider prices to fall 1% per month due to competition.

Each month applies: × 1.08 × 1.03 × 0.99 = × 1.0997 (roughly +10% per month).

Month	Cost	Cumulative
0 (now)	$800	—
3	~$1,064	~$2,764
6	~$1,416	~$7,048
12	~$2,509	~$17,250

By month 12, the monthly run rate has more than tripled — but the annual total is $17,250, well above what a naive “current spend × 12” estimate of $9,600 would suggest. This is why the cumulative figure matters more than the month-12 snapshot when setting an annual budget.

Understanding the three input levers

User growth rate — The most powerful driver, because it compounds multiplicatively with the other two. A 5% monthly rate (modest, sustainable) compounds to 80% year-over-year; 10% monthly compounds to 214%. Use the growth rate you can actually defend to finance, not the optimistic scenario.

Tokens per user change — This captures feature evolution. Early on, users try short prompts; as they trust the product, they submit longer inputs or the product triggers more background calls. A positive 2–5% monthly rate is common in growing products. If you plan to introduce summarization, retrieval-augmented generation, or multi-turn conversations, this rate should reflect those expansions.

Price trend — LLM list prices have generally fallen as model efficiency improves and competition increases. A small negative monthly trend (for example, -1% to -3%) is often realistic for planning purposes. If you expect to upgrade to a premium model mid-year, model that as a positive trend or introduce a step-change at the upgrade month.

Common modelling mistakes

Projecting a single flat line. Extrapolating current cost × 12 ignores compounding. The tool shows you the non-linear shape, which is always the actual shape.

Conflating user growth with cost growth. Not every new user drives the same cost. Power users consume 5–10× the tokens of casual users. If your growth is skewed toward heavy users (a sign of product-market fit), the tokens-per-user trend should increase, not stay flat.

Ignoring the cumulative spend. Month 12’s run rate looks manageable. The sum of all twelve months — what you actually spend — can be 30–40% higher than twelve times the final month because the early months are cheaper and still count toward the total.

Tips for a realistic model

Be conservative on growth. Sustained double-digit monthly growth is rare; taper it if you are modelling beyond an early launch phase.
Separate price from usage. If you plan to upgrade to a pricier model, put that in the price trend, not the tokens-per-user field.
Watch the cumulative line. Your finance team budgets the annual total — the sum of all twelve months — which is what this tool surfaces alongside the end-of-year run rate.
Run a pessimistic scenario. Double the user growth rate and set the price trend to +0% and see what happens. If the pessimistic cumulative still fits your budget, you are planning conservatively enough.