What the Fastest-Growing AI Companies Can Teach FP&A Teams

Finance teams love precedent. Comparable companies, historical trends, prior-year actuals. That makes sense. It is how the job has always worked.

But what happens when there is no precedent?

Anthropic reportedly went from $1 billion in annualized revenue to $14 billion in just fourteen months. Cursor, a code editor created by four MIT students, crossed $1 billion in ARR before turning three, with almost no marketing spend. OpenAI raised at a $122 billion valuation and is said to be generating around $2 billion a month. These companies are no longer rare exceptions. They are becoming the new standard, and the finance teams supporting that kind of growth are operating very differently from traditional FP&A teams.

To be clear, these figures come from recent funding rounds and public reporting as of early 2026. Private company metrics move quickly, so it is better to think of them as directional rather than exact.

What High-Performing FP&A Teams Are Doing Differently

So what are the best teams doing differently?

They are no longer spending their time defending old budgets.

A lot of finance teams still spend months explaining why actual results do not line up with a plan that was built long ago using assumptions that no longer hold. It is frustrating, and more importantly, it does not help the business move forward.

Fast-growing companies have moved away from that mindset. Not because they are careless, but because they know a static budget loses value quickly in a fast-changing environment. Nvidia, for example, posted $130.5 billion in revenue in FY2025, up 114% year over year. That kind of scale is not driven by guesswork. Teams like that rely on rolling forecasts that stretch 12 to 18 months ahead and get updated constantly. The forecast becomes a living management tool, not a document people defend every quarter.

OpenAI’s VP of Strategic Finance, Stacie Faggioli, has summed it up well: finance teams need to spend more time looking forward than backward. When revenue can swing 30% in a single quarter, last year’s budget is mostly just historical context.

Of course, shifting to rolling forecasts is not as easy as it sounds. The hard part is rarely the model itself. The real challenge is helping business partners understand that revising a forecast is not the same as missing a target. If people are punished every time they update the numbers, they will stop being honest. That cultural shift matters just as much as the systems behind it.

Why Gross Margin Matters More Than Revenue

At scale, gross margin becomes the number that matters most.

Revenue gets attention, but gross margin tells the real story.

Faggioli has spoken publicly about how focused her team is on margins, especially when deciding where GPU capacity should go and how to balance free versus paid offerings. At a company spending heavily to grow, that level of discipline is what keeps growth sustainable.

Many FP&A teams still look at gross margin in one blended number, maybe with a basic segment breakout. But that kind of view hides the details that actually matter. You miss which products are truly profitable, which customer groups are pulling margins down, and which cost drivers are expanding faster than revenue.

The better approach is to build a full waterfall. Break margin down by product, customer type, and cost category. Make individual leaders responsible for specific lines instead of asking them to react to one blended percentage in a deck. When difficult decisions need to be made, that level of detail makes the conversation clearer and far less political.

The Spreadsheet Bottleneck

Everyone knows spreadsheets are slowing the business down.

Finance teams have been talking about the spreadsheet problem for years. Almost everyone agrees it is an issue, yet most companies still fall back on them.

Faggioli has been direct about this too. She said headcount planning in spreadsheets can materially slow the business. OpenAI moved toward real-time headcount planning for a simple reason: finance had become a bottleneck at the exact moment the company needed to move faster.

The solution is not hard to describe. Teams need consolidated data, real-time visibility, and one shared version of the numbers. The hard part is making that change stick. It requires more than funding. It needs executive support, real ownership, and strong change management. When organizations treat it like a side IT project, progress usually stalls.

A good place to start is not with the biggest transformation idea. Start with the process that causes the most pain right now. Fix that first.

The Truth About “AI-Ready Data”

A lot of companies talk about “AI-ready data,” but very few have it.

Every finance software pitch now seems to mention AI-powered planning. That is fine, but Faggioli has been consistent on one important point: AI only becomes useful when the underlying data is clean, structured, and well governed. Without that, AI does not improve decision-making. It just produces flawed answers faster.

Before asking what AI can do for planning, finance leaders need to ask whether their data is actually ready for it.

Is the chart of accounts consistent across the business? Do CRM and ERP systems use the same customer identifiers, or is someone still reconciling that manually every month? Are historical actuals reliable, or are they full of adjustments nobody fully understands anymore?

Most companies that take an honest look find gaps. That is normal. It is what happens when systems are layered together over time. But no planning platform can fix a weak data foundation on its own. That work has to come first.

Why Cash Planning Cannot Be Reactive

Cash issues usually look obvious in hindsight.

Nvidia continues to invest enormous sums into infrastructure while still maintaining positive cash flow. That does not happen by accident. It reflects a finance function that models liquidity with the same level of rigor it applies to revenue.

Many companies do not actually manage cash in that way. They review it. That is not the same thing.

The strongest finance teams treat cash planning as an ongoing process, not something that gets attention only when a problem appears. Payment terms, supplier behavior, borrowing costs, capex timing, these are modeled regularly, not because something is already wrong, but because that is how you avoid being surprised.

A practical way to do this is to build a simple scenario matrix around a few cash drivers that really affect liquidity. Things like days sales outstanding, a major infrastructure investment, or a financing event. Run those combinations against your 13-week cash position and revisit them monthly. The goal is not to predict everything perfectly. It is to make sure that when conditions shift, the CFO hears, “We saw this coming,” instead of, “We have a problem.”

Where Kepion Fits In

Where Kepion fits into this

None of these practices are especially hard to understand. The challenge is putting the right infrastructure in place to support them.

Kepion connects with Microsoft Fabric, Excel, Power BI, and Azure, allowing actuals, forecasts, and operational data to sit in one governed model instead of being spread across disconnected systems and manually reconciled. Rolling forecasts, driver-based planning, and real-time scenario modeling are not add-ons. They are part of how the platform is designed.

That supports the same broader goal Faggioli has described at OpenAI: getting finance out of the business of managing spreadsheets and data cleanup, and into the business of influencing decisions.

The Bigger Takeaway

The finance teams behind the fastest-growing AI companies are not necessarily using more complex models. They are using live ones. They are not magically better at margin management. They are simply tracking the right drivers in enough detail to act on them. They are not perfect at cash planning. They are just not treating it like a reactive exercise.

None of this is out of reach for most FP&A teams. The knowledge exists. The tools exist. Kepion is one option among them.

The real question is not whether the software is available. It is whether the organization is willing to change how finance actually operates.

That part still comes down to leadership.