This combination creates a high degree of confidence in the direction of Oracle’s growth. It also makes precision illusive, if not impossible. There is simply too much activity — too many simultaneous capacity announcements, too many new contracts, too many cross-currents from database and applications — to form a single precise expectation for revenue and earnings.
On top of the infrastructure story, Oracle has substantial potential to grow its database and applications businesses as enterprise customers lean into AI as both a growth unlock and a defense against competition. Every one of those motion picks up revenue in layers the old model did not contemplate.
Attempting narrow precision offers no real benefit over the directional path. Any forecast we put on paper is likely to require constant upward revisions as money starts flowing at even greater velocity in 2026 and beyond. That is something a disciplined forecaster must accept, not resist.
The instinct in financial analysis is to reach for a single number and defend it. That instinct serves the forecaster well in stable businesses. It serves them poorly in inflection businesses. When the underlying drivers are expanding at multiples per year and the contracted backlog is redefining itself each quarter, the accuracy of any point estimate is destroyed before the model is even circulated.
The correct frame is to identify what can be said with conviction — the direction, the shape of the curve, the year in which the evidence becomes undeniable — and to resist the temptation to dress uncertainty in a false precision. The forecast’s job here is to give us a baseline we can mark the world against, not to predict the number Oracle will print.
It helps that Oracle’s business is now largely driven by contracts. Contracts give management visibility into minimum revenues. It further helps that the counterparties on those contracts are a small group of businesses with immense capital at their disposal — hyperscalers and frontier AI labs whose ability to honor their commitments is itself a function of the same secular demand that underwrites the thesis.
In short: the contract structure narrows the dispersion of outcomes around the projection. Oracle is not projecting into a void. They are projecting from a signed backlog.
Given all of that, the most logical approach is to use management’s projections as a starting point, then apply a premium on top — because we believe management has already started what will be a series of upward revisions. Every public statement from management in the past year has been directionally more aggressive than the last. We are at the front of a cycle, not the middle.
This approach preserves the discipline of anchoring to something observable (contracted revenue) while acknowledging what the thesis argued: that the actual results will outrun the projection as capacity comes online, as inference spending compounds, and as database and applications revenue accelerates in an AI-transformed enterprise.
The shape of the curve matters more than any individual year. From fiscal 2022 through fiscal 2025, reported revenue climbed modestly — $42B to $57B — reflecting the legacy mix dominating the result. Starting in fiscal 2026 the slope steepens as contracted backlog begins to translate into recognized revenue. By fiscal 2028 the incremental dollar added each year exceeds the entire annual revenue Oracle reported five years earlier.
Two further points on interpretation. First, this is not a revenue forecast drawn on a blank page. It is anchored to management’s projections with a moderate premium layered on top, consistent with Section 2’s methodology. Second, we believe even this path will require upward revision. The thesis argues Oracle’s growth is a direct function of its ability to build data centers, and that the demand curve does not plateau — both of which imply reality will outrun the chart.
The EPS curve is materially steeper than the revenue curve in the out years. That is not arbitrary — it is the margin structure of a business whose incremental revenue carries high operating leverage once the capacity is built and contracts are running. Capacity buildouts are capital-intensive; the revenue against them is high-margin.
The year that matters most for our valuation is fiscal 2028. We have chosen to use FY28E EPS as the base because it is the first year in which the leverage is fully visible in reported results. Fiscal 2026 still carries the buildout weight; fiscal 2027 is the transition; fiscal 2028 is when the shape of Oracle’s earnings power is undeniable. That is the number against which the market will price the stock as it looks forward.
As with revenue, this is a baseline. If the thesis holds, actual EPS runs ahead of this projection — because pricing power in scarce AI capacity, cross-sell into AI-native database workloads, and the applications tailwind from agentic AI all compound on top of what the chart already shows.
We believe in establishing reasonable floors. Forecasting under explosive growth and a changing competitive set does not reward heroism — it rewards discipline. A defensible floor gives the reader a clear target to beat and leaves the upside unencumbered by a forecast that tries to catch every dollar.
Oracle is currently trading at approximately 25× NTM earnings. That is above Oracle’s long-term average, and a reasonable starting objection from a skeptic would be that the stock is already priced for a re-rating. Our response is structural: Oracle is not the same business it was when that long-term average was formed. The applicable peer set has changed. The growth rate has changed. The revenue mix has changed. Holding Oracle’s future multiple to its pre-AI historical average is comparing two different businesses.
The method is deliberately conservative. We pick a multiple we believe the market will be willing to pay even in a measured re-rating — not the peak multiple we could justify. If the re-rating runs further, the stock delivers more than our case. If it runs only as far as the conservative case, we still have a doubling. That asymmetry is the entire point.
The hyperscaler peer set is itself in the middle of a re-rating that is not fully reflected in current multiples. Microsoft, Amazon, and Google are each being transformed by AI in ways that argue for their own multiples running well above their historical averages. We believe there is a massive re-rating coming for this entire group as capacity is brought online and agentic AI spreads across the enterprise.
If the peer group itself re-rates, the relative argument for Oracle strengthens further. Oracle is not being asked to out-earn hyperscalers — just to earn at the peer multiple, against a growth rate that exceeds theirs. Even that conservative framing produces a return that makes the position compelling.
For conservatism, we are forecasting Oracle at a P/E of 30× NTM earnings. That is below where we would argue Oracle belongs on pure growth differentials, and roughly in line with where the hyperscaler peer set is already trading today — before the peer group re-rating we expect. It is a deliberately modest assumption.
The base for the multiple is fiscal 2028 EPS. Our projected FY28E EPS is approximately $12. The fully expanded base case falls out of that directly.
The tables exist to make the asymmetry visible. Even holding the multiple flat at the conservative 30× anchor, modest upside to EPS — $14 instead of $12 — produces $420 per share, a return of roughly +140%. Modest upside to the multiple — 35× instead of 30×, consistent with the peer-set re-rating we expect — produces $420 at the same $12 EPS, also a +140% return. If both variables move even slightly in our favor (the scenario we believe is most likely), the implied price moves into the $490–$560 range and returns into the +180% to +220% range. The structural arguments in the thesis point to outcomes above that: Oracle runs faster than peers on growth, the peer set itself re-rates, and the backlog compounds into an even larger base.
The downside requires both variables to move against us — EPS at the low end of the table and the multiple compressing below current levels. That is a scenario in which Oracle fails to monetize its backlog while the market simultaneously decides hyperscaler multiples should fall. Both conditions have to hold. Neither is consistent with what the thesis has already established about demand, contracts, and capacity.