Oracle Update: The OpenAI Risk Faded, the Debt Risk Became a Downgrade

In March we published a full 5D case study on Oracle and called it what it was: a leveraged bet on AI-infrastructure demand with a razor-thin margin of safety. Four months on, the share price is almost exactly where we left it, around $141, but it got there the hard way. Oracle ran up to roughly $200 after a record June quarter, gave all of it back, and in early July S&P downgraded its credit to one notch above junk.

This is a focused update, not a fresh case study. Two things have genuinely changed since March, and they point in opposite directions.

The soft signals have improved

Start with the good news, because there is real good news.

Analyst support has gone from decent to strong. On our data the share of buy ratings has climbed into the mid-80s percent and is still rising, the average price target sits near $254 against a $141 stock, and the recent target revisions have been mostly upgrades. The insider and fund selling that was the single ugliest signal in our March write-up has reversed, with both now net buyers on our holdings read. And the biggest unknown from March, whether OpenAI, which accounts for more than half of Oracle's backlog through the Stargate deal, could actually fund what it had committed to, looks better after OpenAI closed a $122 billion raise. The backlog itself has grown to about $638 billion.

The one place Oracle does not screen well, even against those same peers, is leverage. Which brings us to the other side of the story.

The balance sheet got worse, and S&P noticed

The risk we flagged in March did not fade. It intensified, and on 9 July it showed up in the credit rating: S&P cut Oracle to BBB minus, one notch above speculative grade, with a negative outlook. The stated reasoning was blunt, rising business risk and weaker cash flow from the AI build-out.

The numbers underneath are stark. On our data, fiscal 2026 free cash flow was minus $23.7 billion, a swing from roughly breakeven a year earlier, as capital spending jumped from $21 billion to $56 billion in a single year. S&P now projects the fiscal 2027 free cash flow deficit widening to around minus $42 billion, nearly double its previous estimate, with leverage pushing into the mid-4x range. Total debt is around $167 billion and climbing.

Here is the part that matters for shareholders specifically: Oracle is increasingly funding the gap with equity, not just debt. It issued $5 billion of mandatory convertible preferred stock in February, has flagged a further $20 billion equity raise this year, and has signalled tens of billions more over the next few years. That is dilution, and it lands directly on the per-share value the analysts are targeting.

What our DCF actually needs

This is where the March case study deserves a proper update.

Back then our discounted-cash-flow model produced an intrinsic value around $304, and we cautioned that it leaned entirely on analyst revenue estimates of roughly 40% compound growth, many times Oracle's historical rate.

As of the 11 July 2026 model run, the model value is about $301, still well above the $141 price, and it still runs on analyst revenue, now about 30% a year, with revenue tripling to roughly $218 billion by fiscal 2030. To see what today's price is really assuming, we flexed the revenue growth rate and re-solved the model:

Revenue growth (5yr CAGR)	Model value	vs $141 price
30% (analyst estimate)	$301	+114%
25%	$232	+65%
20%	$180	+28%
15%	$136	roughly fair
10%	$99	-30%

But fiscal 2026 landed almost exactly on the estimate, $67.4 billion against a projected $67.2 billion, so the first year of the aggressive ramp actually happened. And the table shows today's price only needs about 15 to 16% revenue growth to stand up, just below the 17% Oracle delivered last year. On the revenue assumption alone, then, the downside looks better protected than the 114% headline upside suggests.

But two features of that model value should keep anyone honest.

First, it is almost entirely back-loaded. Nearly 90% of the model's value sits in the terminal years, beyond the explicit forecast. The reason is the capex wave: the model has Oracle's free cash flow starting at minus $15 billion, staying negative for four years, and only then climbing toward plus $90 billion a decade out as the build-out rolls off and the data centres start paying for themselves. The entire bull case depends on that recovery arriving roughly on schedule. S&P's minus $42 billion fiscal 2027 estimate is a reminder of how negative the near term can get first.

Second, the model value is per share on today's roughly 2.9 billion shares. With $20 billion or more of equity issuance coming, the denominator is about to grow. Even if the revenue and the eventual cash flows show up exactly as modelled, each existing share is entitled to a smaller slice of them.

Where this leaves us

Our March conclusion was that Oracle was not a classic value play but a leveraged bet on AI-infrastructure demand, and that the market was pricing the risk roughly correctly. Four months later we are, if anything, slightly more cautious.

The customer side genuinely improved. OpenAI is better funded, analysts are more bullish, the insiders stopped selling, and on the pure revenue math the stock is not demanding. But the risk we actually cared about, Oracle's own balance sheet, did not recede. It turned into a credit downgrade, a widening cash deficit, and a dilution programme. The valuation only works if a large, back-loaded free-cash-flow recovery arrives on schedule, funded along the way by debt the company is now being penalised for and equity that dilutes existing holders.

That is a coherent bet, and it may well pay off if the backlog converts. But it is a financing-and-execution bet before it is a value one, and the ratings agencies have started to say so out loud. We are staying cautious, with both eyes on the balance sheet.

Explore the current numbers on our Oracle valuation page.