Oracle Corporation (NYSE: ORCL), initiation of coverage, Nov 9, 2026

Oracle’s backlog is huge, but we think it converts too slowly for the Street’s plan

Oracle has $664bn of contracted AI cloud work, but much of it depends on OpenAI and costs more to build than Oracle earns. We match this year’s guidance and sit 25% below analysts on 2028 earnings.

Our rating
Hold
12-month target
$124.99
Price, Nov 6, 2026
$116.10
Expected return, with dividend
+9%

By Penmark Team

ORCLHoldTarget $124.99+9% expectedFull report

What Oracle is now

A database company turning into an AI landlord, spending far ahead of its cash

Oracle sells the database and business software that run company records, and now rents AI computing capacity. Cloud infrastructure was $18.1bn of revenue last year; we expect $119.3bn by 2030.

Contracted future revenue, called backlog, reached $664bn. Building for it costs more than Oracle earns: capex was $55.7bn last year, so free cash flow was −$23.7bn, and we see it positive only in 2029.

AI cloud grows to two thirds of sales
Revenue, FY26A$49.3bnRevenue, FY30E$119.3bn$56.1bn
Cloud infrastructureRest of Oracle

Fiscal years to 31 May. Source: company filings, Penmark Research estimates.

Free cash flow stays negative until 2029
−$0.4bnFY25−$23.7bnFY26−$25.1bnFY27−$24.3bnFY28$6.7bnFY29
ReportedOur estimate

Free cash flow is operating cash flow less capex, including customer prepayments. Source: company filings, Penmark Research estimates.

How Oracle makes money (page 7)

What you are really buying

Lenders and customers have first claim on about a quarter of the business

Our cash flow model values the whole business at $481.5bn. Borrowings, leases and $16.0bn of customer prepayments come first, net of $37.1bn of cash, leaving $366.9bn for shareholders, or $120.46 a share.

That is close to the $116.10 price. The market already assumes a 9.98% discount rate, near our 9.82%, so the shares are neither cheap nor stretched on our middle case.

From the value of the business to shareholders’ value
$481.5bnValue of thebusiness−$134.5bnBorrowingsand leases+$37.1bnCash−$17.2bnPrepaymentsand other$366.9bnShareholders’value

Value of the business is the discounted cash flow enterprise value at 31 Aug 2026; other claims are minorities and preferred dividends. Source: Penmark Research estimates.

The valuation bridge (page 20)

The bull case

If the backlog converts on time, the shares could nearly double

Bulls take management’s plan at face value: $226bn of revenue by 2030, a 66% infrastructure margin and capex of $100bn in 2028. That case is worth $229.80 a share.

The biggest single lift is the backlog converting on schedule on prepaid terms, worth $24.34. Hyperscalers keeping capex high would add $14.5, and three more points of margin $12.11.

What each upside adds per share
Backlog converts on time+$24.34Hyperscalers keep spending+$14.5Margin 3 points higher+$12.11Capex $15bn lower in 2029+$4.5

Each item is a full model re-run against the $120.46 base value; backlog is remaining performance obligations. Source: Penmark Research estimates.

Backlog conversion (page 14)

The bear case

If OpenAI cannot pay, the shares could lose three quarters of their value

S&P estimates roughly half of the backlog, about $320bn, is linked to OpenAI. OpenAI’s revenue came in about $20bn below investor estimates, Microsoft is winding down a rental and Tencent cancelled a $7bn lease.

In the bear case, commitments are renegotiated, margin falls to 52% and 2030 revenue is $118bn: $26.50 a share. OpenAI alone costs $32.61, and leases arriving early $20.19.

What each risk takes off our value, per share
OpenAI cannot fund its ramp−$32.61Leases start ahead of revenue−$20.19Political backlash−$16.86Higher funding costs−$13.24

Each item is a full model re-run against the $120.46 base value. Source: Penmark Research estimates.

The risks (page 22)

What decides it

The March backlog update decides it: about $145 or $88 a share

In March, Oracle reports its backlog for the third quarter. Commitments above $664bn on prepaid terms would lift our base value to about $145; a cut to OpenAI’s would take it to about $88.

At 35% each, the two net to −$2.90 a share, which is why we sit at Hold. The discount rate matters as much: each point moves value by $27.76.

Value per share after the March update
Backlog converts$144.80Our middle case$120.46Today’s price$116.10OpenAI cut$87.85

Base-case value per share at the price date after each model re-run. Source: Penmark Research estimates.

What moves the stock (page 13)

The odds

The cases span −77% to +98%, and weighting them gives only modest upside

Our middle case, 50% likely, meets this year’s guidance and reaches $175bn of revenue in 2030 with a 60% margin: $120.46. The good case, 20% likely, delivers management’s plan: $229.80.

The bad case, 30% likely, sees OpenAI renegotiate and margin fall to 52%: $26.50. Weighted, they give $114.14, which rolls to our $124.99 target, 7.7% above the $116.10 price.

  • Good, 20%Management’s plan arrives on time: $226bn of revenue in 2030, a 66% infrastructure margin and $100bn of capex in 2028: $229.80.
  • Middle, 50%This year matches guidance; OpenAI concentration slows growth to $175bn of revenue in 2030 and margin settles at 60% as leases start: $120.46.
  • Bad, 30%OpenAI commitments are renegotiated and older capacity churns; margin falls to 52% and 2030 revenue is $118bn: $26.50.
Where the price could be in 12 months
−$89.60 down+$113.70 upToday $116.10$26.50Bad case$120.46Middle case$229.80Good caseOur target $124.99

Circle size shows how likely we think each case is. The target weights all three and adds 12 months of expected return. Source: Penmark Research.

All three cases (page 20)

What would change our mind

A credit downgrade or a backlog surprise would move us off Hold

Conviction is low: three of four alternative discount-rate tests change the rating. One point lower gives a $159.89 target, a Buy; one point higher gives $99.09, a Sell.

The dates to watch are December results, the March backlog update, credit rating reviews in the first half of 2027 and the June capex guide. A second equity raise would break our base case.

What bulls get right. Demand is real. GPU utilisation is 97.9%, renewals are about 20% higher, and the backlog converting on time would add $24.34 a share.
Dates to watch
  1. Infrastructure growth near 116% with margin about 60% confirms our view; growth below 90% or margin below 55% breaks it.

  2. Backlog above $664bn on prepaid terms adds $24.3 a share; a cut to OpenAI’s commitments removes $32.6.

  3. A cut to Baa3 at Moody’s or BB+ at S&P would raise funding costs, −$13.2 a share per half point.

  4. We model $90bn for next year. Above $100bn without new prepaid contracts would break our view.

Catalysts and signposts (page 4)

Inside the full report

What each page covers after the summary on pages 1 and 2. Each opens the full report at that page.

See all 23 pages
  1. 3Returns fall as capex is depreciated, and the one big gap to the Street is FY28E EPS: $8.29 against $11.05
  2. 4The shares are cheap on trailing EBITDA and history, but the cheapness is the market pricing capex and funding risk
  3. 5Six exhibits carry the thesis: the stock trades on backlog news, and we differ from the Street after FY27
  4. 6Three debates decide the stock: we are more cautious than the market on backlog and margin, slightly less on the cost of capital
  5. 7Cloud is now half of Oracle’s $67.4bn revenue, and 56 cents of each dollar goes to cash operating costs
  6. 8Infrastructure supplies 92% of FY27E revenue growth, and its margin is the lever that matters most
  7. 9Demand for AI capacity exceeds supply, but Oracle shares the market with peers earning 19-63% margins
  8. 10Oracle pairs 30% growth with a 52% margin, and trades on the lowest EV/EBITDA of its peers (13.9x)
  9. 11Management has put $97bn into capex in five years, and the FY30 plan is 28% above our base case
  10. 12The stock trades on backlog, margin and funding news rather than on EPS, and the market has priced about 70% of the weighted backlog risk
  11. 13Rates and the EPS print do not move Oracle; the next numbers are RPO in March and infrastructure margin in December
  12. 14Backlog converts below the management path: FY30E revenue $175bn against the $225bn target
  13. 15Leases hold infrastructure margin near 60%, and each point is worth $4.04 a share
  14. 16The discount rate decides the call: each point of WACC moves value $27.76 a share
  15. 17Revenue and EPS have compounded for four years, but profit stopped converting to cash in FY25A
  16. 18Our forecasts match FY27 guidance, but FY30E revenue is 22% and EPS 32% below the Investor Day targets
  17. 19The target of $125 rests on a DCF worth $120.46, cross-checked by a peer multiple at $152.52
  18. 20Bear, base and bull cases tie to the drivers, and the weighted value of $114 supports a Hold
  19. 21Five dated events decide the rating, and two of them can add $24 or remove $33 a share
  20. 22Four risks cost $13-33 a share, and the thesis ends if the March RPO update shows a cut or a conversion
  21. 23Institutions hold at least 42% of Oracle, short interest is 1.6%, and the Street is 83% Buy
  22. 24The model builds to $175bn of revenue and $29bn of free cash flow in FY30E, after two years of negative cash flow
  23. 25Reference: adjusted EPS runs $0.36-1.80 above GAAP, and every figure comes from one model dated 6 November 2026

Penmark Research. This summary draws on our initiation report on Oracle Corporation, published Nov 9, 2026. It is not investment advice. Analyst certification, rating definitions and important disclosures are in the full report.

Sign in to Penmark