Oracle's AI Data Centers Show Modest Returns, Analyst Still Rates Stock a Buy

Oracle stock trades at a suppressed 14x forward price-to-earnings ratio, and one analyst argues that discount understates the company's earnings growth potential even though its AI data centers generate what he calls modest returns at best after funding costs.

The analysis, published by financial analyst Bashar on Seeking Alpha, maintains a 'Strong Buy' rating on Oracle (ORCL). The rating rests on a counterintuitive reading of the AI infrastructure buildout: profitable but thin-margin data center economics, a valuation the market has beaten down, and a software portfolio that cushions margins in ways that pure-play cloud rivals cannot match.

What did Oracle disclose about data center profitability?

Bashar's core finding draws on Oracle's disclosure of profitability figures for its AI data centers on a project-level basis. The numbers, he wrote, "confirm what I long suspected about bare-metal compute: it generates modest returns at best."

That is a warning sign, in his framing. Bare-metal compute — servers leased directly to customers without virtualization layers — carries real returns only after Oracle covers the substantial funding costs of building and equipping the facilities. Once those capital costs are counted, the residual profit is thin.

The stakes are considerable. Oracle has committed to heavy AI capital expenditure, and the market has responded skeptically. ORCL's suppressed 14x forward P/E reflects two concerns Bashar identifies directly: skepticism over heavy AI CapEx and a low return on assets (ROA), a measure of how efficiently a company converts its asset base into profit. Low ROA on an expanding data center fleet means capital is being absorbed faster than earnings.

Bashar's argument is that the market has anchored on those risks and underpriced the offsetting factor: earnings per share growth potential that he considers underappreciated.

Why could fewer datacenter permits help Oracle?

The most contrarian claim in the analysis concerns regulation and supply. Bashar argues that regulatory or supply constraints could enhance the value of existing data centers.

The logic runs as follows. If permits for new facilities become harder to obtain, or if supply chains for chips, power, and construction tighten, then capacity already built or underway becomes scarcer. Scarcity supports pricing for the compute that existing facilities sell. Oracle's installed AI data centers, already profitable at the project level, would see the economics of that installed base improve rather than degrade.

In this framing, constraints that Wall Street reads as headwinds for the sector could function as a moat for incumbents with capacity in hand. Oracle's disclosures, according to Bashar, give investors a rare project-level view of whether that capacity actually earns money — and it does, even if modestly after funding costs.

How does Oracle's software portfolio change the math?

Bashar draws a sharp distinction between Oracle and the "Neoclouds" — the newer, smaller cloud infrastructure providers focused on GPU-heavy AI compute. Oracle's robust software portfolio, he writes, provides a margin cushion that Neoclouds lack.

The distinction matters because of business model structure. A company selling enterprise software alongside cloud infrastructure can absorb thinner margins on raw compute by leaning on higher-margin software revenue. A company selling compute alone has no such buffer. If AI infrastructure margins stay modest — exactly what Oracle's project-level numbers suggest — the diversified operator survives the pricing environment more comfortably than the pure-play operator.

That cushion is central to the Strong Buy thesis. The data center business does not need to be spectacular on its own. It needs to be profitable, which the disclosures indicate it is, while software carries margin quality.

Is Oracle's debt load a problem?

Debt is near peak, Bashar writes. That admission sits at the center of the bear case against Oracle: an aggressive AI buildout financed at scale, meeting modest bare-metal returns, produces balance sheet strain.

Bashar's counterargument rests on two mitigating factors. First, operating cash flows: the ongoing business generates cash that services the debt burden. Second, customer prepayments: customers paying Oracle in advance for capacity convert future revenue into present-day liquidity, reducing the need for new borrowing even amid heavy capital spending.

Together, he argues, these flows mitigate balance sheet stress and support the investment thesis. The recent share price decline — the context in which the analysis was published — does not, in his view, reflect an improvement in these underlying fundamentals.

Who is the analyst, and what are the disclosures?

Bashar writes on Seeking Alpha, focused on growth stocks, contrarian setups, and market mispricing. His research looks for companies where consensus is missing a shift in earnings power, competitive positioning, or industry structure. He states that he does not invest personally in the stocks he covers.

The disclosure accompanying the analysis states that the author holds no stock, option, or similar derivative position in Oracle, and has no plans to initiate any such position within the next 72 hours. The author wrote the article himself, expresses his own opinions, and receives no compensation for it other than from Seeking Alpha. He discloses no business relationship with any company mentioned.

Seeking Alpha's own disclosure notes that past performance is no guarantee of future results, that no recommendation or advice is being given as to whether any investment is suitable for a particular investor, and that its third-party analysts include both professional and individual investors who may not be licensed or certified by any institute or regulatory body.

What is the so-what for AI infrastructure investors?

The Oracle case distills to a broader question hanging over the AI buildout: whether anyone can earn attractive returns selling raw compute once capital costs are paid. Oracle's project-level disclosures, as read by Bashar, suggest the answer is yes — but only modestly.

For investors, Bashar's position is that the market has overcorrected. A 14x forward P/E on a company with underappreciated EPS growth potential, a software margin cushion, cash flows and prepayments absorbing near-peak debt, and existing data centers that scarcity would make more valuable, adds up in his analysis to a Strong Buy despite recent share price declines.

Whether that thesis holds depends on the variables he himself flags: whether AI CapEx converts into earnings at the pace he expects, and whether regulatory and supply constraints materialize in a form that raises, rather than strangles, the value of capacity already on the ground.