Is the Dangote Refinery IPO Overpriced?

The Dangote Petroleum Refinery and Petrochemicals FZE public offer has attracted considerable interest, including from Nigerians entering the equity market for the first time.

According to the official IPO website, the company is offering 4.1 billion ordinary shares at ₦525 each, with a minimum subscription of 10 shares. The offer opened on 14 September 2026 and is scheduled to close on 13 October 2026.

The excitement extends beyond the refinery’s existing operations. The company currently operates a 700,000-barrel-per-day integrated refinery and plans to expand capacity to 1.4 million barrels per day by 2029. Investors are therefore evaluating an established operation alongside ambitious expectations for higher production, stronger cash flows and continued market expansion.

That growth story may be compelling, but an exceptional business does not automatically represent an attractive investment at every price.

During dbrownconsulting’s AI Model Audit, David Brown and Paul Barnhurst, widely known as the FP&A Guy, submitted the prospectus to several AI configurations. Each tool was asked to analyse the document, build a financial model, estimate fair value and produce a one-page report recommending whether investors should subscribe.

The exercise centred on one question: Does the ₦525 offer price provide sufficient upside, or are investors already paying for growth that has not yet materialised?

What Copilot, Claude and ChatGPT Produced

The tools received the same source material and substantially similar instructions. Their answers, however, differed sharply.

Across several runs, individual bear, base and bull-case estimates stretched from approximately ₦17 to ₦792 per share. One model returned a base value of about ₦174, another estimated approximately ₦242, while a Microsoft Copilot run produced a base case close to ₦499.

The differences extended beyond fair value. One tool completed an initial draft within minutes. Another ran for more than an hour, consumed hundreds of thousands of tokens and produced a detailed workbook containing operational schedules, financial statements, sensitivities and a comparables section. Yet some assumptions were unsupported, and parts of the comparable-company analysis remained unpopulated.

The outputs looked professional, but they did not provide a stable basis for an investment decision.

Why the Valuations Differed So Dramatically

A valuation is shaped by its assumptions. Production volumes, capacity utilisation, refining margins, exchange rates, reinvestment requirements, return on invested capital, cost of capital and terminal growth can each materially change the result.

The AI tools interpreted these variables differently. They also disagreed on how much detail the assignment required. Some created simplified discounted cash-flow models, while others tried to forecast multiple financial statement and operational line items.

Similar instructions also produced different answers when the exercise was repeated. This reflected the probabilistic nature of generative AI, but it raised an important professional concern: an investment recommendation cannot be considered reliable simply because an AI tool generated an organised workbook to support it.

How Much Projected Growth Appears Reflected in the Offer Price

At ₦525 per share, the offer appeared expensive under the assumptions examined during the experiment.

The valuation suggested that investors were being asked to pay for both the refinery’s current operations and a significant portion of its expected growth. That future value depends on successful capacity expansion, sustained demand, capital execution, stable exchange rates and refining margins that remain attractive after current market conditions normalise.

The refinery may eventually justify or exceed the offer price. Its expansion plans could increase earnings materially, while stronger production and export performance could create further value. The concern is that these outcomes are being priced as though they carry limited execution risk.

Based on these, the IPO appeared expensive relative to the estimated value of current operations because the market was already placing a substantial bet on future growth.

David Brown and Paul Barnhurst’s Review of the Assumptions

After examining the AI outputs, David Brown applied a simplified value-driver approach using operating profit, return on invested capital, reinvestment, growth and cost-of-capital assumptions.

The review produced an indicative value of approximately ₦380 to ₦400 per share, subject to more detailed analysis. Paul Barnhurst reached a similar view after reviewing the prospectus and the AI-generated models.

Both considered the ₦525 offer price demanding. Their concern was not that the refinery lacked long-term potential. Rather, the offer appeared to treat a considerable portion of that potential as already secured, leaving investors with a narrower margin of safety if expansion takes longer, costs more or produces weaker returns than anticipated.

The Key Lessons for Investors and Financial Modelers

The experiment produced four practical lessons:

• A strong business can still be offered at an unattractive entry price.
• Future growth should be discounted for execution and market risk.
• A polished model is only as credible as its assumptions.
• AI-generated valuations require qualified human review before capital is committed.

These findings align with the Financial Modeling Institute’s Human Financial Modeler report. None of the 63 experts surveyed would rely on an AI-generated financial model for a high-stakes decision without independent human review.

AI can accelerate prospectus review, extract financial information and build a first valuation model. It cannot determine which assumptions deserve confidence or accept accountability when a recommendation proves wrong.

dbrownconsulting’s bespoke AI for Work, AI for Strategic Leaders and AI for Board Members programmes help professionals and organisations apply AI within controlled workflows while strengthening validation, governance and decision-making. Each programme can be tailored to an organisation’s roles, data environment and strategic priorities.

To get started, please send an email to [email protected]