Prosus

IThe business

Prosus is a Dutch holding company with two kinds of assets. It owns 22.5% of Tencent, and it runs a group of consumer internet businesses in Latin America, Europe and India that earned $1.27B of adjusted EBITDA in FY2026, the year to 31 March 2026.⁸¹⁰¹² It was carved out of Naspers and listed in Amsterdam on 11 September 2019.¹⁰ In FY2026 revenue was $9.7B, up 57%, or 12% in local currency excluding acquisitions, and adjusted EBIT was $0.62B.R (reported) Revenue before Just Eat Takeaway.com (JET) and La Centrale, bought during the year and the basis of management’s FY2026 target, was $7.74B.C (calculated)^10111217

The income statement leaves out most of what matters. Tencent is equity accounted, so its profit sits below operating profit: Tencent supplied $8.12B of the $8.34B of core headline earnings, Prosus’s own measure, or 97.3%.R (reported)¹⁰¹² Its cash is smaller. Tencent paid Prosus $1.24B of dividends in FY2026, 15% of the earnings Prosus counts from it, and a further $1.4B arrived in June 2026.C (calculated)¹⁰¹²

The market value of Prosus is below the value of that one holding. For $38.18 a buyer gets $47.21 of Tencent shares and, on the analysts’ values, $15.50 of everything else after net debt, which is $62.71 in all: Prosus’s $130.6B over 2,082.8M economic shares.C (calculated)⁴⁴⁴⁵⁴⁶ The next two figures show it: first per share, then in dollars across the whole group.

The consolidated businesses report in four regions. Europe is the largest by revenue because it holds OLX, eMAG, iyzico and, since October 2025, JET. Latin America earns the most for its size, because iFood and Despegar sit there.¹²

How it makes money

• Tencent. 2,043.2M shares, 22.5% of the company, worth $98.3B at the Q4 2026 close and 71.2% of gross assets.C (calculated) Prosus sells a slice each year to fund its buyback: 95.6M shares for $6.5B in FY2026. The stake was 31.1% at the 2019 listing.⁴⁴⁴⁶¹⁰
• iFood, Brazil, 96.2% owned. 1,877M orders in FY2026, up 48.8%, but core food delivery orders were 994.6M, up 8.3%; the difference is non-food orders, including first-party grocery. Revenue was $1.87B and adjusted EBITDA $400M. iFood Pago, the payments and credit unit, added $463M of revenue and its first profit, $38M.R (reported)¹²
• OLX Group, classifieds in 17 international markets. Revenue $992M, up 16% in local currency excluding acquisitions, adjusted EBITDA $481M, a margin of 48.5%. La Centrale, the French car classifieds site bought in November 2025 for $1.3B, sits inside it with $41M of revenue and $17M of adjusted EBITDA.R (reported)¹⁰¹²
• JET, bought in October 2025 for EUR 4.2B, $4.9B. In the six months to March 2026 it took 292M orders, down 8.8%, on $1.92B of revenue and $83M of adjusted EBITDA, 4.3%.R (reported)¹⁰¹²
• eMAG, PayU, iyzico, Despegar. eMAG, in Romania and neighbouring markets, had $2.54B of revenue, down 2% in local currency, at a 3.9% margin. PayU in India processed $90B of payments, earned its first adjusted EBITDA, $18M, on $781M of revenue. Despegar, the Latin American travel agency bought in May 2025, earned adjusted EBITDA of $131M.R (reported)¹⁰¹²
• Listed stakes beside Tencent. Delivery Hero $2.1B, Meituan $1.7B, Meesho $1.2B, Swiggy $1.1B and five smaller holdings, $6.6B in all.C (calculated)⁴⁴

What changed in twelve months

• Prosus bought JET, Despegar and La Centrale: $7.65B at the cost the accounts carry, $8.0B on headline consideration, 9.6% of today’s market value.C (calculated)¹⁰
• The European Commission made Prosus cut Delivery Hero below its largest-shareholder position as the price of approving JET, so it sold 28.8M shares, 9.5% of Delivery Hero, for about EUR 605M to Uber and Aspex Management. On 16 July Uber offered EUR 41.50 a share in cash for the whole of Delivery Hero, and Prosus gave an irrevocable undertaking to tender its roughly 17%; closing is expected in the second half of 2027.¹⁰³⁸
• On 12 May the chief executive guided iFood’s FY2027 adjusted EBITDA to $100M to $150M against $400M, and the share fell 6.3% that day on 3.5 times the previous day’s volume.C (calculated)⁴²¹⁷
• Tencent fell from HK$484.00 at 31 March, when Prosus’s stake was worth $127.6B, to HK$377.60, when it is worth $98.3B on 2,043.2M shares. It closed at HK$663.00 at the end of September 2025.C (calculated)⁴⁶⁹¹⁰

Prosus closed at EUR 62.18 on 13 November 2025. It is 45% below that. Its own deep dive put the net asset value at about $150B on 26 June 2026; the figure is $130.6B now.C (calculated)⁴²⁴⁴³⁴

A share of Prosus is Tencent first, and the market prices everything else at minus $18.8B.

IIThe moat

I think Prosus has two kinds of moat and one structure that works against its owners. The moats are local network effects in food delivery and classifieds, and the Weixin network behind its Tencent stake. The structure is a holding company with a controlling shareholder.

The first network is iFood. Restaurants list where the diners are, and diners order where the restaurants are. iFood has 469,110 merchants, 621,000 delivery partners and 1,638 cities in Brazil, and management says it holds about 80% of app orders nationwide, down about 5 points in 18 months.¹²¹⁸ The margin is the test. The core food delivery business earned a 31.8% adjusted EBITDA margin in the first half of FY2026 and 22.1% in the second, the half in which, on management’s account, Meituan’s Keeta and DiDi’s 99Food were subsidising orders, and its revenue growth in local currency fell from 24% to 1%.R (reported)¹²¹⁸³³ Management says rivals are burning about $150M a month and that similarly priced products burn 10 times more per order through subsidies than iFood’s. Those are management’s figures; I found no independent measure.¹⁸

The second network is OLX, where buyers go to the sellers who are listed and sellers list where the buyers are. Its adjusted EBITDA margin rose from 27.9% in FY2024 to 40.4% in FY2025 and 48.5% in FY2026. Paying listers rose from 1.78M to 1.82M in FY2025 and fell to 1.76M in FY2026, down 0.5% in the first half and 6.1% in the second, and active listings fell from 64.3M to 60.8M.R (reported)¹² Revenue grew 16% in local currency excluding acquisitions, so revenue per paying lister rose about 20% in local currency.C (calculated)¹² The cost base, revenue less adjusted EBITDA, was $474M in FY2024 and $487M in FY2026 without La Centrale, up 3% while revenue rose 45%, so the margin did not come from cutting costs: nearly all the new revenue reached profit.C (calculated)¹² La Centrale earns 41%, below OLX’s margin; without it OLX earned 48.8%.C (calculated)¹² I read this as price and mix carrying growth while the base of paying customers fell. I cannot yet separate price from mix. Prosus had scheduled an OLX briefing for 26 and 27 October³¹ and has published the slides. They repeat the margin path, 28% in FY2024, 40% in FY2025 and 48% in FY2026, and add revenue and margin by vertical for FY2026: Motors $429M at 59%, Real Estate $185M at 46% and Jobs $87M at 43%. They give no count of paying listers and no split of price from mix, so the question stays open until the half-year results.⁴⁹

Where the moat is missing, the numbers show it. JET’s orders fell for 55 months until September 2026 by management’s count. PayU earned $18M on $781M of revenue. eMAG earned 3.9%. Stack Overflow’s page views fell 56.6% and Prosus impaired $280M of its goodwill.R (reported)¹⁰¹²¹⁸

The structure is a fact of the share register. Naspers holds 43.4% of Prosus’s N shares, 43.1% of its economic shares and all of the B shares, which gives it 75.66% of the votes. Koos Bekker, a former Naspers chief executive, chairs both boards, and Prosus lists nine deviations from the Dutch corporate governance code, including that its chair is not independent.R (reported)¹⁰¹⁶ Prosus’s own audited statements name the reasons holding companies trade at a discount: management costs, tax leakage, governance and shareholder structure, information asymmetry and perceived reinvestment risk.²⁰ One test of the structure: Naspers puts a second layer of control on top of Prosus and trades 46.5% below its own net asset value, wider than Prosus’s 39.1%.C (calculated)⁴⁷²⁰

Is it widening or narrowing? I think the OLX network is widening on price, the iFood network is intact on share and narrowing on cost per order, and the Tencent stake carries Tencent’s moat, which my Tencent analysis calls widening in what each user is worth, flat in the number of users and untested in AI.¹²¹⁸³⁷

OLX raised its margin from 27.9% to 48.5% in two years on a flat cost base, while its paying listers fell 3.3% in the latest year and 6.1% in the second half. iFood’s margin fell 9.7 points in six months.

IIIThe competitive advantage

A holding company competes in two places. It competes for capital, against other ways to own the same assets, and it competes in each business against named rivals. On the first Prosus costs less than Tencent itself for each dollar of Tencent. On the second it leads in Brazil, trails in Europe, and has not shown its margin in India.

Three ways to own the same Tencent

Prosus costs $0.81 for each dollar of Tencent it holds, and it comes with a portfolio the market prices below zero and $7.5B of net debt. Naspers costs $0.73, because it owns 43% of Prosus and trades 46.5% below its own net asset value.C (calculated)⁴⁵⁴⁷⁴⁸ The discounts stack. Naspers controls Prosus, so the controlling holder has not closed its own gap either. If the Prosus discount closes, Naspers gains more, because it carries a second discount on top and holds 43% of Prosus; the cost is a rand listing, a second layer of control and $1.3B of other assets I do not value.

Costing less per dollar of Tencent becomes a better return than owning Tencent directly only if the discount narrows or the unlisted businesses are worth consensus. In my base case, with the discount unchanged, Tencent bought directly returns more than Prosus; the table is in the valuation section under Prosus against Tencent itself.C (calculated)³⁷

Against named rivals

Where Prosus leads

Brazil comes first. iFood has the density, 80 million recurring customers by management’s count, and the products around the order. Its loyalty club, Clube, is 45% of food delivery volume, and iFood Pago turned profitable in the second half of FY2026.¹¹¹²¹⁸ A rival that wants share has to pay for orders at a loss, and the rival’s parent, not Prosus, decides how long that lasts. Prosus holds 193.7M Meituan shares, $1.7B, which means it is a shareholder in the company that is attacking iFood.C (calculated)⁴⁴

The second lead is the funding source. Tencent pays $1.24B a year in dividends and the stake can be sold at about $6.5B a year, so Prosus can fund a subsidy fight without raising equity. It is also spending that same stake on its buyback and acquisitions: central cash fell from $17.2B to $9.5B in a year, and interest cover is 3.8 times.C (calculated)¹⁰¹³

Where it trails

Europe is the weak side. JET’s orders were down 8.8% in October to March while DoorDash’s rose 17% excluding Deliveroo in April to June, and Uber’s Delivery bookings rose 25% in constant currency.¹²³⁹⁴⁰ Management says JET returned to order growth in September 2026, after 55 months of decline, and gives no figure.¹⁸ Prosus paid a 49% premium to the three-month average price for JET, and the analysts’ value is now 22% below the $4.9B consideration and 16% below the $4.5B cost the accounts carry.C (calculated)⁴⁴¹⁰³⁰

In India Prosus controls PayU, which is subscale: its first adjusted EBITDA, $18M, is 2.3% of its revenue.C (calculated)¹² It holds Meesho and Swiggy as listed minorities worth $2.4B together.⁴⁴

IVManagement

Fabricio Bloisi, 49, has been chief executive since 10 July 2024, with a term to 30 June 2028. He founded Movile, whose holding company is now the group’s food business, and he also runs Naspers. Nico Marais, 53, has been chief financial officer since 29 April 2025 and with the group since 1999. Koos Bekker, 73, chairs Prosus and Naspers.¹⁰¹⁶

The operating record since the chief executive took over is good on profit. Adjusted EBITDA of the consolidated businesses was $383M in FY2024, the year before he started, $689M in FY2025 and $1,267M in FY2026.R (reported)¹⁰¹²²⁶ It was a loss of $89M in FY2023, which belongs to his predecessor. A quarter of the rise since FY2024 was bought: Despegar $131M, JET $83M and La Centrale $17M, so without the three FY2026 was $1,036M.C (calculated)¹⁰¹² Core headline earnings excluding Tencent went from a loss of $1,613M in FY2023 to a profit of $223M.R (reported)¹³¹⁹ Free cash flow excluding the Tencent dividend, $275M, was the largest since the FY2024 definition.C (calculated)¹² The FY2026 targets, more than $7.3B of revenue and more than $1.1B of ecommerce adjusted EBITDA excluding JET and La Centrale, were met on management’s definitions, and on mine: $7.74B and $1.17B.C (calculated)¹⁷ The next step is harder. The CEO said on the results call that FY2027 was not a year he expected to increase profitability.¹⁴ The June 2025 ambition of more than 3.5 times FY2025’s adjusted EBITDA by FY2028 implies above $2.4B, which is 1.9 times FY2026’s $1.27B.C (calculated)¹⁰²⁶²⁸

Capital allocation

The buyback has done what it set out to do on its own terms: Tencent per Prosus share rose 5.4%, from 0.931 in March 2024 to 0.981 now. The price was a smaller stake, 22.5% against 31.1% at the 2019 listing, and at today’s prices the swap is $1.3B behind, because Prosus fell more than Tencent. The acquisitions are marked below cost, and the write-downs are in the table.C (calculated)⁴⁵¹⁰²⁰

Over four years Prosus paid $33.0B for its own shares, 41% of today’s market value, and funded it from sales of Tencent, which the programme description says it pairs with the purchases.C (calculated)^1019202132 The swap works in per-share terms because Prosus sells Tencent at the market price and, at today’s prices, retires its own stock at 81 cents on the Tencent dollar. A round of $7.4B, the FY2026 cash paid, retires 9% of the shares and adds 3.5% to value per share at today’s price and 3.1% in my base case; at the guided $5B it retires 6% and adds 2.3% and 2.1%.E (estimate)⁴²¹⁰ In intrinsic terms the swap adds value only while Prosus trades at a lower fraction of its net asset value than Tencent trades of its own value. Today that is 52% of my net asset value against 74% of my Tencent value, which is why the gain falls from 3.5% at the market price to 3.1% in my base case.E (estimate)³⁷ Management guides the buyback at about $5B a year across Prosus and Naspers, which is 6% of Prosus’s market value.C (calculated)¹³¹⁷ Prosus’s own pay report says the discount stayed at about 42% in FY2026,¹⁵ when Prosus returned $7.9B.C (calculated)¹⁰

Incentives

The chief executive’s FY2026 single figure was $1.85M: $825K of salary, a $653K bonus (79% of salary), and $368K of benefits and pension, paid 10% by Naspers and 90% by Prosus.¹⁵ The weight is in two awards. A hire award vests after four years on total shareholder return, with a potential value of $27.6M at 31 March 2026. A one-time moonshot award of $100M, in shares, vests only if the combined free-float market value of Naspers and Prosus doubles from $84B in June 2024 to $168B by June 2028, holds for a year, and total shareholder return beats the median of the peer group.¹⁵ On the plan’s formula that value was $76.0B in Q4 2026, so it has to rise 2.2 times in 20 months.C (calculated)⁴²¹⁵⁴⁸ The short-term bonus carries a goal to cut the discount, weighted 10% for the chief executive and 15% for the chief financial officer. It paid nothing for FY2025 or FY2026.¹⁵²⁹

Ownership is real but small against the awards. Bloisi holds 581,477 shares, EUR 19.9M. On 11 and 14 July 2025 he bought 414,900 in the market at EUR 47.94 to EUR 48.43, and they are down 29%; on 9 January 2026 his associate bought 18,800 more at EUR 52.71 to EUR 53.78.C (calculated)⁴²¹⁵ Marais holds 117,874. Koos Bekker’s family trust sold 1,556,807 shares on 17 to 19 December 2025 at EUR 51.24 to EUR 53.68 and kept 90% of its holding.¹⁵

Put together, the largest award keys on total market value and relative return, not on net asset value per share, and the discount goal paid nothing in two years.¹⁵ That pays management to lift the share price, not to close the gap an owner waits on. Bloisi has his own money in, and those purchases are down 29%.C (calculated)⁴²¹⁵ Over the cycle, $23.0B bought back Prosus at $44.74 a share, now worth 14.7% less.C (calculated)¹⁰¹⁹²⁰

There is one related party to read with the guidance. Bloisi holds 3.2% of the group’s food holding company, Movile Mobile Commerce Holdings, whose minority holders have written put options. Prosus carries a $312M liability for his put, up from $306M, and in FY2026 he sold part of his interest to the group for $24M.¹⁰ The accounts say puts are valued from the underlying business: 52% of the total balance by discounted cash flow on ten-year forecasts, and the note names transaction prices as the other method. They do not say which applies to his.¹⁰ If his is a discounted cash flow, a lower iFood forecast lowers what the group owes him, so the FY2027 guide runs against him on this item. Pro rata, $312M for 3.2% implies $9.75B for 100% of the food holding company, 57% above the $6.2B consensus for iFood and 2.6 times the $3.7B I count.C (calculated)⁴⁴¹⁰ It is a book figure under the put’s terms, not a price, and the holding company may hold more than iFood. I do not use it, and it sits well above every other value in this memo.

High marks for the profit turn since FY2024, a quarter of it bought; the FY2023 to FY2024 part belongs to his predecessor. A mixed mark for capital allocation: the acquisitions are marked below cost, the buyback paid more than today’s price, and the biggest award keys on total market value and relative return, not on net asset value per share.

VFinancials

Reported profit says little about Prosus. Net profit attributable to shareholders was $11.64B in FY2026, and $4.86B of that was gains on partial disposals of equity accounted investments, $4.7B of it from Tencent. Headline earnings, which take those out, were $6.32B. Core headline earnings, Prosus’s own measure, were $8.34B, and Tencent supplied $8.12B of them.R (reported)¹⁰¹² The core measure is not IFRS: it adds back $413M of amortisation and $985M of share-based pay at the equity accounted investments, and the IFRS share of results of all of them was $6.87B.R (reported)¹⁰

The consolidated businesses are the part that is changing. Revenue grew from $4.62B in FY2022 to $9.71B, 20.4% a year; in FY2026 alone it rose $3.54B, $2.77B of it from the three acquisitions.C (calculated)¹⁰¹²²⁷ Group adjusted EBIT went from a loss of $537M in FY2022 to a profit of $619M.R (reported)¹⁰²⁷ Free cash flow, on Prosus’s definition, was $1.51B, of which Tencent’s dividend was $1.24B.¹⁰¹²

The cash that went out is bigger than the cash that came in. In FY2026 Prosus spent $7.9B on acquisitions, $7.4B on its own shares and $0.5B on dividends and capital repayments, $15.8B in all, against $1.5B of free cash flow and $6.9B from selling Tencent shares.C (calculated)¹⁰ Group cash and short-term investments fell from $19.0B to $12.5B and interest-bearing debt rose from $16.2B to $17.0B, or $17.5B with leases.¹⁰¹⁹ At the holding company, cash was $9.5B and debt $15.9B on 31 March 2026, which is a net debt of $6.4B, a gross loan-to-value ratio of 9.8% on Prosus’s measure, which counts unlisted assets at half their value, and ratings of Baa2 at Moody’s and BBB at S&P.¹³ Prosus’s pro forma adjusts those balances for asset sales (Delivery Hero and Remitly shares), investments (Rapido and Alan), share repurchases and bonds raised and settled in July 2026: cash $9.0B, debt $16.4B, net debt $7.5B. Its page showed $7.4B on the same cash and debt a week earlier, so I read the $0.1B as rounding. I use the pro forma figure throughout.⁴⁴

Accounting to watch

1. Net asset value is consensus. The $33.2B for the unlisted businesses is the figure from 18 analysts, which Prosus says it did not verify, and it is 24.0% of gross assets.C (calculated)⁴⁴
2. Tencent’s profit is not Tencent’s cash. The $8.12B counted in core headline earnings compares with $1.24B received. Tencent’s owner free cash flow was RMB108.6B in the twelve months to June 2026 against RMB151.5B in 2025, as the AI build took more cash, and if that continues the gap widens.¹²³⁷
3. Definitions moved. Adjusted EBITDA on the FY2026 definition excludes all share-based pay, free cash flow changed definition in FY2024, and per-share figures before FY2023 are not comparable because of the September 2023 capitalisation issue.¹⁰¹⁹²⁰
4. The book value of Tencent. Under IFRS the stake sits inside $44.8B of associates against a market value of $127.6B at 31 March 2026. Prosus treats the gains on sales as non-taxable: tax was $117M on $11.8B of pre-tax profit.¹⁰
5. Puts. Written put options held by minority holders total $797M, $738M of it exercisable within a year, settled in cash or shares at the group’s choice.¹⁰

VIValuation

First, how this goes badly wrong

Before any numbers, I asked what would most likely hurt an owner. Three things can.

• Tencent falls and the discount stays. The stake is 71.2% of what Prosus owns.¹⁰¹² At my bear value of HK$263 it is worth 30% less, and the bear case is $22 a share.E (estimate)³⁷⁴⁴
• The cash runs out first. In FY2026 Prosus spent $15.8B against $1.5B of free cash flow and $6.9B from selling Tencent shares, and holding company cash fell to $9.5B.C (calculated)¹⁰¹³
• The food bet earns too little. iFood is guided to $100M to $150M from $400M, against about $290M for my count.¹²¹⁷¹⁸

None needs a surprise.

What an owner can take out

Core headline earnings were $8.34B, after adding back $985M of share-based pay at associates.¹⁰ Of Tencent’s $8.12B, $1.24B arrived as dividend, about 15% C (calculated)¹²³⁷ Free cash flow without it was $275M, against $645M a year of interest.¹⁰¹²¹³ The rest reaches owners only if Prosus sells Tencent shares. So I value the stake from Tencent’s owner cash flow, not its profit.

What I can’t judge

I can’t tell you what iFood, JET and PayU are worth. The $33.2B for the unlisted businesses is an average of 18 analysts, which Prosus has not verified.⁴⁴ No filing names a route to close the discount.¹⁰ So read the $49 with the range $22 to $90. A 25% discount alone adds $6.59 to my base.E (estimate)⁴⁴

What the price implies

Start with the sum of the parts, valued at the Q4 2026 close.

The market value is $79.5B. The Tencent stake alone is $98.3B. Take the stake out of the market value and what is left for iFood, OLX, JET, PayU and everything else, after the net debt, is minus $18.8B. If the Tencent stake is worth its price, the market values the unlisted businesses at minus $17.9B, against the analysts’ $33.2B.C (calculated)⁴⁴⁴⁵⁴⁶

Count the unlisted businesses at nothing and the Tencent stake and the other listed holdings, less the net debt, are still $97.5B, which is 23% more than the market value. The price rests on the structure.C (calculated)⁴⁴⁴⁶

Hold Tencent at its close and the price is a 39.1% discount to the net asset value Prosus prints, and 34% to mine.

Two net asset values appear on this page, and every discount says which one it is measured against. Prosus’s takes Tencent and the other listed holdings at the close, the unlisted businesses at the consensus of 18 analysts, and subtracts net debt: $130.6B. The price is 39.1% below it, and the reported series in the chart is on the same measure. Mine counts 75% of the consensus for the unlisted businesses, which takes off $8.2B, and subtracts $2.1B of holding company costs: $120.3B. The same price is 33.9% below it. The 5-point gap is those two lines. I apply the discount to mine, and the cases table restates each case on Prosus’s measure so it can sit beside the history.⁴⁴²⁰

The reported series is in the chart: 54% in April 2022, 38% in March 2023, 37% in March 2024, 42% in March 2025 and 42.8% in March 2026.¹⁵²⁰²⁹ The discount fell 16 points in FY2023, from 54% to 38%; the buyback began that June. Tencent closed at HK$385.80 on 31 March 2023 against HK$378.80 on 1 April 2022, 2% higher, after a weekly close as low as HK$200.80 in October 2022. By March 2026 the discount was 4.8 points wider than in March 2023, after Prosus had spent $23B on its own shares.C (calculated)²⁰¹⁵⁴¹ Those are year-end readings. Between them it moved further: about 32% on 1 December 2025 by my sum, before Tencent fell.C (calculated)⁵¹

The question can be asked the other way, by holding the discount and solving for Tencent. This table shows what Tencent has to be worth for the price to be fair.

At the four March readings since 2023, 37% to 43% on Prosus’s measure and 32% to 38% on mine, the price is fair for Tencent between HK$365 and HK$408 at my count of the unlisted businesses. At today’s 33.9% it needs HK$378, where Tencent closed, and at 43%, my bear-case discount, HK$451.C (calculated)³⁷ With no discount at all and the rest of the group at consensus, the price needs Tencent at about HK$189. In the Tencent model, HK$190 meant owner free cash flow falling 4.1% a year for 15 years from a trailing year that the AI build had already pulled down; I have not solved the model again for HK$189. That is an extreme case, because no holding company trades at its net asset value. Tencent’s close in the first version of this page, HK$386, needed growth of 7.0% a year, and I have not solved it again for HK$378. Both rates are a constant 15-year growth in owner free cash flow with 3% growth after year 15 and a 10.0% discount rate, from the Tencent model.C (calculated)³⁷

Bear, base and bull

The model is a sum of the parts per economic share, 2,082.8M of them, in dollars at the ECB rates of Q4 2026.⁴³⁴⁵ Tencent is valued from my own Tencent analysis, published on this site in Q4 2026: HK$263 in the bear case, HK$508 in the base and HK$734 in the bull, from a 15-year owner free cash flow model at a 10.0% discount rate with cash flows at year end, dated from 30 June 2026. That page also shows that mid-year timing dated Q4 2026 would lift each case by about 7%, to HK$278, HK$542 and HK$787. I use the stated values, which are the lower; at HK$542 my base would be $52.E (estimate)³⁷ The other listed holdings, $6.6B, stay at the close in every case, because I have no view on them.⁴⁴ The unlisted businesses start from the consensus of 18 analysts, $33.2B.⁴⁴

I count the unlisted businesses in two groups. The first is the businesses valued on profit that already exists: OLX, OLX Brasil, Despegar and eMAG, $12.7B of consensus, at 13 to 20 times their FY2026 adjusted EBITDA. I used no peer multiple; the range is what the consensus itself implies.C (calculated)⁴⁴¹² The second is the rest, $20.5B. The analysts value iFood at 41 to 62 times its FY2027 guide, JET at 38 times its FY2027 target of $100M, PayU at 200 times its first positive year and iyzico at 43 times, and $5.7B is in lines Prosus does not itemise.C (calculated)⁴⁴¹²¹⁷ In the base case I count the first group in full and the second at 60%, which gives $25.0B, 75% of consensus. The NAV page does not say whether each value is for the whole business or for Prosus’s share; the holdings are 96.2% of iFood, 99% of OLX and 88.2% of eMAG, so I did not adjust.⁴⁴¹²

The count puts iFood at $3.7B, 9 times its FY2026 profit and 30 times the midpoint of its FY2027 guide. At 13 times, the low end of the multiples I accept for profit that exists, $3.7B needs about $290M of adjusted EBITDA, 15% of FY2026 revenue of $1.87B. iFood earned 21% in FY2026, and the FY2027 guide of $100M to $150M is 5% to 8% of that revenue. So the count needs iFood’s margin to recover to about 15%, below its FY2026 level and above the guide.C (calculated)⁴⁴¹²¹⁷

Then I take off holding company net debt, $7.5B, and the holding company’s costs, $214M a year capitalised at the 10% I use for Tencent, 5 points over the 5.00% ten-year Treasury yield, $2.1B.E (estimate)⁴⁴¹²³⁵ A discount applies to what is left. Prosus’s reported discount has no cost line, and Prosus lists management costs among the reasons for the discount, so taking costs off before the discount counts them in part twice. The effect is $1.03 a share before the discount.E (estimate)²⁰

The discount is the assumption the page is most exposed on. My base case does not move it: it stays where the price puts it, 33.9% of my net asset value, which is 38% of Prosus’s in that case. The buyback has not narrowed it in three years and I found no announced step that would, so a base that needed it to close would need an event I cannot name. The bear case takes the widest reading Prosus has reported, 54% in April 2022, which is 43% of my net asset value in that case because the bear case already takes more off the unlisted businesses. The bull case takes the narrow end of Prosus’s own audited band, 15%: the statements say holding company discounts are normally in a 10% to 40% range and that Prosus has mostly traded between 15% and 40% below its equity value since listing.²⁰ On my measure the bull case is 15%, on Prosus’s 16%. Narrowing to 25%, the middle of the band, is not in any case; it adds $6.59 a share to the base.E (estimate)

The base value is the price plus one step. Tencent at HK$508 instead of HK$378, with the discount unchanged, takes the value from $38.18 to $49, a gain of $10.75 a share that is the whole gap between the price and the base.E (estimate)³⁷ The base does not need the discount to move. The bull case adds three more steps: $18.68 from Tencent at HK$734, $2.60 from counting the unlisted businesses in full, and $20.08 from a discount of 15%. Half of the distance from the base to the bull is the discount.E (estimate)

Net asset value at my values is $154.2B, $74.03 a share. It starts from $62.71, Prosus’s $130.6B over 2,082.8M shares. The rows I add up give $62.73, because the unlisted rows sum to $33.2B and Prosus prints $33.1B, and I use the rows from here. Tencent at HK$508 instead of HK$378 adds $16.26 a share. Counting 75% of the unlisted businesses takes off $3.94, and the holding company’s costs $1.03. The 33.9% discount then takes off $25.10 and leaves $49.E (estimate)⁴⁴³⁷

The bear case is April 2022 again. Tencent falls to HK$263, 30% below its close, which is where my Tencent analysis puts it when advertising falls and AI spending stays high. The unlisted businesses are worth $15.7B, 47% of consensus: the second group is counted at 30%, which puts iFood at $1.9B, 15 times the midpoint of its FY2027 guide, and JET at $1.1B. The discount is 43% of my net asset value, which is 54% of Prosus’s, the April 2022 reading. The value is $22, 42% below the price.E (estimate)¹⁷³⁷

The bull case is the analysts’ unlisted values holding, Tencent at HK$734, 94% above its close, and a discount that closes to 15%. The value is $90, 136% above the price.E (estimate)⁴⁴³⁷

The three cases are joint scenarios. Each moves Tencent, the unlisted count and the discount together, so they are not independent odds, and I do not weight them or quote a probability-weighted value. The cases also lean on one source: if my Tencent analysis is wrong, this page is wrong in the same direction.³⁷ Most of the range is Tencent. Holding Tencent at its close and varying only the rest, the bear-to-bull range is $30 to $52. Holding everything else at the base case and varying Tencent from its bear to its bull value, it is $29 to $68.E (estimate) Each HK$100 of Tencent value is worth $8.26 a share at a 34% discount, and ten points of discount is worth $7.40 a share at my base net asset value.E (estimate)

The cells below the $38.18 price sit in the top right, where Tencent is worth little and the discount is wide. A reader who rejects my Tencent value can read across the HK$378 row: at a 25% discount the value is $43, at 34% it is $38 and at 43% it is $33. A reader who rejects my discount can read down the 43% column: Tencent at my base value gives $42, and at my bull value $58.E (estimate)³⁷

What closes the discount

The base does not need the discount to close, but the bull case does, so here is what could close it and what the record says about each. The buyback is the one tool at work. The discount fell 16 points in FY2023, the year the buyback began, and in FY2026 $7.9B returned to shareholders left it flat at about 42%, March to March.C (calculated)¹⁰¹⁵²⁰ Management now guides about $5B a year across Prosus and Naspers, and the CFO describes a deliberate reduction in the pace of Tencent disposals.¹⁰¹⁷ The structure is the second. The annual report says the board focused on simplifying the group structure; I found no announced transaction, and Naspers still holds 75.66% of the votes.¹⁰ The third is cash. If Uber’s offer closes, Prosus receives $2.36B in the second half of 2027, 3.0% of its market value and small against the gap.C (calculated)³⁸ The fourth is the incentive: the discount goal in the chief executive’s bonus paid nothing in FY2025 and FY2026 and stays for FY2027.¹⁵ No filing names a route, so the bull case is my view that one appears.

Prosus against Tencent itself

At the same price, $38.18 in Tencent shares earns the Tencent return, and in my base case that is 34% against 28% through Prosus. The discount does not move in the base case and the rest of the group does not rise with Tencent, so Prosus lags. Prosus beats Tencent only if the discount narrows or the unlisted businesses are worth consensus, and it loses more if either goes the other way.C (calculated)³⁷

Three items sit outside the cases.

1. Delivery Hero. Prosus will tender its 51.1M shares at Uber’s EUR 41.50, against a close of EUR 37.50. That is EUR 2.12B, $2.36B, against $2.13B at the close, a gain of $0.23B or $0.11 a share, if the offer closes in the second half of 2027.C (calculated)⁴⁴³⁸
2. Claims below the holding company. I deduct no subsidiary debt, leases, minority interests or put options, and no tax on a Tencent sale, because Prosus books the gains as non-taxable. Sized from the filing, subsidiary debt is about $1.1B (group $17.0B less central $15.9B), leases $0.5B ($17.5B with leases against $17.0B without) and written puts $0.8B: $2.4B, or $0.76 a share in my base case and $1.15 before the discount. I do not know whether the analysts’ values are before or after these, so I leave them out of the cases. Each $1B of any of them is $0.32 a share in my base case, $0.48 before the discount.C (calculated)⁴⁴¹⁰
3. Fewer shares. At the guided $5B a year and today’s price Prosus would retire about 6% of its shares in a year if all of it were Prosus’s. Value per share today does not count the 2.1% a year that adds in my base case.C (calculated)⁴²¹⁰

The margin of safety

• Price$38.18Euro close in dollars, Q4 2026 [42]
• Base value$49My estimate E (estimate)
• Margin of safety below$3235% under my base value C (calculated)

I set the margin of safety at 35% below my base value, which gives $32, EUR 28.6. My rule for a discount to base value is 25% for a business with net cash and no customer above a tenth of revenue, 30% when one cycle or two customers drive results, and 35% when fixed charges such as leases sit ahead of the shares. Prosus is in the top tier for two reasons. One asset is 71% of what it owns and 97.3% of what it earns, which is the 30% tier, and 5 more points go for control and fixed charges: Naspers holds 75.66% of the votes, and the group’s $645M a year of interest expense is covered 1.9 times by Tencent’s dividend alone. Holding company net debt is $7.5B on Prosus’s page, from $16.4B of gross debt and $9.0B of cash before rounding.C (calculated)^44101316 The line sits on top of the 34% discount already in my base case, so it is 43% of my base net asset value, which has already taken a 25% haircut off the unlisted businesses. The price is 20% above the line, so the stock would have to fall about 17% to reach it. The lowest close in the past year was EUR 33.93, on 22 October, 19% above the line.C (calculated)⁴²

Where this could be in 5 and 10 years

Prosus has no cash flow model of its own, so I roll forward the part I can: the Tencent stake. Everything else stays at today's case value: the unlisted businesses at $15.7B, $25.0B and $33.2B, net debt and costs, the discount and the share count. That leaves out growth in the unlisted businesses and the buyback, which adds about 2.1% a year in my base case. I cannot give a year-10 value for iFood or Just Eat Takeaway.com, so I do not try.E (estimate)³⁷⁴⁴

For the base to be right at 5 years, Tencent has to be worth about HK$650 a share, against HK$378. By year 10 it is HK$756. That needs owner free cash flow to go from RMB 63B in year 1 to RMB 270B in year 5 and RMB 344B in year 10. It rests on revenue growth of 6.8% a year, a margin before depreciation rising from 40.5% to 45%, capital spending falling from 28% of revenue to 17%, and a 10.0% discount rate. The discount to net asset value stays at 34%, so the base needs no event to close it.E (estimate)³⁷

The bear has spending stuck near 21% of revenue, advertising down 7% in year 2, the unlisted businesses at 47% of consensus and a 43% discount. The bull has year-10 owner cash of RMB 499B, the unlisted businesses at full consensus and a 15% discount. The most likely way the base is wrong is capital spending. Each 2 points of Tencent's long-run spending is worth about HK$27 a share, or $2.20 of my base value for one Prosus share today.E (estimate)³⁷

These are estimates of value, not forecasts of the share price. The market price can sit above or below value for years: Prosus says it has mostly traded 15% to 40% below its equity value since listing.²⁰

The next best use of the same dollar

The nearest alternative on this site is Tencent itself, with a base return of 12.4% a year over ten years against 9.8% here.E (estimate)³⁷ Prosus gets ahead only if the discount narrows or the unlisted businesses reach consensus.

How good the business is and how good the price is are two questions. The business is one strong asset and marketplaces still proving they earn cash. But $23B of buybacks left the gap wider, so a low price has not yet been a good return.C (calculated)¹⁰¹⁵²⁰

The bull case needs Tencent at HK$734, the unlisted businesses at full consensus and a 16% discount. At one chance in three, one in three and one in four, that is about 3%.E (estimate) These odds are guesses, not independent, and not in my values.

VIIThe case for the price

The price, $38.18, is already below my $49 base value. So the question here is what closes the gap.

What most people miss

The buyback no longer leans only on Tencent. Since November 2025 sales of other assets also pay for it. The CFO counts about $2B last year and almost $1B more in “the first quarter of this year.”¹⁴ Tencent shares sold fell from 177.9M in FY2024 to 95.6M in FY2026.²⁰¹⁰ Tencent behind each Prosus share still rose from 0.963 to 0.981 in the seven months to October.C (calculated)⁴⁴⁴⁵

The 37% I quote is the narrowest year-end reading, not the narrowest ever. Prosus says it has mostly traded 15% to 40% below its equity value since listing.²⁰ Business Day put it at about 30% on 1 December 2025, $143B on $203B of net asset value.⁵¹ That market value seems to count treasury shares. Net of them I get about 32%.C (calculated) The Tencent stake was then $164B, against $98.3B today.

The bull case

As I read it, the market pays 81 cents for each dollar of Tencent and less than nothing for the rest, and a company buying its own shares at that price can force the gap shut. That last step is my reading. The closest holder is David Herro’s Oakmark International. In 2022 it welcomed the buyback because it “takes advantage of the massive discount in Prosus shares relative to its NAV.”⁵² In June 2024 it called Prosus “the optimal way to gain access to Tencent.”⁵⁸ The later commentaries I could open do not mention Prosus, but it was 2.8% of the fund at 30 September 2026.⁵³ Old Mutual’s private client team wrote in July that it expects the discount at Naspers, the parent, to narrow as the group delivers.⁵⁴

Fabricio Bloisi told investors in June he will build “another 50 to $100 billion ecosystem in Latin America.” Analysts put today’s Latin American businesses at $8.3B.¹⁴⁴⁴ CFO Nico Marais said Prosus wants to stay a large long-term Tencent shareholder.¹⁴ On Investing.com’s tally, 18 analysts average EUR 59.49, about $66, from EUR 37 to EUR 80. JPMorgan’s Marcus Diebel is at EUR 71, Goldman Sachs at EUR 40. I have not read the notes.⁵⁵⁵⁶

Against reality

My base value is $10.75 above the price, and two things can close that. If Tencent reaches HK$508, 34% above its close, Prosus gets to $49 with the discount unchanged. If Tencent stays at HK$378, Prosus’s discount has to fall from 39.1% to 22%, narrower than any figure I found since the buyback began in 2022. A mix asks less: Tencent at HK$450, up 19%, with a 32% discount also gives $49.E (estimate) My $90 bull value is much further off. It needs HK$734, the analysts’ $33.2B intact and a 16% discount.E (estimate)³⁷⁴⁴

So if Tencent rises by a fifth, the discount gets back near its December 2025 level, and iFood and JET hit their targets, $49 is possible. Today’s numbers are not there yet. The buyback alone will not do it. At the guided $5B, value per share grows about 2.3% a year, so with all else flat the gap takes about 11 years.E (estimate)⁴²¹⁷ About $23B of buybacks since March 2023 left the discount wider, and no filing names a route.C (calculated)¹⁰¹⁵²⁰ Profit is behind too. iFood is guided to $100M to $150M against the $290M or so my count needs. The FY2028 ambition, above $2.4B, is 1.9 times FY2026, and the CEO says FY2027 is not a year for higher profit.C (calculated)^17141028

What they have to hit

Where I come down

The new funding of the buyback moved me most, because Prosus now adds Tencent per share while selling less of it. Bloisi’s ecosystem goal and the brokers’ targets did not move me, since none of them names a route to close the discount. A named change to the Naspers and Prosus structure, or iFood above $150M with JET at $100M, would raise my values.

VIIIRisks, and what would change my view

The risk Prosus puts first is the one I weight most:

Our significant Tencent stake represents a material concentration.

1. Tencent and policy. The stake is 71.2% of gross assets and 97.3% of core headline earnings. The report says geopolitics could force divestments or “stranded value”. Prosus has sold about 0.9% of Tencent a year to fund the buyback, and its holding fell from 31.1% to 22.5%.C (calculated)¹⁰¹²
2. The discount stays or widens. It was 54% in April 2022 and 42.8% in March 2026, and Naspers holds 75.66% of the votes. About $23B of buybacks since March 2023 has not narrowed it. My base case holds it where it is, so this risk is in the bear case and not in the base.C (calculated)¹⁰¹⁵²⁰
3. The Brazil subsidy war. iFood’s FY2027 adjusted EBITDA is guided to $100M to $150M from $400M, and the second half of FY2026 already lost 9.7 points of margin. Management expects rivals to keep spending, and no independent market share exists.¹²¹⁷¹⁸
4. JET. Orders fell 8.8% in the six months to March, the analysts’ value is 22% under the $4.9B consideration, and the FY2027 target is $100M on $3.6B of revenue. Management says orders turned in September 2026 and gives no number.^10121718
5. AI and content. Stack Overflow’s traffic has fallen by more than half and OLX’s paying listers fell 6.1% in the second half while price carried growth.¹⁰¹²
6. Capital allocation and funding. Prosus spent $7.9B on acquisitions in a year. Holding company cash fell from $17.2B to $9.5B, interest cover is 3.8 times, and unrealised currency losses on euro bonds were $372M in FY2026. Tencent’s dividend covers interest expense 1.9 times.C (calculated)¹⁰¹³
7. Control and conflicts. One family-linked voting pool controls Naspers, which controls 75.66% of Prosus, and the chief executive holds a stake and a put in the food holding company.¹⁰¹⁶
8. My own inputs. The base case uses my Tencent value and my count of the unlisted businesses. If Tencent is worth its close, HK$378, instead of my HK$508, my base falls from $49 to $38, the price. The Tencent value comes from a separate analysis; if that changes, this page changes in the same direction.E (estimate)³⁷

What would move me

The first test comes on 25 November 2026, when Prosus reports the first half of FY2027, the first period in which iFood runs under its own guide.⁵⁰ I will grade it against the signposts above: iFood against its $100M to $150M guide for the year, JET orders and profit, OLX paying listers with the margin at 48% or better, and free cash flow excluding Tencent against $275M. The October OLX briefing gave no first-half figure for any of them: no lister count, and the first-half margin bar without a number.⁴⁹