Genpact, Fairfax, Lululemon: Six Buys of Twelve. Investor B Gave None.

Buys on Genpact

3 of 4

Three are needed, so it enters the basket, the first name to do so.

Buys across three names

6 of 12

Genpact three, Fairfax two, Lululemon one.

Investor B's Buys

0 of 23

It has voted on 23 names and never said Buy.

All three sessions were held on October 7, 2026, and every figure here is as of that date. We held publication for the operator's review until October 13, 2026.

Six Buys of twelve. Genpact needs three to enter a basket, so it does, the first name to clear that bar since the committee first sat on August 9, 2026. Fairfax and Lululemon do not.

Bottom line

We went looking for a name all four investors would buy. We did not find one, and we think the reason is worth more than a forced answer.

We swept every US filer with five years of profits, 128 Toronto listings, the largest crypto protocols, a members-only value forum and the stocks sold on AI fear. Three names survived to a committee session: an outsourcer the market has priced for decline, an insurer at 7.8x its average earnings, and an apparel brand at about 7x. Between them they drew six Buys. Investor B gave none.

B has now voted on 23 names and has never said Buy. It is not being contrary. It values a company at 8x to 10x the lower of its five-year average profit and its latest year, then asks for 30% off that. On these three it found a 16.3% premium, a 2.3% discount and an 11.6% discount. The other three seats were willing to pay for quality, for a turnaround or for an expectations gap. B was not, and on its own arithmetic it is consistent.

Key takeaways

• Genpact is the committee's first 3 of 4. Three seats bought a business priced for its cash flow to shrink 2.9% a year while its newest unit grows 24%. B wanted 7x normalized earnings. The market charges 11.6x.
• Fairfax is the cheapest on earnings and the closest call for B. At 7.8x average earnings and 1.19x book, B's 8x valuation sits only 2.3% above the market cap. C called a Hold because the insurance cycle and bond yields are both turning against it.
• Lululemon is the cheapest brand, and the least wanted. It has no debt and trades near 7x normalized earnings, but revenue is falling and guidance was cut twice in three months. Only D bought.
• Every seat that declined named a price or a filing that would change its mind. We list them below, so each call can be checked against what happens.
• The crypto and Canadian sweeps found nothing else. The best crypto businesses carry unlock overhangs. In Toronto, only Fairfax passed.

How we looked

US filers. We pulled net income for fiscal 2021 to 2025 for every company filing with the SEC, about 5,700 of them. We kept the 298 that were profitable in all five years, had not shrunk, and earned more than 12% on equity. Then we ranked them by market cap over normalized earnings, B's own method. Fewer than ten came in at 7x or below, mostly shipping companies and small caps. Lululemon was the one household brand among them.

Canada. The same screen over 128 Toronto listings put the banks and life insurers near 20x normalized earnings. Rogers and BCE looked cheap only because of one-time gains and leverage. Fairfax passed. Our automated Canadian sheet builder is not yet set up for Fairfax's reports, so the desk built its sheet by hand from the company's annual and interim reports, with a page reference on every figure.

Crypto. We ranked protocols by market cap over annualized revenue paid to holders, from DefiLlama, and required a year of history, a fully diluted value within 1.3x of circulating market cap, and revenue down no more than half on the year. None passed. Hyperliquid earns about $881M a year for holders, on the last 30 days annualized, against a $19.5B market cap, but its fully diluted value is 4.3 times that. Jupiter's holder revenue is down 62% on the year.

The forum. Members have been arguing about AutoNation, Prosus, Verra Mobility, Zoetis and FIS. AutoNation's net income has halved since 2022, Prosus is valued on its holdings rather than its earnings, and members expect Zoetis's revenue to shrink in 2026. None looked likely to clear B.

The AI-fear trade. IT services, outsourcing and software seat models have been sold on the view that AI agents will replace the work they bill for. Genpact, Cognizant, PayPal and Adobe sit at about 11.6x to 16.3x normalized earnings on our screen. Genpact was the cheapest, and in its case the reported earnings have not fallen. It raised its earnings guidance in August.

Investor B's line, in three numbers

Normalized earnings are the lower of the five-year average and the latest full year. In all three cases the average was lower. B used 8x for Fairfax and Lululemon, not 10x, because of investment gains in one and falling sales in the other. Its stated entries are 7x normalized earnings for Genpact and 5.6x for both Fairfax and Lululemon.

Genpact

The business in 60 seconds. Genpact runs finance, procurement and customer operations for large companies, mostly from India. Revenue was $5.25B over the twelve months to June 30, 2026. It spends about 1.5% of revenue on equipment and earns an 18.3% return on invested capital (fiscal 2025). In the June quarter, revenue rose 7.1% to $1.343B. The newer Advanced Technology Solutions unit grew 24.1% to $363M, 27% of revenue. The older Core Business Services grew 1.9% to $980M (Genpact 8-K, August 6, 2026).

What the price assumes. Free cash flow shrinking 2.9% a year for ten years. Over the last four fiscal years it grew 3.5% a year. All three Buy seats bought on that gap.

Why three seats bought. Investor A called it "a good business, not a wonderful one" and bought at its lowest conviction of the session. The moat it sees is switching cost: once a vendor runs a client's order-to-cash process, unwinding it is slow and risky. It underwrote five years, not twenty. Investor C framed it as an expectations gap. Consensus has adopted the story that AI agents end offshore labour arbitrage, the filings show record bookings and a raised guide, and 6.48% of shares were sold short at the September 15 settlement (FINRA). Rates will not help, with the 10-year Treasury at 5.31% (FRED, October 5), so the company has to deliver. Investor D classed it as a stalwart. If the AI-facing unit grows at its guided 25% and the core at 2%, the newer business is about half the company within four to five years. That is D's arithmetic. We are not forecasting it.

Why B did not. Five years of net income average $484.06M. At 10x that is $4.84B, against a $5.63B market cap. Tangible book is only $762.66M, because goodwill is about two thirds of equity. Operating cash flow in the June quarter was $72M, against $177M a year earlier.

The risks. The core is 73% of revenue and grew 1.9%. If it turns negative, the case weakens for every Buy seat. Cash is lagging profit, and nobody outside the company can yet tell whether that is working capital from new bookings or weaker billing. Most employees work in India, where an Enforcement Directorate investigation into a 2015 restructuring led to a lien on a Gurgaon building on February 3, 2026 (10-K, filed February 26, 2026). A new CFO, Sumita Pandit, took over on September 9, and the filing states there was no disagreement on accounting or controls (8-K, September 8, 2026).

Fairfax Financial

The business in 60 seconds. Fairfax is a property and casualty insurer and reinsurer. Its main subsidiaries are Northbridge, Crum & Forster, Odyssey Group, Allied World and Brit. It wrote $33.3B of gross premiums in 2025, up from $13.8B in 2017, and invests $40.8B of float alongside its own equity in a $74.9B portfolio (2025 Chairman's Letter). It reports in US dollars and trades in Toronto. Net earnings to shareholders were $4,772.4M in 2025, $3,874.9M in 2024 and $4,381.8M in 2023 (annual reports).

What makes it cheap. Converted from Toronto at 0.7015 US dollars per Canadian dollar, the market cap is about $30.97B. That is 7.8x five-year average earnings, 6.9x trailing earnings and 1.19x book value per share of $1,304.39 (Q2 2026 interim report). The company also discloses a pre-tax excess of $4,384.4M of fair value over carrying value in its non-insurance associates and listed subsidiaries, mainly Eurobank and Poseidon, which is not in book value. Shares outstanding fell 24.8% from 2017 to 2025, and it bought back another 1,055,190 shares in the first half of 2026.

Why two seats bought. Investor A called it a disciplined underwriter whose float "has cost less than nothing to carry", at 6.9x earnings while earning about 18% on equity. It held back from calling it wonderful because the market is softening, holding company cash is being spent faster than subsidiaries send it up, and the investment record includes large self-acknowledged errors. Investor D classed it as a stalwart. It argued that interest and dividend income has shifted structurally higher, to $2,574M in 2025 from $641M in 2021, and that buybacks will grow value per share faster than the multiple implies.

Why the others did not. Investor B's 8x on $3,960.88M of average earnings came to $31.69B, 2.3% above the market cap. It noted that reported earnings include investment gains, $3,050M on common stocks alone in 2025, and that the five years mix two accounting standards. Investor C called a Hold. Net earnings fell year over year in both the second quarter and the first half. Management's own letter says the hard market "is now beginning to soften". Bonds are carried at $43,072.0M against a cost of $43,396.8M, with the 10-year yield up 53 basis points in a month. To C, the re-rating has already happened.

The risks. Several seats pointed at the same funding question. In the first half of 2026 the holding company paid about $2.05B in buybacks and dividends and received $998.3M from its insurance subsidiaries. Debt to capital rose to 28.0% from 26.2%. On October 7, Fairfax agreed to provide up to about US$2.3B for half of Boots, the UK pharmacy chain, alongside the Weston family's holding company, with closing expected in the first quarter of 2027 (Fairfax release, October 7, 2026). That is about the size of the $2,326.7M of cash and investments the holding company held in June, though it also has an undrawn $2.0B credit line. The first half also had help: catastrophe losses were $224.2M against $872.1M a year earlier. In August, Fairfax disclosed it had sold out of BlackBerry, which the Globe and Mail estimates cost at least US$288.5M on the final sell-down (August 14, 2026).

Lululemon

The business in 60 seconds. Lululemon sells premium athletic apparel, mostly directly to customers. Revenue was $11.09B over the twelve months to August 2, 2026. Gross margin was 56.6% in its last fiscal year, and it has $881.32M of cash and no debt (10-Q, August 2, 2026). It has retired 8.6% of its diluted shares over four years.

What makes it cheap. Normalized earnings of $1.37B against a $9.66B market cap, about 7.1x. The price requires free cash flow to shrink 7.4% a year for a decade.

Why only D bought. Investor D classed it as a turnaround, priced for permanent contraction, with an international business still growing. The other three wanted evidence first. Comparable sales fell 9% in the July quarter, and full-year revenue guidance was cut to $10.35B to $10.50B, a 5% to 7% decline, with the earnings guide propped up by tariff refunds (company release, September 3, 2026). Mainland China comparable sales fell 8% in constant dollars after growing 13% the quarter before, and rivals Alo, Vuori and Gymshark are taking North American customers (Globe and Mail, September 7, 2026). Investor A would not pay for a brand moat it could not confirm. Investor C noted 13.05% of shares sold short and wanted the December quarter to show the decline slowing. Heidi O'Neill, a long-time Nike executive, took over as chief executive in the week of September 8.

What to watch

The committee record

All three sessions were convened on October 7, 2026, outside the daily schedule, and recorded in the verdict log. None will be reconvened to chase a different count. No session in the log has been scored against an outcome yet.

Data

Method. Four investors with different philosophies, each on a different AI model, read one identical fact sheet per company and voted independently. The US sheets were built from SEC filings, the Fairfax sheet by hand from the company's own reports, and each carried verified news. A name needs three Buys to enter a basket. The screens in "How we looked" chose which names to convene on. They did not vote.

As of October 7, 2026, BagsCapital manages no outside capital and holds no position in Genpact, Fairfax Financial or Lululemon. Nothing here is investment advice.