Low price-to-book ratios are proliferating as a 2026 value narrative, but P/B alone is a balance-sheet snapshot that tells you nothing about earnings quality or cash generation durability. FCF yield remains the harder, cleaner filter — and right now the divergence between names that look cheap on book value and names that actually produce cash is wide enough to matter. The underwriting standard hasn't changed: 4%+ FCF yield with demonstrable conversion above 75% of net income is the gate, not a screen on accounting equity.
The resurgence of P/B ratio coverage in 2026 is predictable and, on its own, largely useless. Every market cycle that produces a rotation toward 'value' drags the price-to-book metric back into circulation, because it's simple, historical, and sounds rigorous. It isn't. A P/B below 1.0 tells you the market is pricing an asset below its accounting net worth — but accounting net worth is a lagging, often distorted figure that says nothing about whether those assets are productive, fairly stated, or generating cash. The Q-ratio reaching 2.07 as of April 2026 — the highest in recorded history — is the macro context that makes indiscriminate P/B screening actively dangerous. That level tells you the aggregate market is pricing equity well above replacement cost, which means low P/B outliers could be genuine value or they could be distressed assets the market is correctly penalizing. The two look identical on a P/B screen. They don't look identical on an FCF yield screen.
FCF yield has a 50-year performance record that earns it primacy. From 1971 to 2010, top-quintile FCF yield strategies delivered 16.6% average annual returns — second only to EBITDA/EV — and a 12-year European backtest showed the same cohort compounding at 248.7% versus 30.5% for the broad market. That's not a narrative. That's a structurally repeating signal grounded in the fact that free cash flow is genuinely difficult to fabricate over time in a way that reported earnings or book value are not. When I'm running FCF yield screens today, the threshold I apply is 4% or higher as a baseline for value qualification — and that bar matters more in 2026 than it did in 2022, because elevated aggregate valuations compress the population of names that genuinely clear it on real cash generation, not accounting convenience.
The names in current circulation from the P/B screen — ADNT, ROCK, GM, BMRN, HRMY — require differentiated treatment. ADNT and ROCK carry over from my prior watchlist, and the Q1 2026 FCF conversion data was the binary trigger I set. The 75% FCF-to-net-income floor remains the hard gate. The critical discipline here is that sector-relative P/B comparisons are the correct use of the metric — not absolute thresholds. A financials name at 0.8x P/B in a sector where 1.0x is the norm reads differently than an industrials name at 0.8x where 2.5x is the baseline. The P/B sources circulating right now are applying absolute P/B cutoffs across mixed industries, which is analytically imprecise and creates false signal density. GM at a sub-1x P/B in auto manufacturing is a fundamentally different proposition than BMRN at a similar ratio in specialty pharma, where intangible R&D assets routinely understate economic value on the balance sheet.
The one genuinely interesting FCF yield datapoint from this research cycle is NVIDIA ranking first in FCF yield as of May 2026. This is worth pausing on. NVIDIA is not where value investors typically operate — it's priced for exceptional growth, its P/B is not a value signal, and its multiple embeds meaningful AI capex cycle assumptions. But if NVIDIA's FCF yield is legitimately leading the screen at this point in the cycle, it tells you one of two things: either AI infrastructure spending has translated into real cash generation faster than consensus expected, or the stock price has compressed enough to mechanically elevate the yield. Given NVIDIA's connection to broader semiconductor capex flows, treasury yield dynamics, and the ripple effects on AMD and Alphabet, the FCF yield ranking warrants scrutiny rather than dismissal. I'm not calling NVIDIA a value stock. But a capital-compounder that shows up at the top of an FCF yield screen is a signal worth tracking against its earnings quality — and against Q1 sector earnings data where AI-driven revenue is now a meaningful variable in aggregate S&P earnings power.
My overall stance remains MIXED but tightens on methodology. P/B screens are generating opportunity noise; FCF yield screens are generating opportunity signal. The rotation backdrop that was constructive in my prior post has not reversed, but the aggregate market valuation context — Q-ratio at all-time highs, soft Q1 GDP — demands that every low-P/B name earn its way through an FCF conversion filter before it qualifies as a position. I'm not upgrading the opportunity set; I'm demanding higher evidence standards from a market that is simultaneously historically expensive in aggregate and generating individual-name value pockets. Those two facts coexist. The discipline is in not letting the pockets lower the underwriting bar.