Phase Space AI

Financial Model Notes

RB Global [RBA]

RB Global [RBA] — Financial Model Notes

Share count — verified, and the screen used the wrong one

The screen used shares outstanding; the valuation uses diluted 187.5m (+0.64%). Small, but the same class of error that cost LITE 31.5% of EV. netIncome / shares ≈ EPS: $124.6m available to common / 187.5m = $0.664 against a filed diluted $0.66. Clean.

SPLITS: 3.0000 on 2008-04-28 and 2.0000 on 2004-05-07 — both far outside the analysis window. No corporate-action basis error.

TTM revenue — verified

897.7 + 336.9 = $1,234.6m (Q1'26) + 1,222.0 + 1,093.0 + 1,186.0 = $4,735.6m. Screen: $4,716.7m (0.4% low). Used $4,736m.

Net cash — REBUILT; the screen was wrong by 143%

From the Q1'26 10-Q balance sheet ($m):

Short-term debt                                282.0
Current portion of long-term debt               51.2
Long-term debt                               2,271.0
Current operating lease liabilities            132.8
Long-term operating lease liabilities        1,472.3
Series A Senior Preferred (temporary equity)   482.0   <- redeemable, cumulative dividend
                                            --------
Total claims senior to common                4,691.3
Less: Cash and cash equivalents               (667.2)
                                            --------
NET DEBT                                     4,024.1

Screen: −$1,655m. Correct: −$4,024m. Error $2,369m = 143%. EV understated 10.6% ($22,290m → $24,659m).

Three separate omissions: short-term debt and current LTD ($333.2m), operating leases ($1,605.1m), and the Series A preferred ($482.0m). The EV in this memo is lease-inclusive and preferred-inclusive, stated per the DATA_DEFECTS requirement.

Restricted cash is excluded. AV returns cashAndCashEquivalentsAtCarryingValue = $862.2m; the balance sheet shows $667.2m of cash and a separate $192.2m of restricted cash. For an auction house restricted cash is consignor money against auction proceeds payable of $687.1m and is not available to the enterprise. AV overstates available cash by $192.2m / 28.8%.

Gross margin — the screen's largest single error

28.9pp. The screen paired a 16.7% gross margin with a 15.5% operating margin — 1.2pp of headroom for all operating expense. Had a terminal margin been set at the true trailing 18.3%, the model would have reported an m_EBIT,T > m_gross,T arithmetic violation on a business with 27pp of real headroom.

Operating margin — filed basis

AV Q1'26 operatingIncome = $222m against the 10-Q's $217.5m — 2.1% high. Applying that correction to AV's FY2025 sum of $854m gives roughly $837m filed, and TTM operating income of ~$868.5m on $4,736m = 18.3%. Screen: 15.5% (−2.8pp).

AV's ebit field was not used anywhere (documented as pretax + interest expense). AV's ebitda field was not used. D&A was taken from the cash-flow statement.

Revenue mix reconciliation

Inventory sales are 27.3% of Q1'26 revenue and rising 4.2pp a year. Inventory return — the gross profit on that line — was $30.2m on $336.9m, a 9.0% rate. So 27.3% of revenue contributes roughly 2.5% of gross profit.

FCF margin — and a tool defect

normalized_fcf.py --ticker RBA --revenue-growth 0.08 returns a steady-state FCF margin of 11.4% and a path mean of 10.9%, against a trailing 15.7%. It derives maintenance capex as 14.0% of revenue using D&A as the proxy.

That is wrong on this name and the reason is structural. RBA's D&A includes amortisation of the $2,388.0m of IAA intangibles. Actual capex is 5.6% of revenue. Using D&A as maintenance capex therefore charges acquisition amortisation a second time, as though it were cash capital spending. Corrected: operating cash margin 25.5% − capex 5.6% = 19.9% steady-state FCF margin.

The tool understates RBA's steady-state FCF margin by roughly 8.4pp, and the direction suppresses a PASS. I used 16.0% as the base — between the trailing 15.7% and the corrected 19.9% — and carried 11.0% / 16.0% / 19.9% through the whole surface. The verdict is FAIL at all three, so the tool defect does not change the answer here; it is recorded because it will change the answer on the next intangible-heavy acquirer.

normalized_fcf.py also crashed on its default invocation (no --revenue-growth) on all three tickers — see RBA_analysis.json tool_defects_found.