Delta Air Lines ($DAL) roared down the runway of Q3 earnings season today, with revenue up briskly, total costs up even more, and fuel costs specifically up so much you don’t even want to look at the line-item. 

Delta has been the first to file quarterly earnings for several years now, and today’s earnings release did not disappoint. It’s a fascinating glimpse into how a large, complicated business is navigating the economic pressures of today.

We can start with the big numbers. Total revenue for the quarter was $20.2 billion, up 21.1 percent from the year-ago period. Total costs, however, were $18.7 billion, an increase of 25 percent. All of that ultimately led to declines in operating income (down 13.7 percent), net income (down 45.6 percent), and EPS (down 47.2 percent to $1.15).

But nobody is here for the headline numbers, right? We love airlines for all the non-GAAP disclosures they also report, which give us much deeper insight into the firms’ overall performance.

Let’s start with fuel costs. They are, no pun intended, sky-high. Delta’s fuel costs were $4.35 billion for the quarter, an increase of — you may want an air sickness bag here — 69.3 percent. Delta’s average fuel cost per gallon was $3.80, compared to $2.25 one year ago. 

Figure 1, below, tracks average fuel costs for all six major U.S. airlines for the last three years. Right now we only have Delta’s numbers for Q3, although we’ll add the other five airlines as they file their earnings later this month. You can still see, however, that Delta’s average fuel costs are at least not rising as painfully as they were earlier this year.

Perhaps even more important for airline metrics are total revenue per available seat mile (TRASM) and cost per available seat mile (CASM). Those metrics help analysts understand how much money an airline is making per passenger.

Figure 2, below, compares TRASM and CASM at Delta since the start of 2024. The good news is that TRASM (in blue) has exceeded CASM (in red) in every quarter — but look at the sharp upward trend for both lines since the start of this year. That coincides with the war in Iran spiking fuel costs, and the broader but milder inflation we’ve seen everywhere else lately.

So far, Delta seems to have been able to pass along those higher costs to passengers. Total passenger revenue rose 15 percent in Q3 to $15.53 billion. Main cabin ticket revenue specifically rose 12 percent, to $6.8 billion. If you’re a Delta customer, those are numbers to think about next time you’re sitting at the gate waiting for your flight.

But will Delta be able to maintain that continued growth when so many costs and other pressures are bearing down on the flying consumer? 

Well, consider one more metric that Calcbench tracks for you: forward guidance. Delta reported that too, and said it now expects full-year 2026 EPS to land at $5.10 to $5.60, and full year free cash flow of $2.5 billion.

We stared at those numbers for a moment and then quietly clicked on the “See Previous Period” tab to see what Delta’s guidance was three months ago, at the end of Q2. Delta was predicting full-year EPS of $6.50 to $7.50, and full-year free cash flow of at least $3 billion. 

That’s quite a drop in altitude in just three months. One wonders how low those numbers might be come next quarter.

Today we return to the issue of corporate debt, and which firms might feel more pressure in 2027 as they try to refinance old debt at today’s higher interest rates. Our previous post last week gave a few suggestions for how analysts can use Calcbench to track such data; now let’s consider a few more examples based on actual corporate disclosures. 

We pulled the schedules of investments from more than 160 business development company filings for fiscal Q2 2026. We then looked for portfolio companies valued between 50% and 70% of cost: names that have lost at least 30% but still carry most of their value. Here's what stood out:

• Over 200 portfolio companies marked down at least 30%. Together they carry about $10.8 billion at cost, now valued at roughly 60 cents on the dollar, an unrealized loss of about $4.3 billion.
• The biggest holdings are software. Medallia, Cornerstone OnDemand, Kaseya, Symplr and Barracuda lead the list by cost, each held by eight or more BDCs.
• 84 of these companies are held by more than one BDC, and the BDCs often disagree. Curia Global's debt is marked at about 9% of cost by one lender and over 80% by others. Medallia's debt ranges from 39% to 51% depending on who holds it.
• The same company often appears under many names. Kaseya shows up under 14 different names across filers. Getting to a clean company-level view means reconciling all of them first.

The bottom line: the 50–70% band is where the hard valuation calls in private credit are made, and comparing marks across BDCs shows how much those calls vary. All of the underlying data is available at Calcbench, by holding and by BDC, straight from the filings. Have look at the table below for details.

Cruise and hospitality giant Carnival Corp. ($CCL) delivered a solid earnings report last week, which gives us another opportunity to talk about one of our favorite financial analysis topics here at Calcbench: non-GAAP disclosures!

Wall Street has been abuzz with concern lately about rising interest rates. The Federal Reserve is raising rates, the mortgage industry is raising rates, and soon enough the corporate debt markets will be looking at higher interest rates too.

The headline number

The four largest cloud providers now report nearly $2.4 trillion of remaining performance obligations (RPO): revenue under signed contracts that has not yet been recognized. The totals are easy to add up. They are much harder to compare, because each company describes the timing of that backlog in a different way. Before we get into the analysis, at the bottom are the tags and the datapoints that we used to get the data directly out of Calcbench.

In this post we translate each disclosure into a common unit: the average quarterly revenue implied by the near-term backlog. That number says more about the pipelines than the headline totals do. Meta, the fifth company in our hyperscaler peer group, does not report a comparable backlog.

Four companies, three timing methods

ASC 606 requires companies to disclose when they expect to recognize RPO, but it lets them choose how. The four hyperscalers use three different approaches, and the windows don't match.

The definitions differ too. Amazon counts only contracts with original terms over one year. Microsoft splits out a commercial figure. Alphabet splits out Google Cloud. Oracle reports one company-wide number that includes its support contracts.

Translating to a common unit

With some simple algebra, each disclosure can be turned into implied average quarterly revenue from backlog. For the three companies that give a percentage p over a window of X months:

For Amazon, which gives only a weighted-average remaining life L in years, we take the reciprocal of the life as an implied annual recognition rate:

Applied to the June 2026 quarter:

• Microsoft: $684B × 30% ÷ 4 ≈ $51B
• Alphabet: $519.5B × ~50% ÷ 8 ≈ $32B
• Amazon: $496B ÷ (4 × 6.4) ≈ $19B
• Oracle: about $19B, using the RPO and 12-month percentage in its fiscal 2026 10-K (quarter ended May 31, 2026)

What the chart shows

Implied quarterly revenue from backlog has grown at all four companies, but the ranking by growth is very different from the ranking by headline RPO.

Alphabet grew fastest. Its implied figure rose roughly fourfold in a year, from about $7B in Q2 2025 to about $32B in Q2 2026. Most of the jump came after Q3 2025. Google Cloud's actual revenue grew 82% over the same period, from $13.6B to $24.8B, so contracted revenue is growing far faster than revenue itself. Part of Alphabet's level reflects its 24-month window, discussed below.

Microsoft is still the largest. At about $51B a quarter, it is well ahead of the others, up roughly 40% from a year earlier. Its RPO includes a large base of Microsoft 365 and other commercial subscriptions, not only cloud infrastructure, so it starts from a higher floor.

Amazon's backlog is getting longer, not just bigger. Amazon's implied figure rose about 60% in a year, far less than its RPO. The reason is the weighted-average remaining life, which stretched from about four years to 6.4 years. The filing points to two large, long contracts: an expansion with OpenAI of $100 billion over 8 years in Q1 2026, and an expansion with Anthropic of more than $100 billion over 10 years in Q2 2026. More backlog is landing further out.

Oracle is accelerating. Its implied figure rose from about $11B to about $19B over the year. Oracle's August 2026 10-Q, filed after the period in the chart, pushes it higher: $664B × 13% ÷ 4 ≈ $21.6B. Against actual revenue of $19.3B in that quarter, coverage rises to about 112%.

How much of a quarter is already contracted

Comparing implied backlog revenue with actual revenue in the same quarter shows how much each business already has under contract. The spread is wide: from under half at AWS to more than 100% at Google Cloud.

Google Cloud's coverage above 100% is the clearest sign that these are averages, not forecasts of the next quarter. Its 24-month window reaches deep into a ramp as new capacity comes online. Oracle's contracted book now matches its entire current revenue run rate. AWS sits at the other end: much of its revenue comes from usage outside long-term commitments, and Amazon's RPO counts only contracts with original terms over one year.

Caveats

The implied figures are approximations. Each rests on an assumption that the filings do not let us test directly.

• Straight-line within the window. Dividing by X /3 spreads revenue evenly across the window. These backlogs are likely back-loaded, so early quarters would be smaller than the average. The longer the window, the bigger the effect, which inflates Alphabet relative to the 12-month companies.
• Amazon's reciprocal. RPO ÷ weighted-average life assumes contracts recognize straight-line and have similar lengths. When lengths vary, short contracts contribute more to near-term revenue than the average suggests. For example, $100 on a 1-year contract and $100 on a 9-year contract have a 5-year average life, implying $40 in year one; actual year-one revenue is $111. Back-loading pulls the other way, so the net bias can't be signed.
• Rounding. Percentages are disclosed as “approximately” or “just over.” On Oracle's $664B, one percentage point is worth about $1.7B a quarter. Amazon's life is given to one decimal; 0.1 years moves its implied quarter by about $0.5B.
• What RPO excludes. RPO leaves out cancellable contracts and pay-as-you-go usage. Oracle also elects the practical expedient to omit certain variable consideration. RPO is a contracted floor, not a revenue forecast.
• Microsoft's duration is a different measure. The reciprocal of Microsoft's 2.3-year duration is about 43%, well above the roughly 30% it says will be recognized in 12 months. Its duration appears to measure the average timing of recognition rather than remaining contract life, so it can't be used to check the Amazon method.

Getting the data from Calcbench

Every number in this post comes from XBRL-tagged filing data that Calcbench stores. RPO is tagged as RevenueRemainingPerformanceObligation . Where companies tag the timing, the percentages and periods are captured under RevenueRemainingPerformanceObligationPercentage and RevenueRemainingPerformanceObligationExpectedTimingOfSatisfactionPeriod1 , broken out by expected start date.

That means the analysis here isn't limited to five hyperscalers. The same RPO and timing data is available for every filer that reports it, across quarters, in the Calcbench platform and through the API. Narrative disclosures that aren't fully tagged, such as contract descriptions, are searchable in Calcbench's footnote and disclosure text.

Today we interrupt our usual attention to corporate earnings data to call out another type of disclosure that institutional investors might also find useful: the pay versus performance data one can use to evaluate CEO compensation.

Earnings quality — the question of how closely a company’s reported profit tracks the actual cash moving through the business — is one of the oldest and most durable ideas in fundamental analysis. When earnings run consistently ahead of cash flow, it’s usually a signal worth investigating further; when the two move together, it’s generally reassuring.

With standardized financial data available through the Calcbench API, the Multi-Company page, or the Excel Add-in, you can construct earnings-quality screens ranging from straightforward to fairly sophisticated — and apply them across the entire market at once!

To illustrate that range, we built two versions ourselves.

Two ways to measure the same idea.

The simple version: cash flow relative to net income. Divide operating cash flow by net income for a given period, and you have a quick, intuitive gut-check. A ratio near 1 is good and normal; a ratio well above or below 1 is worth a second look. 

This ratio is also easy to explain and easy to compute, but easy does not mean it's foolproof. Ratios like this get distorted by ordinary things, such as a company having an unusually small net income in a given quarter, a one-time gain or charge, or seasonal timing. Building this ratio smartly means building in some guardrails, not just dividing two numbers and reading the result uncritically.

The more rigorous version: an accrual ratio.1 Rather than comparing two income-statement-and-cash-flow-statement totals, this approach looks directly at the balance sheet — comparing how a company’s operating assets and liabilities have shifted from one period to the next relative to the size of the company. 

When configured correctly the two measures above point in opposite directions, so it’s worth keeping this straight before looking at the tables below.

• CFO/Net Income: bigger is usually better. A ratio near or above 1 means earnings are backed by real cash; a number well below 1 or even negative is a caution sign. (An extremely large ratio isn’t extra-good, however; it usually just means net income was unusually small that period.
• Accrual Ratio: smaller — more negative — is usually better. Numbers closer to zero or negative mean earnings are grounded in cash; higher, more positive numbers mean earnings are increasingly built on non-cash items such as growing receivables or inventory.

We ran both measures across the S&P 500 (companies in the financial and utility sectors excluded, since their balance sheets don’t lend themselves to this kind of comparison), comparing Q2 2026 to Q2 2025. Below is a look at the 25 companies at each end of the spectrum for both measures.

A Note on How to Read These

Accrual Ratio — bottom 25 / top 25

CFO / Net Income — bottom 25 / top 25

*Tied at the winsorization cap (1st/99th percentile) rather than a coincidental identical value.

¹ For readers who want the exact construction: the Accrual Ratio here follows the framework introduced by Sloan (1996), adapted for standardized Calcbench fields —
Accrual Ratio = (ΔCurrent Assets − ΔCash − ΔShort-Term Investments − ΔCurrent Liabilities + ΔShort-Term Debt − Depreciation) ÷ Average Total Assets
where each Δ is the year-over-year change (Q2 2026 vs. Q2 2025) and Depreciation is the current period's expense only, not a delta. Some versions of this measure also include a deferred-tax adjustment; we left it out of this run, since deferred-tax detail is disclosed far less consistently in quarterly filings than in annual reports. ↩

Since our last update, we've added 28 companies to the list, though most of that reflects catching up on research into filings from July and early August rather than new activity. Companies that actually reported in the last week or so include American Eagle Outfitters ($179M), Academy Sports and Outdoors ($83.7M, via a Participation Agreement that pre-sold a portion of its claim to a third-party buyer), Caleres ($55.6M), Destination XL Group ($4.6M), J.Jill ($18.4M), Lovesac ($20.3M), ChargePoint Holdings ($6.1M), and Petco ($6.8M). The total has grown from $19.3 billion across 105 companies to approximately $20.3 billion across 133 companies.

Companies with a quantified IEEPA tariff refund figure identified by Calcbench as of September 10, 2026, listed alphabetically. Total: approximately $20.3 billion. See the main analysis for methodology, sourcing, and the largest recoveries by dollar amount.

Total across 133 companies: approximately $20.3 billion.

Four new companies: Lululemon ($134.5M), PVH ($106.7M), Ross Stores ($253M), Victoria's Secret ($135M) — all filed within the last few days (Sept 1–4).

Companies with a quantified IEEPA tariff refund figure identified by Calcbench as of September 4, 2026, listed alphabetically. Total: approximately $19.3 billion. See the main analysis for methodology, sourcing, and the largest recoveries by dollar amount.

Total across 105 companies: approximately $19.3 billion.

Following Friday's post about Tariff refunds and all the gory details, we thought we would publish the full list of firms that we have found with tariff refunds to date.  

We may have missed some firms, but this is a solid list.  And we will continue to add to it.  Enjoy!

Full List: 101 Companies Disclosing IEEPA Tariff Refunds (Alphabetical)

Companies with a quantified IEEPA tariff refund figure identified by Calcbench as of August 28, 2026, listed alphabetically. Total: approximately $18.7 billion. See the main analysis for methodology, sourcing, and the ten largest recoveries by dollar amount.

Total across 101 companies: approximately $18.7 billion.

Figures reflect filings and disclosures identified as of August 28, 2026.

Following the invalidation of tariffs imposed under the International Emergency Economic Powers Act (IEEPA) earlier this year, public companies have begun reporting significant tariff refunds, expected recoveries, approved claims and refund receivables.

Calcbench has identified approximately $18.7 billion of gross tariff refunds across 101 public companies . The 11 largest company-specific amounts identified so far account for approximately $12.1 billion of the total, and they’re listed in the table below. (Normally we would do a top 10 list, but UPS and General Motors are tied at $500 million each, so we are treating the group as 11 rather than cutting them off at 10.)

The distribution is notably top-heavy. The 11 companies in the table above represent roughly 65 percent of the $18.7 billion we’ve identified across the whole 101 firms in our sample group; and the five largest account for 45 percent. 

That concentration rhymes with a broader pattern visible across many parts of today’s marketplace: a relatively small number of very large companies often account for a disproportionate share of the aggregate dollars.

* Apple's tariff-refund impact was disclosed in its fiscal Q3 2026 earnings press release, not in the 10-Q. The company reported quarterly revenue of $109.4 billion and gross margin of 50.1%, including a favorable impact of approximately two percentage points from tariff refunds, and diluted EPS of $2.02, including a $0.11 favorable impact from tariff refunds. Applying the gross-margin effect to quarterly revenue produces an estimated gross recovery of approximately $2.19 billion; we did not rely on the EPS impact and share count, which would produce a similar but not identical estimate.

** General Motors described a net $0.5 billion favorable adjustment primarily attributable to previously charged, refundable IEEPA tariffs. We use that amount as the best disclosed estimate rather than treating it as a separately reported cash receipt. GM separately disclosed an estimated $2.5–3.5 billion full-year EBIT-adjusted impact from the broader tariff environment; that figure is a forward-looking net cost estimate covering largely non-IEEPA tariffs and is unrelated to the refund pool measured here. Masco disclosed a similar net-only figure: approximately $95 million of net tariff benefit, principally in its Plumbing Products segment, for both the three and six months ended June 30, 2026 (indicating the full amount was recognized in the second quarter). As with GM, no separate gross or total-paid figure was disclosed alongside it. Dollar General took this a step further, disclosing only basis-point impacts directly (81 basis points to gross margin, 66 basis points to operating margin, both explicitly "after related reinvestments") and a $0.25 per-share EPS benefit, with no dollar figure anywhere in its release. We estimated approximately $55.5 million by backing into implied diluted shares from disclosed net income and EPS ($550.3 million ÷ $2.48 ≈ 221.9 million shares) and applying the per-share benefit — a derived, net-of-reinvestment, after-tax estimate, not a company-disclosed figure.

Why these refunds are difficult to find

Among the 11 largest recoveries, narrative-only disclosure accounts for 52 percent of the dollar amounts. Company-specific XBRL extension tags account for for 19 percent, and standard FASB-taxonomy tags for 5 percent. 

The remaining 24 percent — the single largest dollar amount in this group — was disclosed only verbally, on Walmart's earnings call, and does not yet appear in any SEC filing. A search of the standard FASB XBRL taxonomy alone would miss the large majority of the money sitting in just the largest recoveries. Companies have used at least four different disclosure approaches:

1. Standard FASB taxonomy tags and dimensions. Amazon and a number of other companies reported refund amounts using a standard tag such as RecoveryOfDirectCosts , combined with an IEEPA tariff-refund dimension.
2. Company-specific XBRL extension tags. Nike used InternationalEmergencyEconomicPowersActExpectedRecoveryOfTariffsPaid . FedEx used ProceedsFromInternationalEmergencyEconomicPowersActTariffs . UPS used IEEPATariffRefundClaimsSubmitted .
3. Narrative or earnings disclosures. Apple, Ford, Target and General Motors disclosed material amounts without a directly usable numeric XBRL fact identifying the refund.
4. Spoken commentary only, with no numeric disclosure in any filing. Walmart's CFO stated on the company's Q2 FY27 earnings call that Walmart "was eligible for approximately $2.9 billion of tariff refunds, amounting to approximately 0.5% of our annual U.S. net sales," and that "to date, we have received substantially all of these tariff refunds." At the time of this analysis, that figure does not appear in Walmart's earnings release, its 37-page investor presentation, or any SEC filing — only in the recording of the call itself.
5. Disclosed only inside a non-GAAP reconciliation table. Boston Scientific's $83 million IEEPA tariff refund appears exclusively as a line item in its Q2 2026 non-GAAP earnings reconciliation exhibit, treated as a one-time add-back excluded from adjusted results in the same manner as restructuring or litigation charges. Fortive followed the identical pattern at a smaller scale: its 10-Q risk-factor language confirmed that refund claims were being submitted and approved amounts recognized within cost of sales, but disclosed no figure there; the $4.5 million pretax amount appeared only in the separate earnings release's non-GAAP reconciliation. Motorola Solutions disclosed its $60 million pretax benefit ($0.25 per share) the same way, in a single reconciliation footnote within its earnings release. Urban Outfitters followed the same pattern with a $95.66 million figure, shown using the standard parenthetical-negative convention as an add-back subtracted from GAAP results to arrive at adjusted figures.
6. Acknowledged but never quantified. First Solar disclosed that receivables for outstanding IEEPA refund claims as of June 30, 2026 were included within "Other current assets," a balance sheet line that totaled $741.4 million as of June 30, 2026 (up from $643.1 million at December 31, 2025) — but the company never isolated what portion of that total, or its increase, related to tariff refunds specifically. We could not extract a comparable figure and have not included First Solar in the totals above.
7. Quantified, but recognized as zero. Deckers Outdoor disclosed a total gross amount of tariffs paid of approximately $120,000 — immaterial at this analysis's scale — and stated explicitly that it had recognized none of it as of the filing date, applying a gain contingency model under ASC 450 and noting that cost-sharing arrangements with its independent manufacturers may mean its own net exposure is smaller than the gross figure regardless. Not included in the totals above.

Caterpillar is an especially useful example, even though its $392 million recovery has since fallen just outside this top group. It tagged the recovery as ReceivableForRecoveryOfImportDutiesNet. The tag itself does not mention IEEPA or tariffs. The accompanying text explains that the amount represents expected IEEPA recoveries for claims submitted and accepted through the government's CAPE system. 

Caterpillar also disclosed that its total IEEPA tariff costs since the tariffs took effect were approximately $1 billion — meaning roughly $600 million of its own paid tariffs ae not yet reflected in any refund total, ours included, pending further CAPE claim submissions and approvals. 

Capri Holdings pushed the same problem further; it accounted for its $65 million IEEPA tariff refund receivable under ASC 410-30, Environmental Obligations — an accounting standard written for environmental remediation, applied here by analogy — which means its XBRL facts are tagged using environmental-remediation elements with an IEEPA dimension attached. A search for tariff-specific or import-duty-specific tags would never surface it.

Company-specific totals require review

The facts also cannot simply be added together to give you a clear sense of tariff refunds. A company may disclose a consolidated recovery, business-segment components, cash received to date, a remaining receivable and a related liability; adding them all together could lead you down a wildly inaccurate path.

Nike, for example, reported a consolidated recovery of $986 million, disclosed in Note 1, Summary of Significant Accounting Policies of its 10-K filed July 15, 2026. It also reported $965 million for North America, $21 million for Converse and $302 million of proceeds received. The $965 million and $21 million are components of the $986 million total, while the $302 million is a cash-receipt subset. Adding all four observations would substantially overstate Nike's recovery.

UPS provides another example. Its disclosure reported approximately $500 million of approved Phase 1 claims, consisting of $200 million received and $300 million recorded as a receivable. The correct company-level amount is $500 million, not $1 billion — but if you added all three items together, $1 billion is what you’d get. 

Resideo Technologies and ADI Global Distribution present a third, structurally different version of the same trap. Resideo recognized approximately $27 million of IEEPA tariff refunds in its consolidated Q2 2026 results. ADI Global Distribution — Resideo's sole operating segment prior to its August 2026 spin-off — separately disclosed approximately $20 million of IEEPA refunds in its own pre-spin-off combined financial statements, covering the same period during which ADI was still wholly owned by Resideo. 

The $20 million is almost certainly a subset of Resideo's $27 million, not an additional amount; we count only Resideo's consolidated figure. A spin-off's carve-out financial statements can create the same kind of overlap as a company's own internal segment reporting, just across two separate SEC filers rather than within one.

What the $18.7 billion represents

For this first stage of the analysis, Calcbench is measuring the gross tariff-refund pool. The total includes disclosed refunds, expected recoveries, approved claims, refund receivables and clearly identified eligible amounts.

Where a company discloses both a cash-received amount and a separately recognized receivable for the remainder, we use the cash-received figure only, not the combined total. 

LCI Industries, for example, disclosed $94.9 million actually received alongside a separate $24.4 million receivable deemed probable but not yet collected; we use $94.9 million. Xylem disclosed $4 million received and a separate $12 million receivable; we use $4 million. 

This is a deliberate choice, and not the only defensible one; a reasonable rationale exists for using the full recognized total instead, since a receivable deemed probable under U.S. GAAP is not a speculative number. 

We chose the more conservative figure for consistency across companies with very different disclosure structures. Where a company discloses only a single recognized amount with no cash-received breakout at all — Solventum and McCormick among them — that total is the only figure available and is used as-is. 

Zimmer Biomet is a clean example of this: it states it "received a portion" of its $77 million in previously paid tariffs during the second quarter but never quantifies that portion separately, so the full $77 million recognized under its loss recovery accounting model is the figure used here. Calcbench's underlying data captures both the cash-received and total-recognized figures wherever a company discloses the split, so a reader who prefers the alternative convention can reconstruct it directly from the same source data. 

One further caveat applies to a single company: International Flavors & Fragrances disclosed $18 million of tariff refunds received, but explicitly net of amounts expected to be remitted to customers — the only entry in this analysis where the figure itself is already net rather than gross, with neither the gross amount received nor the size of the customer obligation being netted out separately disclosed. We include it at $18 million since no other figure is available, but it is not strictly comparable on a gross basis to every other company in this table.

We have not yet reduced the amounts for obligations to return refunds to customers or share them with suppliers, and the size of that reduction varies enormously by company. Target and Amazon, by contrast, disclosed no customer-refund liability at all — both recognized their refunds as a direct reduction of Cost of Sales with no stated pass-through obligation, implying the full amount was retained. 

FedEx received approximately $800 million and separately reported a $749 million customer-refund liability, leaving roughly $51 million retained net — about 6 percent of the gross figure. Medline's $332 million recovery carried an $89 million estimated customer-repayment accrual, implying roughly $243 million retained net, or about 73 percent. Cardinal Health's $200 million receivable produced a disclosed net operating-earnings benefit of approximately $100 million for the quarter, or roughly 50 percent retained, once its own customer repayment obligation is factored in. 

UPS's situation is structurally different from the other three. UPS describes itself as acting as an intermediary for cross-border shipments, at times paying tariffs on behalf of customers and passing those costs through under its contract terms. It recorded the full $500 million of CBP-approved refunds within Other current liabilities rather than as a straightforward asset, and has disclosed no dollar figure for what portion, if any, it expects to retain once entries are reconciled and refunds are passed through. That makes UPS's $500 million harder to compare directly to FedEx's, Medline's, or Cardinal Health's gross figures, which represent tariffs those companies paid on their own imported goods. 

Walmart presents a fifth pattern: no customer-refund liability, but an explicit statement that the benefit is being funneled into price investments rather than kept as pure margin. Walmart's own earnings materials note that adjusted operating income growth in constant currency of approximately 17 percent included a 750-basis-point net benefit from tariff refunds received, after accounting for those price investments — implying a current-quarter net operating income effect of roughly $600 million, a much smaller number than the $2.9 billion cumulative refund figure disclosed on the call.

On Holding sits at the opposite extreme from Walmart. As of June 30, 2026, the Swiss footwear and apparel company had paid CHF 55.6 million in IEEPA tariffs and submitted CHF 52.7 million for refund processing, but recognized none of it — applying a "virtually certain" recognition threshold, stricter than the "probable" standard used by most companies in this analysis, and concluding that bar had not yet been met. By August 11, 2026, the company reported it had actually received approximately CHF 27.9 million, to be recognized in the third quarter.

Hubbell followed a similar pattern on a smaller scale: as of June 30, 2026, it had recognized nothing under its ASC 450-30 gain contingency model, but disclosed as a subsequent event that it had since received approximately $30 million in refunds. We count both On Holding's and Hubbell's actually-received amounts in the total, on the basis that real cash changing hands is a meaningful threshold even when formal accounting recognition lags behind it. Trimble adds a further variation: of its $17.8 million approved refund, it received $13.9 million in cash during the quarter, but structured its planned customer refunds as a reduction of revenue rather than a cost-side liability the way FedEx, Medline, and Cardinal Health did — the same underlying obligation to share the benefit with customers, booked on a different line of the income statement. Williams-Sonoma introduces a direction none of the others do: of its $197.8 million initial claim (all but $3.2 million collected as of August 2, 2026), $47.5 million was provisioned not for customers but to reimburse merchandise vendors who had previously extended the company tariff-related price concessions — benefit flowing upstream to suppliers rather than downstream to customers. Williams-Sonoma also directed $10.0 million of the refund toward a one-time discretionary 401(k) contribution for its employees, a voluntary use of the windfall rather than an accounting obligation of any kind. TJX went much further in the same direction: of its $331 million in gross refunds, it accrued $112 million of incremental incentive compensation and discretionary bonus expense for employees globally, netting to a $219 million pretax profit benefit — a formal accrued expense more than eleven times the size of Williams-Sonoma's, showing that directing refund proceeds to employees is a repeatable choice, not a single company's gesture. The Gap confirmed this is not a one-off: of its $512 million in gross tariff refunds, $95 million was committed to certain vendors as what the company called a "commitment of appreciation," netting to $417 million recognized in cost of goods sold. We include UPS's full $500 million and Walmart's full $2.9 billion in the $18.7 billion gross pool, consistent with how every other company's gross figure is treated, but readers should not assume either company retains anything close to its full disclosed amount.

****** On Holding reports in Swiss francs. We converted its approximately CHF 27.9 million received to approximately $34.3 million using an approximate CHF/USD exchange rate of 1.23 for the relevant period; we did not attempt to pin down the exact daily rate, since the currency effect at this size is immaterial to the totals in this analysis. On Holding's disclosure is, to our knowledge, the most conservative recognition posture of any company covered here: despite CHF 55.6 million paid and CHF 52.7 million already submitted for processing, none of it had been recognized in the financial statements as of June 30, 2026.

The immediate task is to identify and ringfence the gross refund population. Based on disclosures identified so far, that population totals approximately $18.7 billion across 101 public companies . The figure is a moving target in the most literal sense: Target and Lowe's each disclosed material IEEPA tariff refund impacts after this analysis was first compiled, adding a combined $1.08 billion and pushing Target directly into the top three; Walmart then disclosed, verbally, an amount larger than any other company's, adding a further $2.9 billion and taking the top spot outright. Target recognized $994 million as a reduction of Cost of Sales for the quarter and six months ended August 1, 2026, with no disclosed customer pass-through obligation. Lowe's disclosed a $0.11 per-share benefit to diluted and adjusted diluted EPS; applied to approximately 560 million diluted shares, that implies an after-tax impact of roughly $61.6 million, or approximately $81 million on a pre-tax basis grossed up at Lowe's approximate 24% effective tax rate.*** Caterpillar, separately, has disclosed roughly $600 million of its own paid IEEPA tariffs that are not yet reflected in any recovery figure, pending further CAPE claim submissions and approvals, and Cardinal Health noted that the majority of its refund requests fall under CBP's Phase 2 process, launched only in June 2026, with additional phases still to come. Taken together, these are concrete examples of a broader dynamic: CBP's refund process is still working through its early phases, and the total measured here reflects what has been recognized to date, not the full population of IEEPA tariffs eventually expected to be refunded. John Deere, reported separately, disclosed tariff recoveries of $110 million in its third fiscal quarter and $382 million for the first nine months of fiscal 2026 , embedded within its segment-level Production Costs reporting rather than called out as a standalone line — a reminder that some of the more generic "tariff recoveries" language in segment disclosures may be easy to overlook even when a specific dollar figure is present.*****

***** Deere's Q3 FY2026 disclosure does not use the term IEEPA and does not cite the Supreme Court ruling directly. We include it in this analysis on the basis of Deere's own fiscal 2025 Form 10-K, filed in late 2025, which stated: "On November 5, 2025, the United States Supreme Court heard oral arguments on tariffs imposed under the International Emergency Economic Powers Act (IEEPA). The court may provide tariff relief and the potential recovery of amounts previously paid. We are monitoring developments in this case and its impact on our future financial statements and business." Deere's subsequent "tariff recoveries" language, appearing in the first fiscal quarter after the ruling, is treated here as referring to that same IEEPA matter, though the Q3 disclosure itself does not make the connection explicit.

Advance Auto Parts, by contrast, disclosed a small but explicitly labeled figure: $26 million of second-quarter 2026 adjusted gross profit tied specifically to refunds of tariffs previously paid under IEEPA , with the company noting that the rest of its gross margin improvement was driven by unrelated merchandising initiatives. Unlike Deere's ambiguous phrasing, there is no inference required here — a useful reminder that disclosure quality varies as much among smaller amounts as it does among the largest ones.

*** Lowe's disclosed a $0.11 per-share benefit rather than a direct dollar figure. Applied to approximately 560 million diluted shares outstanding, this implies an after-tax net income impact of approximately $61.6 million. Because diluted EPS is an after-tax measure, we grossed this up using Lowe's approximate 24% effective tax rate (23.9% per its most recent 10-K) to estimate a pre-tax/gross figure of approximately $81 million, for rough comparability with the gross figures reported elsewhere in this analysis. Unlike Apple's gross-margin-based estimate, this figure is derived from an after-tax EPS impact and a tax-rate assumption and should be treated as a rougher approximation. At approximately $81 million, Lowe's does not rank among the largest recoveries identified so far.

**** Walmart's $2.9 billion figure comes from its Q2 FY27 earnings call (Bentonville, Ark., August 20, 2026), where the company's CFO stated Walmart "was eligible for approximately $2.9 billion of tariff refunds, amounting to approximately 0.5% of our annual U.S. net sales," and that "to date, we have received substantially all of these tariff refunds." At the time of this analysis, that figure does not appear in Walmart's earnings release, investor presentation, or any SEC filing; we cite the recording of the call directly, at approximately the 23:40–23:54 mark. Separately, Walmart's investor presentation states that adjusted operating income growth in constant currency of approximately 17% for the quarter included a 750-basis-point net benefit from tariff refunds received, after accounting for offsetting price investments — a figure that implies a current-quarter net operating income effect of roughly $600 million, which is not directly comparable to the $2.9 billion cumulative eligibility figure disclosed on the call. We use $2.9 billion as Walmart's entry in this analysis for consistency with how every other company's cumulative gross figure is treated, while noting that it is currently the only figure in this dataset sourced entirely from spoken commentary rather than a written filing or press release.

The broader lesson: structured data is essential to this analysis, but relying on a single taxonomy tag is not sufficient. A complete result requires standard XBRL facts, company extensions, dimensions and narrative disclosure text to be analyzed together — and, as FedEx shows, the gross number is only half the story until the pass-through obligations are netted out.

That’s it, folks — we now call time on our Q2 earnings updates! With more than 3,600 non-financial companies in our sample, we have reams of financial disclosures to review and ponder, so let’s get to it. 

Now that just about all S&P 500 firms have filed their Q2 earnings reports, let’s take a closer look at financial performance and economic trends hidden within those numbers. First up: capex spending.

Capital expenditures are always a useful disclosure to observe because it helps analysts understand broader economic trends. If capex is rising, that means firms are confident enough in their business prospects to make more investments for long-term growth. If it’s falling, that suggests businesses are less confident about future growth and would rather preserve cash.

The debate these days, of course, is whether capex spending for corporations as a whole is being distorted by a small number of tech firms spending zillions of dollars on AI data centers — and whether, if you strip those AI hyperscalers out of the analysis, capex spending is not as good as the overall number seems. So the Calcbench research team (read: intern trying to look busy) used our Multi-Company page to investigate.

Answer: yes. The AI hyperscalers are skewing the capex spending curve for everyone else.

Specifically, the S&P 500 reported $416.77 billion in net capital expenditures in Q2 2026. That’s a 32.5 percent increase from the $314.44 billion reported in Q2 2025. Sounds good so far, right? 

But $188.24 billion of that Q2-2026 number (that is, 45 percent of the $416.77 billion total) came from six data center kingpins:

• Amazon ($AMZN)

• Alphabet ($GOOG)

• Microsoft ($MSFT)

• Meta Platforms ($META)

• Oracle Corp. ($ORCL)

• Micron Technology ($MU)

If we exclude those six data center players, then capex spending didn’t jump 32.5 percent in Q2. It rose only 6.3 percent from the year-ago period, a far less impressive number. 

And who are these biggest spenders, you ask? We simply sorted our findings from largest to smallest, and answered the question instantly. See Figure 2, below.

Wow. No hyperscaler even cracked the Top 5 this quarter. And the first non-hyperscalers that do make an appearance are global behemoths such as Walmart ($WMT), Tesla ($TSLA) and AT&T ($T). Of course those folks would be near the top somewhere.

This also raises another question. If the AI hyperscalers are skewing the total picture on capex spending, are they doing the same on other financial statement items too? We’ve noted previously that Google, for example, booked a $97.8 billion one-time gain on the value of SpaceX ($SPCX) shares that it owns, which inflated overall net income growth considerably this quarter. 

So what other outlier events are exerting a larger gravitational effect on corporate performance, and to what extent? Stay tuned for more analysis in future posts!

Next week marks the official end of Q2 earnings season, when Nvidia ($NVDA) closes out the season with its earnings announcement on Aug. 26. Of course Calcbench will have that data too, but for now let’s see what the big picture for corporate earnings looks like with earnings from 3,500 non-financial companies already on file.

Following the invalidation of tariffs imposed under the International Emergency Economic Powers Act (IEEPA), public companies have begun reporting significant tariff refunds, expected recoveries, approved claims and refund receivables.

Calcbench has identified approximately $9.4 billion of gross tariff refunds across 55 public companies. The 10 largest company-level amounts identified so far account for approximately $7.8 billion of the total:

The distribution is decidedly top-heavy. Those 10 companies with the largest amounts represent approximately 83% of the $9.4 billion total, and the five largest account for approximately 63%. That concentration tracks with a broader pattern visible across many parts of today’s marketplace: a relatively small number of very large companies often account for a disproportionate share of the aggregate dollars.

* Apple reported that tariff refunds added approximately 2 percentage points to quarterly gross margin and $0.11 to diluted earnings per share. Applying the gross-margin effect to quarterly revenue of $109.417 billion produces an estimated gross recovery of approximately $2.19 billion.

** General Motors described a net $0.5 billion favorable adjustment primarily attributable to previously charged, refundable IEEPA tariffs. We use that amount as the best disclosed estimate rather than treating it as a separately reported cash receipt.

Why these refunds are difficult to find

A search of the standard FASB XBRL taxonomy finds a meaningful portion of the population, but it does not find everything. Companies have used at least three different disclosure approaches:

1. Standard FASB taxonomy tags and dimensions. Amazon and a number of other companies reported refund amounts using a standard tag such as RecoveryOfDirectCosts, combined with an IEEPA tariff-refund dimension.
2. Company-specific XBRL extension tags. Nike used InternationalEmergencyEconomicPowersActExpectedRecoveryOfTariffsPaid. FedEx used ProceedsFromInternationalEmergencyEconomicPowersActTariffs. UPS used IEEPATariffRefundClaimsSubmitted.
3. Narrative or earnings disclosures. Apple, Ford and General Motors disclosed material amounts without a directly usable numeric XBRL fact identifying the refund.

Caterpillar is an especially useful example. It tagged its $392 million recovery as ReceivableForRecoveryOfImportDutiesNet but the tag itself does not mention IEEPA or tariffs. The accompanying text explains that the amount represents expected IEEPA recoveries for claims submitted and accepted through the government's CAPE system.

Company-level totals require review

The facts also cannot simply be added together. A company may disclose a consolidated recovery, business-segment components, cash received to date, a remaining receivable and a related liability.

Nike, for example, reported a consolidated recovery of $986 million. It also reported $965 million for North America, $21 million for Converse and $302 million of proceeds received. The $965 million and $21 million are components of the $986 million total, while the $302 million is a cash-receipt subset. Adding all four observations would substantially overstate Nike's recovery.

UPS provides another example. It reported approximately $500 million of approved Phase 1 claims, consisting of approximately $200 million received and $300 million recorded as a receivable. The correct company-level amount is $500 million, not $1.0 billion.

What the $9.4 billion represents

For this first stage of the analysis, Calcbench is measuring the gross tariff-refund pool. The total includes disclosed refunds, expected recoveries, approved claims, refund receivables, and clearly identified eligible amounts.

We have not yet reduced the amounts for obligations to return refunds to customers or share them with suppliers. FedEx, for example, received approximately $800 million and separately reported a $749 million customer-refund liability. UPS also expects to pass approved refunds through to customers. Those obligations are important, but they answer a different question: How much of the gross refund will each company ultimately retain?

The immediate task is to identify and ringfence the gross refund population. Based on disclosures identified so far, that population totals approximately $9.4 billion across 55 public companies. The figure is likely to grow as more companies report and as additional extension-tagged and narrative disclosures are identified.

The broader lesson: structured data is essential to this analysis, but relying on a single taxonomy tag is not sufficient. A complete result requires standard XBRL facts, company extensions, dimensions, and narrative disclosure text to be analyzed altogether.

We are now nearing the end of Q2 earnings season, with data from more than 3,000 non-financial companies in our sample group. At this point the overall picture isn’t likely to change much, and it really hasn’t changed much since last week either: this has been a good quarter for Corporate America.

Everyone knows that the tech giants and AI hyperscalers are betting big on data centers. Calcbench has been taking a deep dive into Q2 disclosures of those companies, and today we offer a better sense of how big those bets are.

This was a busy week for corporate earnings, as more than 1,000 companies across a wide range of industries filed their Q2 earnings reports. We now have 2,050 non-financial companies in our famed Earnings Tracker, so let’s see where things stand.