RTX RTX

Reference table

1. What the business is, and how the disclosure actually works

RTX describes itself through three reportable segments, and both the 10-K and the 10-Q are explicit: "Our operations, for the periods presented herein, are classified into three principal segments: Collins, Pratt & Whitney, and Raytheon." Chief executive: the 10-K identifies the chief operating decision maker only as "our Chairman and Chief Executive Officer" and does not name him in the extracted text; the vendor profile block also omits a CEO name. We do not name one rather than guess. Approximately 180,000 employees as of 31 December 2025, including approximately 54,000 engineering professionals and approximately 32,000 represented by labour unions, located in 52 countries with 69% in the United States. Incorporated with CIK 0000101829, CUSIP 75513E101, listed on the New York Stock Exchange.

The three segments, in the 10-K's own words

• Collins Aerospace — aerospace and defence products plus aftermarket support for civil and military airframers, airlines, and regional, business, general-aviation, defence and commercial-space operators. Aircraft interiors (oxygen systems, galleys, lavatories and wastewater), avionics, power, actuation, wheels and brakes. Investing in operational capacity in the United States (including Puerto Rico), India, Mexico, Singapore and the Philippines.
• Pratt & Whitney — "among the world's leading suppliers of aircraft engines for commercial, military, business jet, and general aviation customers." Produces the PW1000G Geared Turbofan family, which "now powers more than 2,600 aircraft for over 90 operators across three aircraft platforms: Airbus A320neo family, Airbus A220, and Embraer E-Jets E2." The GTF aftermarket network reached 21 facilities worldwide, with PW1100G-JM shop-visit output up approximately 26% year over year in 2025. Also produces the F135, sole powerplant "to exclusively power the single-engine F-35 Lightning II" across all three variants and all partner and foreign-military-sales countries; in 2025 the F135 surpassed one million engine flight hours and P&W was awarded a $2.8 billion undefinitized contract action for Lot 18 and Lot 19 long-lead. The GTF Advantage received FAA and EASA certification in 2025.
• Raytheon — integrated air and missile defence, smart weapons, missiles, sensors and radars, interceptors, space systems and hypersonics. Named programmes in the 10-K include AMRAAM, StormBreaker, LRSO, Early Warning Radar, Next Generation Jammer, Tomahawk, SM-6, AIM-9X, SPY-6, Patriot, LTAMDS, NASAMS, Javelin, Excalibur, Stinger, AN/TPY-2 and SM-3, plus the Coyote counter-unmanned-aircraft system for which the 10-K reports "increased global demand."

The segment table does not sum to revenue, and the 10-K says why

This is the single most important disclosure-structure point in the file, it is a confirmed defect class, and RTX trips it exactly as an aerospace conglomerate should. The vendor's seg_prod block supplies three lines for 2025 and three for 2024 — and no eliminations line. They do not add up:

The 10-K explains the gap in one sentence, and the 10-Q repeats it verbatim: "Segment Total net sales and Operating profit (loss) include intercompany sales and profit, which are ultimately eliminated within Eliminations and other, which also includes certain smaller non-reportable segments." And separately: "The increase in eliminations and other net sales of $0.2 billion in 2025 compared to 2024 was primarily due to an increase in intersegment eliminations, principally driven by Collins." An increase of $0.2 billion on 2024's elimination is consistent with the $2.325B → $2.552B step the arithmetic implies. The segment lines are right; the vendor has dropped the contra line.

The proof that the vendor's own field CAN carry it: for fiscal 2023 the same seg_prod block supplies four lines — Raytheon $26.350B, Collins Aerospace $26.253B, Pratt and Whitney $18.296B and "Eliminations and Other" −$1.979B — which sum to exactly $68.920B, the reported 2023 revenue, to the dollar. The eliminations line is present in 2023, 2021 and 2020 and absent in 2024 and 2025. This is an omission, not a modelling choice.

Independent corroboration that the segment magnitudes themselves are correct. The 10-K states: "Net sales increased $8.9 billion organically in 2025, primarily due to higher organic sales of $4.8 billion at Pratt & Whitney, $2.6 billion at Collins, and $1.7 billion at Raytheon." The vendor's Pratt & Whitney line moves $28.066B → $32.916B = +$4.850B, matching the disclosed $4.8 billion organic almost exactly. Collins moves +$1.912B against +$2.6B organic less the disclosed divestitures of the actuation and flight-control business and Simmonds Precision Products; Raytheon moves +$1.330B against +$1.7B organic plus $0.1B "other" less the CIS divestiture. All three reconcile. We use the segment lines and add the eliminations line back ourselves.

Two label corrections. The vendor calls the third segment "Raytheon Intelligence & Space." That name is superseded — RTX consolidated Raytheon Intelligence & Space and Raytheon Missiles & Defense into a single "Raytheon" segment, and both the 2025 10-K and the June 2026 10-Q use only "Raytheon." The vendor also writes "Collins Aerospace Systems" where the filings say "Collins Aerospace" or simply "Collins." We use the filings' vocabulary throughout.

And the older history in seg_prod is a different company entirely. The 2018, 2017 and 2016 rows carry Otis and Carrier — the elevator and HVAC businesses spun off from United Technologies in 2020 — and the 2010 row carries Sikorsky, sold to Lockheed Martin in 2015. The 2015 through 2011 rows show a single line, "Commercial Aerospace," of $1.7–3.5 billion, which is not a coherent segment disclosure for a company that was then doing $50–65 billion. Nothing before fiscal 2020 in this block is usable and none of it appears in this dive.

Geography — one year is clean, one year is a hybrid of two different tables

The 10-K contains two geographic disclosures and they measure different things: "Geographic External Sales by Origin and Long-Lived Assets" (Note 20), where "Geographic external sales are attributed to the geographic regions based on their location of origin," and a separate "Segment sales disaggregated by geographic region based on customer location," which "uses end user customer location where known or practical to determine." The vendor's seg_geo block mixes them.

The tell is arithmetic. $74,778 on $88,603 of revenue is 84.4% — and the 2023 row's United States figure of $57,539 on $68,920 is 83.5%, an equally implausible customer-location share but a perfectly plausible by-origin share for a company that exports heavily and books foreign military sales through the U.S. government. So the 2023 row is the by-origin table, the 2024 row is the customer-location table, and the 2025 row is the by-origin domestic line pasted on top of four customer-location international lines. We use 2024 as supplied, derive the 2025 United States figure as a residual and label it as derived, and discard the 2023 row for comparison purposes because it is a different lens.

The honest read on geography: roughly 43–47% of RTX's revenue goes to customers outside the United States, and it is rising — international was 42.9% in 2024 and 46.6% in 2025 on the four comparable lines. Europe alone went from $15.975B to $18.799B, +17.7%, which is the NATO rearmament trade showing up in a revenue table.

Customer concentration — one named customer above 10%, and it is not a government

The 10-K's largest-customer disclosure is unusual for a defence prime because the named customer is a planemaker. "The largest contributor to our Commercial aerospace and other commercial sales is Airbus. Sales to Airbus primarily relate to Pratt & Whitney and Collins products, and prior to discounts and incentives were approximately 14%, 14%, and 17% of total net sales in 2025, 2024, and 2023, respectively." And within the engine segment specifically: "Pratt & Whitney's largest commercial customer by sales is Airbus, with sales, prior to discounts and incentives, of 29%, 31%, and 48% of total Pratt & Whitney segment sales in 2025, 2024, and 2023."

That is a genuine concentration and it points in an underappreciated direction: RTX's largest single commercial relationship is with the airframer whose fleet is grounded by RTX's own engine problem. The Powder Metal Matter concerns the PW1100 that powers the A320neo.

The 10-K also discloses U.S. government sales and international sales in tables that the current extraction stripped of numbers — we can see the footnotes ("Excludes foreign military sales through the U.S. government") but not the figures. We state that as a gap rather than estimating around it. What can be said from prose: RTX has "substantial U.S. government sales, which we conduct through all three of our business segments," Raytheon is "a prime contractor or major subcontractor on numerous programs" with the U.S. Navy, Army, Missile Defense Agency, Air Force and Space Force, and defence operations run through all three segments, not only Raytheon.

Competition, as the 10-K frames it

The 10-K does not publish a named competitor list. It says instead: "All of our businesses are subject to significant competition... Many of our competitors have substantial financial resources and significant technological capabilities. Further, some non-U.S. competitors receive government research and development assistance, marketing subsidies, and other assistance." On defence specifically it flags a structural shift worth quoting because it is the exact thesis one knowledge-base voice advances against this stock: "the competitive landscape in the defense industry continues to evolve with trends such as the continued increase in commercial competitors, new entrants with different technology approaches and business models, governmental bid evaluation processes requesting expanded intellectual property disclosures and rights sharing."

The vendor's peer set is unusually good for once — Boeing ($187.4B), GE Aerospace ($391.5B), Lockheed Martin ($136.0B), Northrop Grumman ($78.4B), General Dynamics ($104.4B), L3Harris ($53.1B), TransDigm ($71.3B), Leonardo DRS ($12.1B) — with only Caterpillar and Eaton as obvious misfits. No peer multiples are supplied in the file, so no peer-multiple comparison is drawn.

2. The single most important thing to understand about these numbers: the backlog, and the fact that it is prose

Every other number in this dive is a report of the past. The backlog is a contractual claim on the future, it is the largest such disclosure in this batch, and it exists nowhere in the vendor payload. It is a sentence in the MD&A of a 10-Q.

> "Total backlog was $289 billion and $268 billion as of June 30, 2026 and December 31, 2025, respectively. Total backlog includes commercial backlog of $170 billion and $161 billion as of June 30, 2026 and December 31, 2025, and defense backlog of $119 billion and $107 billion as of June 30, 2026 and December 31, 2025, respectively."

> — 10-Q filed 2026-07-23

> "Total backlog was $268 billion and $218 billion as of December 31, 2025 and 2024, respectively... Total backlog includes commercial backlog of $161 billion and $125 billion as of December 31, 2025 and 2024, respectively, and defense backlog of $107 billion and $93 billion as of December 31, 2025 and 2024, respectively."

> — 10-K filed 2026-02-06

Four things follow from that table, and together they are the entire bull case.

First, this is not a soft measure. The 10-K defines it precisely: "Backlog, which is equivalent to our remaining performance obligations (RPO) for our sales contracts, represents the aggregate dollar value of firm orders for which products have not been provided or service has not been performed and excludes unexercised contract options and potential orders under ordering-type contracts." Options and IDIQ ceilings are excluded. This is signed work.

Second, roughly a quarter of it lands within twelve months. "Of the total RPO as of December 31, 2025, we expect approximately 25% will be recognized as revenue over the next 12 months." Twenty-five percent of $268B is about $67 billion — roughly 70% of the entire 2026 consensus revenue line of $96.114B — already contracted before RTX wins another order.

Third, the defense side is accelerating far faster than the annual figures suggest. Defense bookings were "approximately $61 billion in both 2025 and 2024 and $51 billion in 2023." But the 10-Q reports $23 billion in the June 2026 quarter alone against $12 billion a year earlier, and $37 billion in the first half against $21 billion — +76%. Raytheon's own segment backlog went $63B (2024) → $75B (2025) → $86B (June 2026), and its June-quarter book is itemised program by program:

Fourth, and this is the part the market may be discounting incorrectly: consensus does not extrapolate any of it. The forward estimate block has revenue growing +8.5% (2026), +7.4% (2027), +7.2% (2028), +5.1% (2029), +3.5% (2030) — steady deceleration — against a backlog that grew 7.8% in six months. We return to this in section 6c, because it is where our view differs from the price.

The counterweight, stated plainly so this section is not a sales pitch. Backlog is a claim on revenue, not on profit. RTX's own filings spend more words on estimates-at-completion (EAC) adjustments than on almost anything else: "we have both favorable and unfavorable EAC adjustments in the ordinary course," and on fixed-price contracts "we receive a fixed price irrespective of the actual costs we incur, and we therefore carry the burden of any cost overruns." The 2024 Raytheon Contract Termination — "the termination of a fixed price development contract with a foreign customer" — cost $0.5 billion in cost of sales and $0.6 billion of segment operating profit. A large backlog converted at the wrong margin is a large problem, and RTX has demonstrated that at scale twice in three years.

3. Growth and margin trajectory — and the pension income sitting under it

Annual, from the vendor income series, with margins computed by us:

2023 is the Powder Metal year and 2024 the legal-settlement year — neither is a clean read, and the filings quantify both. More on that in section 8.

Quarterly, cross-checked against earn_cal:

Five consecutive beats, averaging +13.2% on adjusted EPS, and revenue beat by 7.9% in the June quarter ($24.708B against $22.892B expected). Operating margin has expanded from 5.17% in 2023 to 11.38% in the June quarter. Gross margin has gone from 17.54% to 20.8%. This is a genuine, sustained, multi-year operating recovery and it is corroborated segment by segment in the MD&A — Collins organic sales +$1.0B in the quarter on commercial OEM, aftermarket and defence; Pratt & Whitney +$1.3B of which $0.9B was commercial aftermarket and $0.5B military F135; Raytheon +$1.3B on Patriot, Standard Missile and AMRAAM.

The pension line, which is where roughly an eighth of the earnings comes from

RTX carries one of the largest corporate defined-benefit plans in the United States, and GAAP requires the non-service components of it to be reported as a separate line BELOW operating profit. The 10-K is explicit: "Net periodic benefit (income) expense is classified between operating and non-operating, whereby only the service cost component is included in operating profit and the remaining components are included in Non-service pension income."

The filing's numeric table was stripped by the current extraction, so we derived the line ourselves from the income statement and then verified the derivation against the 10-K's prose. Non-service pension income = total other income/expense, net + interest expense − interest income:

The derivation reproduces both disclosed year-on-year changes and isolates the disclosed Q4 2025 settlement charge to within rounding. We are confident in it. The settlement itself is disclosed at Note 10: in December 2025 RTX "completed a buy-out conversion of a group annuity contract previously purchased by the Plan from The Prudential Insurance Company of America (Prudential) transferring approximately $2.3 billion of gross pension obligations," covering "approximately 60,000 Plan retirees and beneficiaries, which represents approximately one-third of retirees and beneficiaries in the Plan," and generating "a one-time, non-cash pension settlement charge of $0.3 billion."

What it does to the earnings. Trailing twelve-month pre-tax income is $9,997M, of which $1,118M — 11.2% — is non-service pension income. At the trailing effective tax rate of 18.73% that is $908.6M after tax, or $0.666 per diluted share on 1,364.7M shares.

> Trailing GAAP diluted EPS $5.68 · less $0.67 of non-operating pension income · = CLEAN operating EPS of $5.01

> P/E on reported trailing GAAP EPS: 38.4x. P/E on clean operating EPS: 43.5x.

This is not an accounting scandal and we are not calling it one. Non-service pension income is real, recurring, and disclosed exactly where the rules require. But it is not operating income, it is not cash, it is a function of discount rates and asset returns rather than of selling engines, and it is 11.7% of the earnings a buyer at $217.93 is paying 38x for. Note also the direction of travel: $1,828M (2023) → $1,524M (2024) → $1,170M (2025), and the December 2025 annuity buy-out removed $2.3B of obligations and the matching assets from the plan. The line has fallen 36% in two years, and every dollar it falls is a dollar of EPS headwind that has nothing to do with the backlog.

One further disclosure most readers will not know exists. Inside operating profit, RTX also runs a FAS/CAS operating adjustment: "the difference between the service cost component of our pension and PRB expense under the FAS requirements of U.S. GAAP and our pension and PRB expense under U.S. government Cost Accounting Standards (CAS), primarily related to our Raytheon segment." The 10-K reports the adjustment changed by $0.1 billion in 2025 and $0.3 billion in 2024 but the numeric table is stripped. So a second pension-driven swing factor sits inside operating profit and we cannot size it from this archive. We flag it and do not estimate.

Research and development, and where the cost pressure is

R&D was $2.807B in 2025, 3.17% of revenue (trailing 3.02%). SG&A was $6.095B, 6.88% of revenue (trailing 6.64%). Both modest for the sector — this is a company whose development is largely customer-funded through cost-reimbursable contracts.

The disclosed cost headwinds are tariffs and the supply chain, and the 10-K names tariffs as a direct margin drag in two of three segments. At Collins: operating profit rose $0.5 billion "partially offset by the impact of tariffs and unfavorable commercial OEM mix." At Pratt & Whitney: aftermarket volume "more than offset the impacts of unfavorable aftermarket mix and higher tariffs and production costs." And in the June 2026 half-year: Collins commercial aerospace profit rose "partially offset by the impact of higher tariffs." On the supply chain, the 10-K is blunt about physical dependency: "we are largely dependent upon foreign sources for certain raw materials, such as cobalt, tantalum, chromium, rhenium, nickel, and titanium, and we rely on foreign suppliers as single-source suppliers of some components."

4. Balance sheet and the enterprise-value rebuild

The first thing to say is that bal_a[0] in this file is dated 31 December 2025 and the archive contains a 10-Q for the quarter ended 30 June 2026 filed 2026-07-23. The annual balance sheet is seven months stale. This is confirmed defect class 1 and it is the reason we did not simply rebuild net debt from bal_a as the standing instruction contemplates — doing so would have produced a worse answer, not a better one.

What the 10-Q actually verifies at 30 June 2026

The 10-Q's balance sheet tables were stripped by the current extraction. Exactly one balance-sheet figure survives in prose, and it is the important one:

> "At June 30, 2026, we had cash and cash equivalents of $8.3 billion, of which approximately 27% was held by RTX's foreign subsidiaries."

Plus, on the debt side:

> "As of June 30, 2026, we had a revolving credit agreement with various banks permitting aggregate borrowings of up to $5.0 billion, which expires in August 2028. As of June 30, 2026, there were no borrowings outstanding under this agreement."

> "At June 30, 2026 and December 31, 2025, we had no commercial paper borrowings outstanding. There were no new borrowings or repayments of commercial paper with maturities greater than 90 days during the six months ended June 30, 2026 and 2025."

> "We made the following repayments of long-term debt during the six months ended June 30, 2026:" — table stripped, but the heading says repayments only, with no issuance heading anywhere in the note.

> "The average maturity of our long-term debt as of June 30, 2026 is approximately 12 years."

> "The $1.5 billion change in cash flows used in financing activities in the six months ended June 30, 2026... was primarily driven by the absence of proceeds from commercial paper borrowings of $1.4 billion in the six months ended June 30, 2025."

> "The decrease in Interest expense for the quarter and six months ended June 30, 2026... was primarily driven by long-term debt repayments."

Every one of those points in the same direction: debt FELL in the first half of 2026. Interest expense confirms it independently — $502M, $480M, $483M, $370M, $406M, $417M across the last six quarters, down 17% from the Q1 2025 level.

The 8-K check for a post-statement financing — run, and it came back empty

This is exactly the test that caught LLY's 23% debt understatement, and RTX passes it. The archive contains three 8-Ks and none is a financing:

• 8-K 2026-04-21 — Item 2.02, cover page announcing first-quarter 2026 results. "The press release issued April 21, 2026 is furnished herewith as Exhibit No. 99." Exhibit 99 is NOT in the extraction. The document contains no financial figures.
• 8-K 2026-05-04 (event date 2026-04-30) — Item 5.07, annual meeting voting results. Discloses that "As of March 3, 2026, the record date for the meeting, 1,345,974,220 shares of Company common stock were issued and outstanding." This is a useful independent share-count checkpoint.
• 8-K 2026-07-23 — Item 2.02, cover page announcing second-quarter 2026 results. "The press release issued July 23, 2026 is furnished herewith as Exhibit No. 99." Exhibit 99 is ABSENT from the extraction — the file is 2,667 bytes of cover page and contains not a single financial figure. (The parent brief asked specifically whether Exhibit 99.1 was present or absent on this filing: it is labelled Exhibit 99, not 99.1, and it is ABSENT.)

There is no bond deal, no shelf takedown, no term loan and no equity issuance in the archive post-dating the June 2026 balance sheet. The 10-Q does note "an existing universal shelf registration statement, which we filed with the Securities and Exchange Commission on September 18, 2025, for an indeterminate amount of debt and equity securities for future issuance" — capacity, not a draw.

The rebuild

We cannot read total debt at 30 June 2026 off a filing, because the long-term-debt table was stripped. We can derive it from the vendor's trailing ratio block and then check it three independent ways. Solving the trailing per-share block on the share count implied by trailing revenue (1,349.06M):

• revenuePerShareTTM 69.3003 × 1,349.06M = $93,500M — matches the sum of the last four quarterly revenues exactly.
• cashPerShareTTM 6.1555 × 1,349.06M = $8,304M — matches the 10-Q's filing-verified "$8.3 billion" of cash.
• debtToEquityRatioTTM 0.58538 × common equity of $66,373M = $38,853M of total debt. debtToCapitalRatioTTM of 0.36924 independently gives the same figure.
• Cross-check: enterpriseValueTTM $324,268M − market cap $293,717M = $30,551M of implied net debt; and $38,853M − $8,304M = $30,549M. Agreement to $2 million.

> Market cap $293.717B (1,347,758,144 shares × $217.93 — the exact 10-Q cover count)

> + total debt ~$38.853B (derived; includes $1.602B of finance leases on the December basis)

> − cash and equivalents $8.304B (filing-verified at $8.3B)

> − short-term investments $0 (RTX reports none — see below)

> = rebuilt enterprise value $324.27B

Correction 1 — the stale balance sheet, and it runs the OTHER way. A naive rebuild from bal_a[0] gives net debt of $32.071B (December). The June figure is $30.55B. The vendor's bal_a overstates current net debt by $1.52 billion, or 5.0%, because it is a December balance sheet and RTX has been repaying debt. We use $30.55B.

Correction 2 — short-term investments are genuinely zero, and this is stated because the standing test assumes otherwise. shortTermInvestments reads $0 in all three years of bal_a. That is correct, not a defect. The 10-Q's liquidity section describes only "cash and cash equivalents of $8.3 billion" — there is no "cash, cash equivalents and marketable securities" formulation of the kind that made KLA's and Berkshire's net debt wrong. The $2.132B of longTermInvestments on the December sheet is customer financing and equity-method interests, not liquidity, and the "marketable securities held in trusts" mentioned at Note 10 are nonqualified-plan assets sitting in Other assets and dedicated to benefit obligations. Tested, and there is nothing to add back.

Correction 3 — the vendor's trailing EBITDA includes the pension income. evToEBITDATTM of 20.1747 implies trailing EBITDA of $16,073M, which reconciles exactly to pre-tax income $9,997M + interest expense $1,676M + D&A $4,400M. That construction sweeps in the $1,118M of non-service pension income. Clean operating EBITDA is operating income $10,485M + D&A $4,400M = $14,885M.

What the balance sheet actually says. RTX is investment grade and improving: "In May 2026, our S&P Global rating was affirmed and our outlook was revised from BBB+/stable to BBB+/positive. In February 2026, our Moody's Investors Service outlook improved from Baa1/stable to Baa1/positive." Average debt maturity approximately twelve years. A $5.0B revolver untouched to August 2028. Net leverage of 2.05x on clean EBITDA is comfortable but is not a fortress — it is roughly ten times KLA's and it sits on a company with:

• $85.188B of goodwill and intangibles — 49.8% of total assets (intangiblesToTotalAssetsTTM 48.27%). The 10-K's own risk factor names it: "A significant portion of our assets consists of goodwill and other intangible assets." The 1 October 2025 annual impairment test was passed at step 0 — a qualitative assessment only, with no quantitative testing performed on any reporting unit.
• Tangible book value of NEGATIVE $11.73 per share and tangibleAssetValueTTM of −$15.827B. Shareholders' equity of $65.245B is entirely accounted for by acquired intangibles.
• Return on invested capital of 7.33% on invested capital of $103.330B. Return on equity 11.80%; return on assets 4.45%. For a company carrying $53B of goodwill from the Rockwell Collins and Raytheon transactions, a 7.3% ROIC is the honest measure of what those deals earn, and it is not obviously above the cost of the capital.
• A working-capital position of $667M on $93.5B of revenue — currentRatioTTM 1.011, quickRatioTTM 0.783. Receivables of $31.793B are 128.5 days; inventory $13.364B is 70.6 days; the cash conversion cycle is 115.8 days. Long-cycle manufacturing looks like this, but note the receivables number is flattered by factoring (section 5).

5. Cash flow and capital returns

Free cash flow has gone from $4.5B to a trailing $12.0B in eighteen months, and roughly an eighth of the improvement is a financing decision, not an operating one. The 10-Q says so directly: "The Company enters into various factoring agreements with third-parties to sell certain of its receivables, primarily related to customer facilitated programs... Factoring activity resulted in a $1.5 billion increase in cash provided by operating activities during the six months ended June 30, 2026 compared to the six months ended June 30, 2025." We flag that rather than celebrating a $12B free-cash-flow print at face value. The remainder is genuine: "higher net income after adjustments... driven by our segment performance and favorable changes in net working capital."

Capex is $2.799B trailing, 3.0% of revenue — light for a manufacturer, and capexToDepreciationTTM of 0.636 says RTX is spending well under its depreciation charge. On a $4.4B annual D&A run-rate that is a real question about reinvestment adequacy in a business ramping five munitions lines, though a large share of that D&A is acquisition-related intangible amortisation rather than replaceable plant.

incomeQualityTTM of 1.824 (operating cash flow ÷ net income) is high, and part of that is the factoring above and part is the $4.4B of non-cash D&A running through a business whose intangible amortisation alone costs $0.27 per diluted share per quarter ($0.54 for the half, per the 10-Q).

The dividend, the buyback that stopped, and the Executive Order

Dividend — and the vendor's label is wrong in the same way it was on KLA. lastDividend reads 2.77 and dividendPerShareTTM reads 2.77, giving a yield of 1.271%. That $2.77 is the trailing-twelve-month total, not the last declared dividend. The filings give the actual sequence: the 10-K records "On February 6, 2026, the Board of Directors declared a dividend of $0.68 per share payable March 19, 2026"; the 10-Q records "On April 30, 2026, the Board of Directors declared a dividend of $0.73 per share... On June 26, 2026, the Board of Directors declared a dividend of $0.73 per share payable September 3, 2026." Three quarters at $0.68 plus one at $0.73 is $2.77 — the vendor's figure, correctly computed but mislabelled. The forward run-rate is $0.73 per quarter, $2.92 annualised — a 1.340% forward yield, not 1.271%. Payout ratio on trailing GAAP EPS is 48.1%; on clean operating EPS it is 58.3%.

Buyback — this is the most under-discussed fact in the file.

RTX bought back $50 million of stock in the whole of 2025 — 0.017% of the current market capitalisation — and the diluted share count has RISEN every period since 2024. The cover-page counts confirm it independently: 1,342,287,676 shares at 31 December 2025 → 1,345,974,220 at the 3 March 2026 record date → 1,347,758,144 at 30 June 2026. That is +0.41% in six months, roughly +0.8% annualised dilution against $1.092B of stock compensation. And the runway is nearly gone: "At June 30, 2026, management had remaining authority to repurchase approximately $0.6 billion of our common stock under the October 21, 2023 share repurchase program." $0.6B is 0.20% of market cap.

Total shareholder yield is therefore about +0.5%: a 1.34% forward dividend, essentially no buyback, less roughly 0.8% of dilution. This is not an income story and it must not be sold as one.

And there is a disclosed political risk to even that. From the 10-K's own capital-allocation discussion:

> "In addition, pursuant to a January 7, 2026 Executive Order, the Secretary of War could seek to limit our ability to pay cash dividends or make share repurchases if the Secretary of War determines that we have underperformed or lacked sufficient prioritization of, investment in or production speed in carrying out / performing under our U.S. government contracts."

And in the 10-Q's risk discussion: "A subsequent executive order was issued that may limit corporate distributions, share repurchases, and executive compensation incentives during periods of defense contractor underperformance... a limitation on our ability to issue distributions or engage in share repurchases related to the defense contractor performance executive order could adversely affect the market price of our common stock."

We have not seen this risk factor at any other name in this batch. It is company-and-sector specific, it is in the company's own words, and it attaches directly to the only component of shareholder return RTX is currently delivering.

6. Valuation — priced in or room?

At $217.93 (market cap $293.717B, 1,347,758,144 shares, rebuilt enterprise value $324.27B):

Estimate coverage is good through 2028 — 15 to 18 analysts on revenue and 11 to 15 on EPS. The 2029 and 2030 EPS rows rest on FOUR analysts each and are excluded from every conclusion in this dive.

Per the data contract, est.ebitdaAvg and est.ebitAvg are not used for forward valuation, and we tested them rather than assuming. No arithmetically impossible row exists in this file: every ebitdaAvg is positive, ebitAvg is below ebitdaAvg in all eight years, and neither equals minus revenue. But both are consistently low against actuals. The 2025 ebitAvg of $8.887B is 4.4% below actual operating income of $9.300B and 15.9% below the vendor's own computed EBIT of $10.568B; the 2025 ebitdaAvg of $14.049B is 6.0% below the actual $14.946B. Wrong in a consistent direction, not nonsensical. All forward valuation here runs on epsAvg and revenueAvg.

Basis note, and it matters more here than at most names. The est and earn_cal blocks are non-GAAP adjusted EPS; inc_a and inc_q are GAAP. RTX's adjustments are chiefly acquisition accounting adjustments — the 10-Q quantifies them at "$0.4 billion, net of tax, which had an unfavorable impact on diluted earnings per share (EPS) of $0.27" per quarter, $0.54 for the half — plus restructuring. The gap is large: 2025 consensus adjusted EPS of $6.211 against GAAP diluted EPS of $4.96 is a 25.2% premium. Every forward multiple above uses consensus non-GAAP; the trailing multiples are shown on GAAP and on our clean operating basis. Note also that consensus adjusted EPS INCLUDES the non-service pension income — the street does not strip it, and neither do we when computing forward multiples, precisely so that the comparison is like-for-like.

6a. What today's price assumes (the inversion)

At $217.93 — 30.1x the 2026 consensus, 27.6x 2027 and 24.9x 2028 — the price embeds the following falsifiable claims, each with a number and a date:

• Revenue reaches $96.114B in 2026 (+8.5%), $103.253B in 2027 (+7.4%) and $110.637B in 2028 (+7.2%). (Consensus-derived; 16, 17 and 18 analysts.) The near-term check is 2026-10-20, where consensus wants $23.9997B of September-quarter revenue — only +6.8% year on year against the +14.5% just delivered. First-half 2026 revenue of $46.784B was +11.7% on the year-ago half; the full-year consensus implies second-half growth of only +5.6%. That is a sharp implied deceleration against a backlog that grew 7.8% in six months.
• Adjusted EPS reaches $7.234 in 2026 and $8.747 by 2028 — roughly 10% annual compounding after 2026. (Consensus-derived; 15 and 11 analysts.) First-half 2026 adjusted EPS was already $3.67 ($1.78 + $1.89), so the second half needs only $3.56 to hit the number — less than the first half, in a year where the second half is seasonally the bigger revenue period. On 2025's half-year shape the implied full year is nearer $7.61.
• Operating margin holds above roughly 11% and keeps grinding higher. (Our number, from the 5.17% → 8.10% → 10.50% → 11.38% path.) This is the assumption most exposed to something outside RTX's control: the 10-K names tariffs as an explicit drag on both Collins and Pratt & Whitney 2025 operating profit, RTX has paid approximately $0.5 billion of IEEPA tariffs whose refund is now under U.S. government appeal, and the 10-Q warns that actual tariff costs "could significantly affect the estimates inherent in our financial statements, including those used in our estimates-at-completion."
• Non-service pension income of roughly $1.1 billion a year, 11.7% of trailing EPS, does not deteriorate further. (Our number, derived and filing-corroborated.) This is the most fragile assumption in the price. The line has already fallen from $1,828M (2023) to $1,524M (2024) to $1,170M (2025), and the December 2025 Prudential buy-out permanently removed $2.3 billion of gross obligations and the matching plan assets. It is a non-operating, non-cash credit that consensus adjusted EPS silently includes, and it is decaying.
• The market keeps paying roughly 25x or better on the forward year at every date between now and 2028. (Our number.) At 20x the 2028 consensus the stock is $175; at 30x it is $262. The multiple, not the backlog, is the widest source of variance in the outcome — and RTX has arrived at this multiple in three months, having lagged the S&P over six.

6b. The return bridge (why the multiple moves)

Expected return over the next twelve months decomposes as: adjusted EPS growth (+10.8%, from 2027E $7.892 to 2028E $8.747) + multiple drift (mild COMPRESSION, from 27.6x to 26x on the respective forward year, −5.8%) + shareholder yield (+1.34% dividend, less roughly 0.8% dilution, net ≈ +0.5%) ≈ +4% to +6%.

Our base case assumes modest multiple COMPRESSION, and it is not a rolldown artefact. Today the market pays 27.6x on the next calendar year's estimate. If in twelve months it pays 26x on the then-next year's estimate, that is a genuine de-rating of about 6%, not the mechanical effect of the calendar advancing. We assume it for three reasons, all arithmetic rather than atmospheric. First, consensus growth itself decelerates on its own path: revenue +8.5%, +7.4%, +7.2%, +5.1%, +3.5% across 2026–2030. Second, the quality of the earnings is deteriorating at the margin — a rising share of a shrinking pension credit is arithmetically impossible, so as the $1.1B line decays it becomes a drag on reported growth that operations must offset. Third, RTX re-rated 26% in three months to arrive here, and multiples that arrive that fast without a change in the growth algorithm tend not to be sticky.

Almost all of the expected return is earnings growth, and the multiple is working against it. If the multiple simply HELD at 27.6x on 2028E the price would be $241 (+10.8%); if it compressed to 22x the price would be $192 (−11.7%) — which is, not coincidentally, the 50-day moving average. The entire spread between a good and a mediocre outcome sits in a multiple we do not control and have no edge on.

The bull case at $273 requires modest expansion — 29x applied to the 2028 consensus high of $9.407 — and is therefore the least conservative of our three anchors, deliberately, because the backlog gives a mechanism for the upgrade. The bear case at $163 requires no operational disaster at all: 20x the 2028 consensus low of $8.171. A 20x multiple is not a crisis multiple for an industrial growing 7%; it is a normal one. That asymmetry — a bear case that needs only a normal multiple and a bull case that needs an above-average one — is the core of the Watch.

6c. Variant perception (where we differ, what would surprise)

• We differ from consensus on the near-term numbers, in the company's FAVOUR, and it is the only place we think we have an edge. Consensus has second-half 2026 revenue growing +5.6% year on year after +11.7% in the first half, and full-year adjusted EPS of $7.234 when the first half alone delivered $3.67. RTX has beaten adjusted EPS in five consecutive quarters by an average of 13.2% and beat June-quarter revenue by 7.9%. Applying 2025's half-year earnings shape to 2026's first half implies roughly $7.61, about 5% above consensus — which would make spot 28.6x rather than 30.1x. Watchable number: the 2026-10-20 print against $23.9997B of revenue and $1.75 of adjusted EPS. A sixth consecutive double-digit beat is the base case on the evidence.
• We differ from the street on what the earnings are made of, and this cuts the other way. No sell-side number we can see in this file strips the $1,118M of non-service pension income — 11.7% of trailing EPS, non-operating, non-cash, and down 36% in two years. Clean trailing EPS is $5.01 and the honest trailing multiple is 43.5x, not 38.4x. Combined with ROIC of 7.33%, negative tangible book of $11.73 per share and $85.2B of goodwill and intangibles never quantitatively tested (the 2025 impairment review stopped at step 0), we think the quality of this earnings stream is lower than a 30x forward multiple implies. Watchable number: the non-service pension income line in the 2026 10-K, and whether the FAS/CAS operating adjustment inside operating profit moves against the company.
• We differ on the capital-return story, and we think the market has not priced the Executive Order at all. RTX bought back $50 million of stock in 2025 with $0.6 billion of authority left, the share count is rising 0.8% a year, and the 10-K discloses that a January 7, 2026 Executive Order allows the Secretary of War to seek to limit dividends and repurchases on a performance determination. A defence prime whose distributions are subject to a political veto is a different instrument from one whose distributions are not, and total shareholder yield of roughly +0.5% is not compensating for that. Watchable event: any Department of War determination naming a contractor, or a restart of the buyback, which would signal the company does not consider the order binding in practice.
• Positive surprise that would force a re-rate: a September-quarter print materially above $23.9997B with operating margin above 11.5%, which would simultaneously kill the second-half deceleration in consensus and validate the margin path; approval of the IEEPA tariff refund against the approximately $0.5 billion paid, which RTX has explicitly not booked as a receivable; passage of the $67 billion supplemental with $21 billion for munitions, or a named Golden Dome award against the $24.4 billion appropriated; or a formal statement that PW1100 aircraft-on-ground levels have normalised ahead of the "through 2026" guidance.
• Negative surprise that would break the thesis: any disclosure that other engine models are materially affected by the powder-metal condition — the 10-K warns explicitly that "other engine models within Pratt & Whitney's fleet contain parts manufactured with affected powder metal"; a material adverse EAC adjustment on a fixed-price development contract of the kind that cost $0.6 billion of segment operating profit in 2024; a Secretary of War determination limiting distributions; or simply a de-rating toward 22x, which needs no news at all and takes the stock to $192.

Synthos fair values

All three anchors are multiples of the 2028 consensus adjusted-EPS distribution (mean $8.747, low $8.171, high $9.407, 11 analysts), cross-checked against 2027E of $7.892 (15 analysts).

• Bear ~$163 — 20x the 2028 consensus LOW of $8.171. Cross-check: 20.7x 2027E; 28.7x the trailing GAAP EPS of $5.68 and 32.5x the clean $5.01. The scenario needs no catastrophe: the second-half deceleration consensus already models turns out to be real, the pension credit keeps decaying, tariff costs stick, one fixed-price development programme produces a 2024-style EAC charge, and the market pays a normal industrial multiple for a 7%-growth conglomerate with negative tangible book. −25.2%.
• Base ~$227 — 26x the 2028 consensus MEAN of $8.747. Cross-check: 28.8x 2027E; 31.4x 2026E; rebuilt EV of $336.4B on 2028E revenue of $110.637B is 3.04x sales. Sensitivity, stated openly: 24x gives $210 — 3.6% BELOW spot — and 28x gives $245, above the street. The whole answer lives in a 24–28x band and we claim no more precision than that. The scenario: the backlog converts roughly on the consensus schedule, RTX beats the near-term numbers as it has five times running, margins keep grinding higher, and the multiple gives back a little as growth decelerates and the pension credit fades. +4.2%, plus about 0.5% of net shareholder yield.
• Bull ~$273 — 29x the 2028 consensus HIGH of $9.407. Cross-check: 34.6x 2027E. The scenario: the FY2027 defence budget lands near the $1.5 trillion request, the $67 billion supplemental with $21 billion of munitions passes, Golden Dome awards land against the $24.4 billion appropriated, PW1100 aircraft-on-ground clears in 2026 and releases the Pratt & Whitney margin headwind, the IEEPA tariff refund is approved, and the buyback restarts. This is the only one of our three anchors that requires multiple expansion, and we say so. +25.3%.

Base is 4.2% above spot; asymmetry roughly 1.00:1 (25.2% down, 25.3% up). That is a coin flip, and it is the arithmetic reason this is a Watch. Our base sits 4.9% BELOW the street's $238.50 and 5.4% below the median $240. The street's own LOW target of $215 is below the current price. When the base case is a symmetric bet, sits under the crowd's, and the entry is 0.30% from a 52-week high with RSI at 80, the honest answer is to wait for the price to do some of the work.

7. Knowledge base — six sector claims, five Nvidia collisions, and a lane that disagrees with itself

Raw hits: 11. Entity matches: 6. Text-only matches: 5. Used as supporting colour: 6. Used as name-level conviction: 0. Discarded for collision: 5.

The collision, found and discarded exactly as expected

All five text-only matches were Nvidia graphics products. The string "RTX" is Nvidia's consumer and workstation GPU brand, and every text-only hit came from the same channel and the same speaker:

Not one names RTX Corporation, and the entity lists confirm it: every one carries NVIDIA or NVDA and none carries RTX, Raytheon, Collins or Pratt & Whitney. Two of them (#3) even register "RTX Pro Server" and "RTX Pro 6000" as entities in their own right, which is the extraction correctly identifying a product line and not a defence contractor. All five are discarded in full. This is the single cleanest collision case in the batch and it would have contaminated the lane by 45% if the text-only hits had been used naively.

The six that survive, verbatim

> 2026-03-03 · bullish · conviction 55 · horizon: thesis · entities: L3Harris, RTX, Rocket Lab, Kratos, Palantir, Elbit Systems · channel: compound_and_friends · NO NAMED SPEAKER

> "21st-century war is fought with cheap one-way kamikaze drones that must be constantly replaced; JEDI ETF captures the munitions/drone basket."

> 2026-03-16 · bullish · conviction 70 · horizon: fact · entities: Raytheon, RTX · channel: odd_lots · NO NAMED SPEAKER

> "Raytheon announced it will ramp five munitions lines, with more increases to come, as the Pentagon pushes primes to maximize output."

> 2025-06-28 · neutral · conviction 60 · horizon: thesis · entities: Raytheon, Lockheed Martin · channel: natalie_brunell · NO NAMED SPEAKER

> "Defense primes face demand to replenish depleted missile stocks but can't produce rocket motors domestically — a manufacturing/supply-chain constraint."

> 2025-05-06 · BEARISH · conviction 68 · horizon: thesis · entities: Raytheon · channel: natalie_brunell · NO NAMED SPEAKER

> "A >$1T defense budget is unsustainable and should be cut ~50%; defense primes profit from repairing old junk and reproducing decades-old systems."

> 2026-04-22 · BEARISH · conviction 66 · horizon: thesis · entities: RTX, GD, BA · speaker: Jim Bianco · channel: jim_bianco · skill 0.8

> "Raytheon, General Dynamics and Boeing are the past; cheap-drone warfare demands new tech-oriented defense, so the primes are correcting and underperforming despite the conflict — a bear's trend."

> 2026-04-22 · bullish · conviction 48 · horizon: thesis · entities: BA, RTX, NOC · speaker: Arthur Hayes · channel: arthur_hayes · skill 1.2

> "War means money printing and ramped bank lending to finance the merchants of death — Boeing, Raytheon, Northrop Grumman get the financing to expand weapons production."

What this lane is, stated without inflation

All six are entity matches, so the attribution discipline is satisfied — RTX or Raytheon genuinely appears in the entity list of each. But not one of the six is an analysis of RTX Corporation as a security. Every one uses the company as a member of a set or as an illustration of a macro thesis:

• The 2026-03-03 claim is an ETF basket call naming six tickers. It assigns no view to any constituent.
• The 2026-03-16 claim is a reported fact (five munitions lines ramping), not a judgement. It is nonetheless the most useful item in the lane because it corroborates, from an independent direction, exactly what the 10-Q's $37 billion of first-half defence bookings shows.
• The 2025-06-28 claim is a supply-chain constraint on the whole prime cohort — rocket-motor capacity — and is a genuine, checkable limiter on how fast any backlog converts. Neutral in stance and correctly so.
• The 2025-05-06 claim is a political argument about the defence budget, not about RTX. Bearish, conviction 68, and it is the direct negation of the appropriations tailwind in section 2.
• The 2026-04-22 Bianco claim is the closest thing to a security view in the lane and it is bearish: the primes "are correcting and underperforming." It is also, on this file's evidence, wrong so far — RTX has returned +39.0% over twelve months against SPY's +24.3%, and +26.0% over three months against +7.6%. We report the miss rather than quietly dropping the claim.
• The 2026-04-22 Hayes claim is a monetary-regime thesis in which defence primes are the downstream beneficiary. Conviction 48 — the lowest in the lane — and the reasoning is about money printing, not about backlog, margin or valuation. Arthur Hayes is a promoter-class macro voice; under house policy that frame is disclosed and the claim is weighted low regardless of its skill multiplier of 1.2.

Four of the six carry speaker: null and are attributed only to a channel. Under the 4-lane attribution policy that is the weakest form of sourcing we accept: a channel-level attribution tells you where a claim was said, not who said it or what their record is. Two of those four come from the same channel (natalie_brunell) and take opposite sides — one neutral on the primes' production constraint, one bearish on the entire defence budget — which is precisely the pattern channel-level attribution cannot resolve, because it is almost certainly two different guests. We flag that explicitly rather than treating "natalie_brunell" as a coherent view.

Per-speaker concentration and reversals. No speaker supplies more than one claim. There are no position reversals visible in the dates. No claim in this lane carries a speaker_role of management, so the half-weight management discount does not apply anywhere — and unlike the Lilly lane, there is not a single company-sourced claim here. That is worth saying: this is a lane with no promotional capture from the issuer, only from the macro commentators.

Net stance is genuinely split — three bullish (conviction 55, 70, 48), two bearish (68, 66), one neutral (60) — and the two bearish claims carry the HIGHER convictions. Skill-weighted, the bullish side has 55×1.0 + 70×1.0 + 48×1.2 and the bearish side 68×1.0 + 66×0.8. There is no net signal here worth trading.

Conclusion, stated as a finding rather than dressed up. The Synthos knowledge base has no view on RTX Corporation. It has a live, unresolved argument about whether exquisite defence primes or cheap attritable drones own the next decade, and RTX appears on both sides of it. It has one useful production fact (five munitions lines) and one useful constraint (domestic rocket-motor capacity), both of which we have used as corroboration in sections 2 and 11. Neither is name-level conviction, neither justifies a conviction premium in the fair value, and the split stance is a direct input to the Watch verdict. Conviction rating Low, breadth 5 channels, net conviction mixed-low.

8. Data integrity — what we rejected from the vendor file and why

Fourteen findings. Listed rather than silently corrected. RTX's file is mid-pack: the share count, market capitalisation, currency and enterprise value are all clean and independently verified, while the segment, geographic and balance-sheet-vintage fields carry real defects.

1. bal_a[0] is SEVEN MONTHS STALE — the annual balance sheet is 2025-12-31 and the archive contains a 10-Q for 2026-06-30. A naive net-debt rebuild from bal_a gives $32.071B. The correct June figure is ~$30.55B. The staleness OVERSTATES net debt by $1.52B, or 5.0% — the opposite direction to LLY, because RTX has been repaying debt, not issuing it. The 10-Q's evidence: no revolver borrowings, no commercial paper at either date, a note heading that reads "We made the following repayments of long-term debt" with no issuance heading anywhere, and interest expense down from $502M (Q1 2025) to $417M (Q2 2026). Filing-verified cash is $8.3 billion. We use the June basis.

2. The 8-K financing check was run and came back EMPTY — no post-statement bond deal exists in this archive. Three 8-Ks: 2026-04-21 (Q1 earnings cover), 2026-05-04 (annual-meeting vote results), 2026-07-23 (Q2 earnings cover). None is an Item 1.01, 2.03 or 8.01 financing. The 10-Q references "an existing universal shelf registration statement... filed... on September 18, 2025, for an indeterminate amount of debt and equity securities" — capacity, undrawn. The LLY failure mode does not apply here, and we say so because the test was run.

3. netDebt and short-term investments — TESTED, and there is genuinely nothing to add back. shortTermInvestments reads $0 across all three years of bal_a. That is correct. The 10-Q's liquidity paragraph describes only "cash and cash equivalents of $8.3 billion" with no "marketable securities" formulation. The $2.132B of longTermInvestments is customer financing and equity-method interests; the "marketable securities held in trusts" at Note 10 are nonqualified-plan assets in Other assets, dedicated to benefit obligations. The AAPL/BRK-B defect class does not apply. Stated because the test was run and passed.

4. seg_prod omits the "Eliminations and other" line for 2024 and 2025 — the segment lines sum to 102.9% of reported revenue. Collins $30.196B + Pratt & Whitney $32.916B + Raytheon $28.043B = $91.155B against reported revenue of $88.603B — a $2.552B gap; 2024 gaps by $2.325B on the same basis. The 10-K names the mechanism: "Segment Total net sales and Operating profit (loss) include intercompany sales and profit, which are ultimately eliminated within Eliminations and other." Proof this is an omission and not the company's structure: the same vendor field DOES carry "Eliminations and Other" of −$1.979B for fiscal 2023, and the 2023 lines then sum to exactly $68.920B — the reported figure, to the dollar. We add the contra line back ourselves and use the segment magnitudes, which independently reconcile to the 10-K's disclosed organic growth by segment ($4.8B P&W, $2.6B Collins, $1.7B Raytheon).

5. The seg_prod segment LABEL is superseded. The vendor writes "Raytheon Intelligence & Space" for 2024 and 2025. RTX merged Raytheon Intelligence & Space and Raytheon Missiles & Defense into a single "Raytheon" segment; both the 2025 10-K and the June 2026 10-Q state "three principal segments: Collins, Pratt & Whitney, and Raytheon." The vendor also writes "Collins Aerospace Systems" where the filings say "Collins Aerospace." Corrected in presentation throughout.

6. The seg_prod history before 2020 is a DIFFERENT COMPANY and is discarded entirely. The 2018, 2017 and 2016 rows carry Otis and Carrier (spun off in 2020); the 2010 row carries Sikorsky (sold in 2015); and the 2015–2011 rows carry a single implausible line, "Commercial Aerospace" of $1.7–3.5B. The 2022 row is also broken on its own terms: RIS $29.175B + Collins $20.597B + P&W $20.530B = $70.302B against reported revenue of $67.074B, with no eliminations line and a "Raytheon Intelligence & Space" figure that appears to conflate two then-separate segments. Nothing before fiscal 2023 appears in this dive.

7. seg_geo for 2025 is a HYBRID of two different 10-K tables and sums to 131% of revenue — REJECTED in part. The five 2025 lines sum to $116.090B against $88.603B of revenue. The 10-K contains "Geographic External Sales by Origin" (Note 20) and, separately, "Segment sales disaggregated by geographic region based on customer location." The vendor's 2025 "Geographic Distribution, Domestic" of $74,778M is 84.4% of revenue — a by-ORIGIN figure — sitting on top of four customer-LOCATION international lines. The 2024 row is clean (US $46,087M + four regions = $80,738M exactly) and is used as supplied. The 2023 row also ties exactly but its US share of 83.5% marks it as the by-origin table, so it is not comparable to 2024 and is excluded from the trend. For 2025 we use the four international lines ($41.312B, 46.6%) and derive the United States figure as the residual ($47.291B, 53.4%), labelled as derived.

8. Non-service pension income is 11.7% of trailing EPS and is NOT operating income — a clean operating EPS is derived and used. The vendor supplies no line for it; we derived $1,118M trailing from totalOtherIncomeExpensesNet + interestExpense − interestIncome and verified the derivation against both of the 10-K's disclosed year-on-year changes ($0.3B in 2025, $0.3B in 2024) and against the isolated $0.3B Q4 2025 Prudential settlement charge. Clean trailing operating EPS is $5.01 against reported GAAP diluted EPS of $5.68; the honest trailing multiple is 43.5x, not 38.4x. Note also that a FAS/CAS operating adjustment sits INSIDE operating profit and the 10-K discloses only its year-on-year change ($0.1B in 2025, $0.3B in 2024) — the numeric table was stripped, so a second pension-driven swing factor cannot be sized from this archive and we do not estimate it.

9. interestIncome is ZEROED in two of nine quarters — flagged because it feeds the derivation above. The field reads $0 for 2026-06-30 and 2025-12-31 while the seven neighbouring quarters range $19M–$51M. The effect is to overstate our derived non-service pension income in those two quarters by roughly $25M each, or about 2%. Immaterial to every conclusion, but disclosed because the reader is entitled to know the derivation's error bar. Related zeroes that are NOT defects: generalAndAdministrativeExpenses and sellingAndMarketingExpenses read $0 because RTX reports a single combined SG&A line of $6.095B, and deferredRevenueNonCurrent reads $0 because RTX classifies its $21.615B of contract liabilities as current — the current-liability subtotal reconciles exactly.

10. est.ebitdaAvg and est.ebitAvg — TESTED for the arithmetically impossible signature, PASSED, but NOT USED. No year has a negative ebitdaAvg, ebitAvg exceeds ebitdaAvg in no year, and ebitdaAvg never equals minus revenue. This file is in the clean 66.5%. But verified against actuals both run low: 2025 ebitAvg of $8.887B is 4.4% below actual operating income of $9.300B, and ebitdaAvg of $14.049B is 6.0% below the actual $14.946B. Consistently low, not nonsensical. All forward valuation runs on epsAvg and revenueAvg, per the data contract.

11. GAAP and non-GAAP are mixed across blocks and the gap is 25% — labelled throughout. inc_a/inc_q are GAAP; est and earn_cal are non-GAAP adjusted EPS. 2025 consensus adjusted EPS of $6.211 against GAAP diluted EPS of $4.96 is a 25.2% premium. The 10-Q quantifies the main bridge: acquisition accounting adjustments of "$0.4 billion, net of tax, which had an unfavorable impact on diluted earnings per share (EPS) of $0.27" per quarter. Forward multiples in this dive use consensus non-GAAP; trailing multiples are shown on GAAP (38.4x) and on our clean operating basis (43.5x). We state the basis on every figure.

12. lastDividend: 2.77 is the trailing-twelve-month total, not the last declared dividend — corrected from the filings. The 10-K records $0.68 per share declared 6 February 2026; the 10-Q records $0.73 declared 30 April 2026 and $0.73 declared 26 June 2026. Three quarters at $0.68 plus one at $0.73 is $2.77 — the vendor's figure, correctly computed but mislabelled. The forward run-rate is $2.92, a 1.340% yield, not $2.77 and 1.271%.

13. quote.yearHigh/yearLow disagree with tech.hi52/lo52 — we use tech and say why. The quote block reports a 52-week high of $221.34 and low of $150.61; the computed technical block reports $218.58 and $151.75. The gaps are 1.3% and 0.8%, and the explanation is that quote uses intraday extremes while tech is computed from the six-year daily CLOSE series. We use tech throughout, because the moving averages, drawdown and relative-return figures come from the same series and mixing sources would make the percentages inconsistent. Note also that tech.max_dd_from_peak (−0.2974%) equals tech.pct_from_hi exactly — the six-year peak IS the 52-week high, because RTX is at a multi-year high. Per house guidance that is NOT a defect and it is not reported as one.

14. Finance leases ARE folded into totalDebt, consistently, at 4.1% — disclosed, not rejected. totalDebt of $39.506B (2025) reconciles exactly to short-term debt $3.616B + long-term debt $34.288B + capitalLeaseObligations $1.602B, and the same identity holds for 2024 ($42.893B) and 2023 ($45.239B). The treatment is consistent across all three years and the lease share is modest — nothing like MSFT's 68.7%. We disclose it and use totalDebt as supplied.

Not defects, correctly reported and independently confirmed. The share count is exact: the 10-Q cover page states "At June 30, 2026, there were 1,347,758,144 shares of Common Stock outstanding", and 1,347,758,144 × $217.93 = $293,716,922,916 against the vendor's reported market capitalisation of $293,716,922,951 — agreement to 0.00000001%. Two further checkpoints corroborate: the 10-K cover gives 1,342,287,676 at 31 December 2025 and the 8-K of 2026-05-04 gives 1,345,974,220 at the 3 March 2026 record date, a coherent rising series consistent with the diluted counts in inc_q. Currency is clean: inc_a[0].reportedCurrency is USD, profile.currency is USD, and the quote is USD on the NYSE — no ASML-style mixing. The enterprise value is right even though the balance sheet feeding it is stale: enterpriseValueTTM of $324.268B less market cap gives implied net debt of $30.551B, which independently reconciles to derived total debt of $38.853B less the filing-verified $8.304B of cash to within $2 million. Interest coverage of 6.26x reconciles to operating income of $10,485M over interest expense of $1,676M. The effective tax rate of 18.73% reconciles to $1,872M of tax on $9,997M of pre-tax income. And the segment magnitudes tie to the 10-K's disclosed organic growth by segment, which is the strongest available check on a table whose contra line is missing.

Non-equity tripwire — checked, and the low beta was investigated rather than assumed benign. RTX is common stock, NYSE-listed, CUSIP 75513E101, ISIN US75513E1010, and the 10-Q cover confirms "shares of Common Stock outstanding." The tripwire is worth running here because beta is 0.298 — genuinely low, and in the range where a baby bond or preferred would sit. It is not one. The price of $217.93 is not par-like ($25, $50 or $1,000); the 52-week range of $151.75 to $218.58 is a 44% band, far wider than any fixed-income instrument shows; the dividend is variable and was raised from $0.68 to $0.73 in April 2026; volume was 5.89M shares, roughly $1.28 billion of turnover; and the 8-K exhibit index shows RTX separately registers "2.150% Notes due 2030" under Section 12(b) — confirming that RTX has listed debt, and that it trades under a different identifier from the common stock this dive concerns. The 0.298 beta is a genuine characteristic of a defence prime with three years of contracted backlog, not an instrument error. This is common equity.

Known extraction limits, stated rather than estimated around. Numeric tables are stripped from the current filing text extraction — assume PROSE-ONLY. In this archive that specifically cost us: the 10-Q and 10-K balance sheets and long-term-debt schedules (so total debt at 30 June 2026 is DERIVED, not read); the segment net-sales and operating-profit tables (so segment margins are unavailable); the U.S. government sales and international sales tables in Item 1; the backlog-by-segment table; the FAS/CAS operating adjustment components; the pension funded-status tables; and the Exhibit 99 press releases behind both earnings 8-Ks. The freshest hard numbers in this dive — the June-quarter income statement, the trailing ratio block and the estimate block — come from the VENDOR, not from a filing, and are labelled as such throughout. The filing-verified June 2026 figures are: cash $8.3 billion, share count 1,347,758,144, backlog $289B / $170B / $119B, defense bookings $23B and $37B, Raytheon backlog $86B, the Powder Metal accrual of $0.4 billion, the $5.0 billion undrawn revolver, and every quotation in this document.

9. Technicals

• Price $217.93. −0.30% from the 52-week closing high of $218.58, and +43.6% above the 52-week low of $151.75. Position within the annual range: 99th percentile. The day's high of $218.60 marginally exceeded the trailing 52-week closing high, so today printed a fresh high on an intraday basis.
• Moving averages, both supportive and both far below: +13.4% above a rising 50-day average of $192.12; +15.6% above a 200-day average of $188.44. The 50-day sits above the 200-day. This is an unambiguous uptrend with no technical damage anywhere in the structure — and no nearby support either.
• RSI 79.99 — the most overbought reading in this batch, and above the conventional 70 threshold by a wide margin. MACD +7.47, positive and rising.
• Maximum drawdown from the six-year peak: −0.30%. RTX is at its highest level in the entire price history this file holds.
• Relative performance, and the shape of it is the finding: 3-month +26.0% vs SPY +7.6% and QQQ +7.7%; 6-month +8.4% vs SPY +11.1% and QQQ +15.7%; 12-month +39.0% vs SPY +24.3% and QQQ +30.8%. RTX LAGGED the S&P over six months and beat it by 18 points over three. The entire twelve-month outperformance was earned in the last quarter. That is a re-rating, not a grind.
• Sentiment: 17 buy, 9 hold, 0 sell across 26 analysts — not a single sell rating. Consensus target $238.50 (+9.4%), median $240 (+10.1%), high $250 (+14.7%), low $215 — 1.3% BELOW spot. The $215–$250 band is a 16% spread, one of the tightest in this batch. Unanimous non-negativity combined with a low target below spot is a crowded, fully-valued configuration, not a contrarian one.

Today's move and what it does to the entry

RTX closed 2026-08-04 at $217.93, up 0.59% or $1.28 from a previous close of $216.65. It opened at $217.38, traded a $214.53 to $218.60 range, and closed 0.31% below the day's high on 5.89M shares against a 5.38M average. This was a quiet day and RTX did not participate in the event that defined the batch. 2026-08-04 was a violent semiconductor melt-up — LRCX +7.85%, MRVL +12.81%, INTC +10.92%, MU +7.62%, AMD +7.00%, KLAC +6.95%, ARM +17.36%, TXN +5.42% — and RTX, an industrial, moved 0.59%. There is no sector beta to strip out here and no company-specific news in this file for 2026-08-04. The last company event was the 2026-07-23 earnings release and 10-Q, twelve days ago.

The honest read, and it is the opposite of the KLA setup in this batch. KLAC gapped 7% from 35% below its high, into resistance. RTX drifted 0.6% higher at the top of its range with RSI at 80, after a 26% three-month run, having lagged the index over six months. In pure chart terms this is a strong, intact, extended uptrend with the nearest meaningful support 11.8% below at the 50-day average of $192.12.

So the gap is not the problem — the absence of one is. There is no dislocation to buy. Even after the re-rating, RTX's fundamental case is intact and improving, but our base fair value of $227 is only 4.2% above the closing price and 4.9% below the street's $238.50, and the street's own low target of $215 is below spot.

What we are therefore telling the reader, plainly: the business is worth owning and this is not the price at which to start. Two setups would change it, and they are the falsifiers:

• A price near $195–200 — roughly 22–23x the 2028 consensus and 25–25.5x 2027 — which is a normal pullback to the rising 50-day average and would put our base $227 at a 14–16% premium. That is 8–10% below today's close and well inside the range this stock has traversed in a quarter.
• Or the 2026-10-20 print landing materially above the $23.9997B revenue consensus with operating margin above 11.5%, which would kill the second-half deceleration embedded in 6a and justify a higher base. At that point paying up is defensible, because the uncertainty being paid for has been resolved.

Buying today means paying the 99th percentile of the annual range with RSI at 80 for a 4.2% base-case return and a 25% bear case, with no company-specific catalyst for 77 days. That is not a trade with an edge in it.

10. Insiders — eight transactions, and unlike most files in this batch these ARE sales

Every insider transaction in this file was filed on 2026-07-28, for transactions dated 24–28 July 2026 — the days immediately after the 2026-07-23 earnings release, which is when a closed trading window normally reopens.

Open-market sales: 17,862 shares, approximately $3.82 million at the transaction prices. Open-market purchases: ZERO.

The reading, stated carefully, because this table means something different from most in this batch. There is not a single F-InKind transaction in this file. F-InKind is the code for shares withheld by the issuer to satisfy tax on vesting restricted stock, and it is not a sale — every one of KLA's eight transactions was that code, which is why that table was read as neutral. RTX's are S-Sale: genuine open-market dispositions.

Three specifics are worth naming.

Troy Brunk, President of Collins Aerospace, exercised 12,600 stock appreciation rights struck at $97.65 and disposed of essentially the entire resulting position. The $210.17 "D-Return" of 5,854 shares is the standard mechanism for covering the exercise cost and withholding; the two S-Sale legs of 1,811 and 6,746 shares at $210.65 and $210.20 are open-market sales. His holding went from 23,220 shares to 8,809 — a 62% reduction. Exercising a deep-in-the-money award and selling is unremarkable behaviour and carries limited information.

Tracy Atkinson, a director, sold 2,295 shares at $218.0497 — within 0.06% of today's closing price — leaving 1,785. That is a 56% reduction in a non-employee director's stake, executed at the top of the 52-week range. Directors do not receive the same volume of equity compensation as operators, so the discretionary content of a director sale is higher.

Kevin DaSilva, Senior VP and Treasurer, sold 7,010 shares across two days at $213.62 and $216.932, leaving 20,099 — a 26% reduction. No offsetting exercise appears, so on this file's evidence these are discretionary sales of previously held stock by the officer responsible for the company's capital structure.

What the file contains is therefore: zero open-market purchases and roughly $3.8 million of open-market sales by three insiders, concentrated in the four trading days after a strong print, at prices between $210 and $218. $3.8 million against a $293.7 billion market capitalisation is not a meaningful capital event and we will not inflate it into one. But it is directionally negative, it is the opposite of the tax-withholding pattern that dominates most insider tables, and not one insider bought a share into a 26% three-month advance. Combined with the halted buyback, the message from people who know the company best is that at $210–218 they are net sellers of their own stock. We weight that as a mild negative and no more.

11. Verdict, kill-criteria and flip conditions

Watch.

This is the best forward disclosure in this batch and the least attractive entry point.

What is genuinely excellent, and none of it is in dispute: a filing-verified backlog of $289 billion at 3.09x revenue, up $21 billion in six months and $71 billion in eighteen, of which roughly a quarter converts within twelve months and therefore covers about 70% of the entire 2026 consensus revenue line before RTX wins another order; defense bookings of $37 billion in the first half of 2026 against $21 billion a year earlier, itemised program by program in the 10-Q down to $542 million of SM-3; June-quarter revenue up 14.5% year on year, the fastest in the file, with gross margin at 20.77% and operating margin at 11.38% against 5.17% at the 2023 trough; five consecutive EPS beats averaging +13.2%; an installed base — more than 2,600 GTF aircraft, the sole F-35 powerplant, 21 aftermarket facilities — that is as durable a commercial annuity as industrials offer; and credit outlooks that IMPROVED at both agencies in 2026 with a $5.0 billion revolver untouched and twelve-year average debt maturity.

What we are declining to pay for: 38.4x trailing GAAP earnings, 43.5x once the non-operating pension income is stripped, 30.1x the 2026 consensus and 27.6x 2027 for roughly 10% annual EPS growth; a base fair value of $227 that is 4.2% above spot and 4.9% below the street's $238.50, giving asymmetry of essentially 1.00:1; 11.7% of trailing EPS coming from a pension credit that has fallen 36% in two years; ROIC of 7.33% against $85.2 billion of goodwill and intangibles whose 2025 impairment review stopped at a qualitative step 0, with tangible book value of negative $11.73 per share; a buyback that has fallen from $12.870 billion to $50 million with $0.6 billion of authority left, a share count rising 0.8% a year, and total shareholder yield of about +0.5% — which a January 2026 Executive Order may permit the Secretary of War to restrict outright; a knowledge-base lane of six sector claims that disagrees with itself; and an entry at the 99th percentile of the annual range with RSI at 80.0 after a 26% three-month run.

The distinction that matters. We are not saying the backlog will not convert. We are saying that the backlog is the single most visible bull argument in the market, it is disclosed in plain English twice a year, the stock has re-rated 26% in three months while we watched, and there is nothing left in the price to compensate for the ways this could disappoint. In the Synthos frame, a name where the fair value sits below consensus, the risk/reward is symmetric, the conviction lane is split, and the chart is at a multi-year high is a Watch by construction. The business is right; the price is late.

Pre-registered KILL criteria — what would take this to Avoid:

• Any disclosure that other engine models are materially affected by the powder-metal condition. The 10-K states plainly that "other engine models within Pratt & Whitney's fleet contain parts manufactured with affected powder metal." The first occurrence cost $5.4 billion of net sales, a $2.9 billion pre-tax operating charge net of the 49% partner share, and $1.55 of diluted EPS. A second would be a thesis-ending event at 30x forward earnings.
• A material adverse EAC adjustment or fixed-price development termination on the scale of the 2024 Raytheon Contract Termination ($0.5 billion of cost of sales, $0.6 billion of segment operating profit). With $289 billion of backlog and a company that warns "we carry the burden of any cost overruns" on firm-fixed-price work, this is the mechanism by which a great order book becomes a bad one.
• Total backlog declining sequentially, or defense bookings falling below roughly $12 billion in a quarter — the level of the June 2025 comparison. Backlog is the whole forward case; a reversal removes it.
• A Secretary of War determination under the January 7, 2026 Executive Order limiting RTX's dividends or repurchases. At a +0.5% total shareholder yield there is very little to take away, which is exactly why the market would reprice the governance risk rather than the cash.
• Operating margin falling back below 10% on tariffs, supply chain or mix, breaking the 5.17% → 11.38% recovery that the entire multiple expansion rests on.
• Non-service pension income falling below roughly $700 million annualised, which would remove another $0.25 or so of clean EPS with no operational cause.

Pre-registered FLIP conditions — what would take this to Buy — Tactical:

• Price below approximately $195–200 — roughly 22–23x the 2028 consensus, 25x 2027, and a normal retest of the rising 50-day average at $192.12 — with no deterioration in backlog or bookings. This is by far the most likely path to a Buy and it is only 8–10% below today's close. At $197 the base case is +15% and the asymmetry moves to roughly 2:1.
• The 2026-10-20 print above $23.9997B of revenue with operating margin above 11.5%, which would invalidate the second-half deceleration embedded in consensus and support a base nearer $245. A sixth consecutive double-digit beat is our expectation, and it is the single most probable positive catalyst on the calendar.
• A restart of the share repurchase programme at scale, with the authority extended beyond the remaining $0.6 billion. That would convert a +0.5% shareholder yield into something worth having and would signal management does not regard the Executive Order as binding.
• Approval of the IEEPA tariff refund against the approximately $0.5 billion paid, which RTX has explicitly not booked as a receivable — a clean, unmodelled benefit.
• A named Golden Dome award against the $24.4 billion appropriated, or passage of the $67 billion supplemental with its $21 billion of munitions replenishment. The 10-Q says "RTX's portfolio is well-positioned"; that is management's voice and is half-weighted, but an actual award is not.
• Genuine name-level knowledge-base conviction appearing — an analytical claim about RTX Corporation rather than about the sector containing it. At present there is none, and the lane's two highest-conviction claims are both bearish.

Where RTX fits in the Synthos Framework Portfolio. The aerospace and defence sleeve, at 0% today with a 2–3% target on a fill near $195–200 or a confirmed October beat. On the batch overlap question: RTX is the only industrial in a batch dominated by semiconductors, and its 0.298 beta and 0.59% move on a day when the semis went up 5–17% make it a genuine diversifier rather than another expression of the same trade. Within aerospace and defence, the vendor peer set makes the internal choice explicit — GE Aerospace ($391.5B), Boeing ($187.4B), Lockheed Martin ($136.0B), General Dynamics ($104.4B), Northrop Grumman ($78.4B), TransDigm ($71.3B), L3Harris ($53.1B) — but no peer multiples are supplied in this file, so we draw no peer-multiple conclusion and will not pretend to one. What can be said is that RTX is the only one of these with all three of a commercial aftermarket annuity, an engine monopoly on the West's primary fighter, and a missile business at the centre of the largest munitions replenishment in forty years. That combination is worth a premium. Whether it is worth 30x is the question this dive answers "not quite, not here." Logged as a tracked Synthos call (Watch) as of 2026-08-04 at $217.93.

Single biggest risk: a de-rating that requires no bad news. At 30.1x the 2026 consensus, RTX is priced as a growth compounder. The consensus growth algorithm is revenue +7–8% and adjusted EPS +10%, the return on invested capital is 7.33%, tangible book is negative, and 11.7% of the trailing earnings are a decaying, non-operating pension credit. A 20x multiple on the 2028 consensus low is $163 — 25% below spot — and 20x is not a distress multiple for an industrial growing 7%; it is an ordinary one. The bear case here is not that RTX stops executing. It is that RTX keeps executing exactly as consensus expects and the market decides it should pay an industrial multiple for an industrial. There is no valuation cushion to absorb that, and the backlog does not protect the multiple — only the revenue.

Most fragile assumption in the price: that the non-service pension income holds. Every other assumption in section 6a is a volume or margin assumption, and volume is observable quarter by quarter against a $289 billion backlog that RTX republishes twice a year. This one is invisible to almost every screen, it is not stripped by consensus adjusted EPS, and it is already falling: $1,828M (2023) → $1,524M (2024) → $1,170M (2025), with $1,118M trailing. The December 2025 Prudential buy-out permanently removed $2.3 billion of gross obligations and approximately 60,000 retirees — one-third of the plan's beneficiaries — and with them the asset base that generates the credit. If the line falls another $400 million, that is roughly $0.24 of clean EPS gone with no operational cause, and at 27.6x that is about $6.60 of share price. Operations can absorb it — organic operating profit rose $0.5 billion in the June quarter alone. But nothing in consensus adjusted EPS tells a reader that the offset is required, and that is precisely why it is the most fragile leg.

Provenance & disclosures

• Traceability: 11 raw knowledge-base hits scanned against 51,928 distilled claims; 6 entity matches used as supporting colour; 5 text-only matches DISCARDED IN FULL as collisions on Nvidia's "RTX" graphics-product brand (all five from the jensen_huang channel, entities NVIDIA / NVDA / "RTX Pro Server" / "RTX Pro 6000", covering the RTX 50-series launch, RTX Pro/MGX, the RTX Pro server, ray tracing, and RTX Spark — not one names RTX Corporation, Raytheon, Collins or Pratt & Whitney). Zero name-level conviction claims; all six survivors are sector, basket or macro claims and this is reported as a finding, not padded. Sources: compound_and_friends (1, 2026-03-03, bullish, conviction 55, no named speaker — JEDI drone/munitions ETF basket), odd_lots (1, 2026-03-16, bullish, conviction 70, no named speaker — five munitions lines ramping), natalie_brunell (2 — 2025-06-28, neutral, conviction 60, no named speaker, domestic rocket-motor constraint; and 2025-05-06, bearish, conviction 68, no named speaker, >$1T defence budget unsustainable), jim_bianco (1, 2026-04-22, bearish, conviction 66, speaker Jim Bianco, skill 0.8 — the primes "are the past"), and arthur_hayes (1, 2026-04-22, bullish, conviction 48, speaker Arthur Hayes, skill 1.2 — war means money printing; promoter-class macro voice, frame disclosed, weighted low). Four of six carry speaker: null and are attributed only to a CHANNEL — the weakest sourcing the 4-lane attribution policy admits — and two of those four share the same channel while taking opposite sides, which is exactly the failure mode channel-level attribution cannot resolve. No speaker supplies more than one claim; no position reversals are visible in the dates; no claim carries a management speaker_role, so the half-weight management discount does not apply anywhere and this lane contains no issuer-sourced promotion. Net stance is split 3 bullish / 2 bearish / 1 neutral, with the two bearish claims carrying the higher convictions. All quotes are verbatim from the stored claim text.
• Data as-of: fundamentals — income statement for the quarter ended 2026-06-30 and balance sheet / cash flow annuals through 2025-12-31, sourced from the vendor payload, cross-checked against the earnings-calendar entry for 2026-07-23 (revenue actual $24,708M matching the income statement exactly; adjusted EPS $1.89 against a $1.66 estimate); most recent filing-verified figures 2026-06-30, from the 10-Q filed 2026-07-23 · estimates 2026-08-04 · prices 2026-08-04, quote timestamp 1785873602 ≈ 2026-08-04T20:00:01Z ($217.93, +0.59%; 50-DMA $192.12; 200-DMA $188.44; RSI 79.99; MACD +7.47) · knowledge-base claims 2026-08-04. Fiscal/calendar note: RTX's fiscal year is the calendar year. "2025" means the twelve months to 2025-12-31 and "Q2 2026" means the June 2026 quarter. All figures come from the Synthos vendor data file for RTX or from the SEC filings in the RTX archive; no figure comes from memory, recall or external retrieval.
• Filing archive contents and its gaps: 10-K filed 2026-02-06 (fiscal 2025); 10-Q filed 2026-04-21 (March 2026 quarter); 10-Q filed 2026-07-23 (June 2026 quarter — the filing-verified anchor for this dive); 8-K filed 2026-04-21 (Item 2.02, first-quarter 2026 results); 8-K filed 2026-05-04 (Item 5.07, annual-meeting voting results, disclosing 1,345,974,220 shares outstanding at the 2026-03-03 record date); 8-K filed 2026-07-23 (Item 2.02, second-quarter 2026 results). BOTH earnings 8-Ks are present as the cover document ONLY. The 2026-07-23 8-K refers the release to "Exhibit No. 99" — labelled Exhibit 99, not 99.1 — and that exhibit is ABSENT from the extraction; the file is 2,667 bytes and contains not a single financial figure. The same is true of the 2026-04-21 8-K. No financing 8-K exists in this archive, and the 8-K financing check that caught LLY's understated debt was run and returned nothing. Numeric tables are stripped from the current filing-text extraction throughout: this cost us the balance sheets and long-term-debt schedules in both the 10-K and the 10-Q (so total debt at 2026-06-30 is DERIVED from the vendor ratio block and cross-checked three ways, not read off a filing), the segment net-sales and operating-profit tables, the backlog-by-segment table, the U.S. government and international sales tables, the FAS/CAS operating adjustment components, and the pension funded-status tables. Filing-verified June 2026 figures, in full: cash and cash equivalents $8.3 billion; 1,347,758,144 shares outstanding; total backlog $289 billion / commercial $170 billion / defense $119 billion; defense bookings $23 billion in the quarter and $37 billion in the half; Raytheon segment backlog $86 billion; Powder Metal customer-compensation accrual $0.4 billion; estimated full-year 2026 Powder Metal cash impact approximately $0.7 billion; $5.0 billion revolver undrawn to August 2028; no commercial paper outstanding; average debt maturity approximately 12 years; remaining buyback authority approximately $0.6 billion; dividends of $0.73 per share declared 30 April and 26 June 2026.
• Where the filings contradicted or corrected the vendor (detailed in Section 8): bal_a[0] is the 2025-12-31 balance sheet and is SEVEN MONTHS STALE against a 10-Q covering 2026-06-30 — a naive rebuild gives net debt of $32.071B against a corrected ~$30.55B, an overstatement of $1.52B, with the 10-Q's filing-verified cash of $8.3 billion and its repayments-only debt note establishing the direction; seg_prod omits the "Eliminations and other" line for 2024 and 2025, so the three segment lines sum to $91.155B against reported revenue of $88.603B (102.9%) — the 10-K supplies the mechanism verbatim ("Segment Total net sales... include intercompany sales and profit, which are ultimately eliminated within Eliminations and other") and the same vendor field DOES carry the contra line for 2023, where the four lines then tie to $68.920B exactly; the vendor's segment label "Raytheon Intelligence & Space" is SUPERSEDED — both filings state the three segments are "Collins, Pratt & Whitney, and Raytheon" — and "Collins Aerospace Systems" is corrected to the filings' "Collins Aerospace"; seg_geo for 2025 is a hybrid of the 10-K's by-ORIGIN and by-CUSTOMER-LOCATION tables and sums to 131% of revenue, with the $74,778M "domestic" line rejected and the United States figure derived as a $47,291M residual; seg_prod history before 2020 belongs to United Technologies (Otis, Carrier, Sikorsky) and is discarded entirely; lastDividend of $2.77 is the trailing-twelve-month total, not the last declared dividend, with the filings giving $0.68 declared 2026-02-06 and $0.73 declared 2026-04-30 and 2026-06-26 for a forward run-rate of $2.92 (1.340% yield); non-service pension income of $1,118M trailing was derived by us and verified against the 10-K's two disclosed year-on-year changes and the isolated $0.3B Q4 2025 Prudential settlement charge, producing a clean operating EPS of $5.01 against reported GAAP $5.68 and an honest trailing multiple of 43.5x rather than 38.4x; trailing EBITDA recomputed from the vendor's pension-inclusive $16.073B to a clean operating $14.885B, moving EV/EBITDA from 20.17x to 21.79x and net debt/EBITDA from 1.90x to 2.05x; quote.yearHigh/yearLow ($221.34/$150.61) rejected in favour of tech.hi52/lo52 ($218.58/$151.75) for internal consistency; and est.ebitdaAvg/ebitAvg verified against actuals as 4–6% low and excluded from all valuation. Where vendor and filing AGREED — worth recording, because these are the checks that failed elsewhere in this batch: the share count is exact (10-Q cover 1,347,758,144 at 2026-06-30 × $217.93 = $293,716,922,916 against the vendor's $293,716,922,951, agreement to 0.00000001%, with the 10-K cover's 1,342,287,676 and the 8-K's 1,345,974,220 forming a coherent rising series); currency is clean (reportedCurrency USD, profile USD, NYSE quote USD); short-term investments are genuinely $0 and the 10-Q's liquidity paragraph confirms it, so nothing needed adding back; enterprise value of $324.268B is correct and reconciles to derived total debt less the filing-verified cash to within $2 million; the segment magnitudes tie to the 10-K's disclosed organic growth by segment ($4.8B Pratt & Whitney, $2.6B Collins, $1.7B Raytheon); interest coverage of 6.26x and the 18.73% effective tax rate both reconcile; and the 2024 seg_geo row ties to reported revenue exactly at $80,738M.
• Basis note: analyst estimates and the earnings calendar are non-GAAP adjusted EPS; inc_a and inc_q are GAAP. The gap is large — 25.2% in 2025 ($6.211 consensus against $4.96 GAAP diluted) — and consists chiefly of acquisition accounting adjustments, which the 10-Q quantifies at "$0.4 billion, net of tax... an unfavorable impact on diluted EPS of $0.27" per quarter, plus restructuring. Forward multiples in this dive use consensus non-GAAP (30.1x 2026E, 27.6x 2027E, 24.9x 2028E); trailing multiples are shown on GAAP (38.4x) AND on a clean operating basis that strips non-service pension income (43.5x), and the difference between those two is a substantive conclusion of this dive, not a presentational choice. Note that consensus adjusted EPS INCLUDES the non-service pension income; the street does not strip it, and neither do we when computing forward multiples, so that the comparison is like-for-like.
• Estimate coverage: 16 analysts on 2026 revenue and 15 on 2026 EPS; 17 and 15 on 2027; 18 and 11 on 2028. The 2029 and 2030 EPS rows rest on FOUR analysts each and are excluded from every conclusion in this dive. The est EBITDA and EBIT rows were tested for the arithmetically impossible signature (negative EBITDA, EBIT above EBITDA, EBITDA equal to minus revenue) and passed on all three — this file is in the clean 66.5% — but were verified against actuals as 4–6% low and are not used.
• Peer note: the vendor peer set is unusually well constructed for this name — GE Aerospace ($391.5B), Boeing ($187.4B), Lockheed Martin ($136.0B), General Dynamics ($104.4B), Northrop Grumman ($78.4B), TransDigm ($71.3B), L3Harris ($53.1B), Leonardo DRS ($12.1B) — with only Caterpillar and Eaton as misfits. No peer multiples are supplied in the file, so no peer-multiple comparison is drawn. The 10-K publishes no named competitor list, describing competition only in general terms, so no filing-sourced substitute is available either.
• Fair-value caveat: the $163 / $227 / $273 anchors are multiples of the 2028 consensus adjusted-EPS distribution — 20x the low of $8.171, 26x the mean of $8.747, and 29x the high of $9.407 (11 analysts) — cross-checked against 2027E of $7.892 (15 analysts). Stated arithmetic, not a discounted cash flow. The base is deliberately sensitivity-disclosed: 24x gives $210, which is 3.6% BELOW spot, and 28x gives $245, above the street. The whole answer lives in a 24–28x band and no greater precision is claimed. The bear case assumes a de-rating to a normal industrial multiple and requires no operational failure; the bull case is the only anchor that requires multiple expansion, and that is stated rather than hidden.
• Timing: second-quarter 2026 results were released 2026-07-23, twelve days before this dive, and the 10-Q was filed the same day, so the June quarter is fully reflected in both the vendor data and the filing archive. The next print is 2026-10-20, 77 days away (consensus revenue $23.9997B, adjusted EPS $1.75) — there is no near-term company-specific catalyst, which is an explicit input to the Watch verdict rather than an oversight. 2026-08-04 was a broad semiconductor melt-up (ARM +17.36%, MRVL +12.81%, INTC +10.92%, LRCX +7.85%, MU +7.62%, AMD +7.00%, KLAC +6.95%, TXN +5.42%, PANW +5.53%); RTX's +0.59% carried no company-specific news and no sector beta from that event — it is one of the two flattest moves in the batch.
• Accessibility note: no information in this dive is conveyed by colour. All emphasis is carried by bold text, table structure and explicit labelling.
• Not investment advice. Independent research, educational and informational only, never personalised. No recommendation to buy, sell or hold any security is made to any person.
• Version: 2026-08-04-full.