ATI ATI

Reference table

1. What they actually sell

ATI reports two segments:

A disclosure caveat that must be stated: the two segments sum to $5.011B against consolidated revenue of $4.587B — a $423.2M gap, presumably intersegment eliminations that are not disclosed in this file. The segment split is therefore directional only and the percentages should not be treated as precise shares of revenue. We flag it rather than quietly normalising the numbers.

Within that caveat, the structure is clear: High Performance Materials & Components (the jet-engine and airframe alloys business) is the larger and the only growing half, up 5.5% in FY25 and up 46.6% since FY21. Advanced Alloys & Solutions — the more industrial, more commoditised half — shrank 0.5% in FY25 and is up only 33.1% from FY21. The mix shift toward the higher-value segment is a genuine positive and is one of the reasons operating margin has expanded.

Geography (FY25) — and this disclosure does foot exactly to $4.587B:

Two observations. The European exposure is substantial — the UK, France and Germany together are $750.8M, or 16.4% of revenue, consistent with a large European aerospace customer base. And China is $324.3M, 7.1% of revenue, and has been shrinking ($351.8M in FY22 → $324.3M in FY25). In a business where the adjacent knowledge-base lane is explicitly about critical-minerals decoupling from China, a 7.1% Chinese revenue exposure is worth knowing and is not discussed by the bulls.

2. The five-quarter stall — the central fact of this dive

Five quarters. Revenue up 0.6% from first to last. Over the same window the stock has risen 151%.

What IS working, honestly stated: operating margin has expanded every single quarter, from 13.2% to 14.8%. Adjusted EPS has grown from $0.72 to $1.00, a 39% increase. Every quarter beat its earnings estimate. The mix shift toward High Performance Materials is real and it is producing genuine margin gains.

What is NOT working: none of the earnings growth is coming from selling more. It is coming from (a) margin expansion, which is finite, and (b) share-count reduction — the diluted count fell from 146.8M (FY24) to 141.8M (FY25) to 139.4M (Q1 2026), a 5.0% reduction driven by $504.1M of buybacks in FY25 alone. On a flat revenue base, buying back 5% of the company mechanically produces 5% EPS growth. That is a legitimate use of capital; it is not demand.

And the most recent quarter beat on earnings while missing revenue by 3.0% ($1,151.5M against a $1,187.2M estimate). That is the signature of a company managing margin and share count into a demand environment that is not cooperating.

The FY25 full-year picture:

Note that FY25 operating margin actually declined slightly (14.0% → 13.8%) at the annual level even as it rose through the quarters — and that revenue growth has decelerated in each of the last three years (+37.0% → +8.8% → +4.5% → +5.2%). This is a business flattening out, not accelerating.

3. Balance sheet, cash flow, and the leverage comparison that matters

Balance sheet (FY25, 2025-12-28):

• Cash $416.7M (down sharply from $721.2M — a 42% decline, funding the buyback)
• Total debt $1.948B ($53.4M short-term, $1.806B long-term)
• Net debt $1.531B → 1.89x FY25 EBITDA of $810.4M (my computation; the vendor's key-metrics figure of 1.76x uses a TTM EBITDA base)
• Debt-to-equity 1.08x; equity $1.805B; current ratio 2.67; interest coverage 6.7x
• Goodwill $225.2M on $5.100B of assets — a real-asset business

The comparison that frames the whole valuation debate. Carpenter Technology (CRS) is covered in this same batch, is nearly identical in market capitalisation ($27.00B versus $26.36B), sells into the same end markets, and appears in ATI's own peer list:

ATI earns materially lower margins, carries five times the leverage, generates a lower return on capital, and has grown revenue by 0.6% against Carpenter's 12.6% — and trades at a HIGHER forward multiple. That is the single most damning fact in this dive, and it is derived entirely from the vendor's own files for the two companies. If an investor wants exposure to aerospace alloys, the arithmetic points at the other name, and our Carpenter dive rates that one only a Hold on price grounds.

Cash flow (FY25):

• Operating cash flow $614.3M (+50.9%)
• Capex -$280.6M (+17.4%, running 6.1% of revenue)
• Free cash flow $333.7M (+98.5%) — a genuine and large improvement over FY24's $168.1M
• Buybacks -$504.1M — more than 1.5 times free cash flow, funded by drawing cash down from $721.2M to $416.7M
• Dividends $0
• Net debt repayment -$182.6M

Free cash flow nearly doubled and that is a real positive. But it was entirely consumed, and then some, by a buyback 1.5 times its size, funded out of the cash balance. Buying back $504M of stock at an average price well below today's was good capital allocation in hindsight; continuing to do so at 43x forward earnings would be a different decision, and it is one worth watching. Free-cash-flow yield at today's price is 2.1%; price-to-free-cash-flow is 79.0x.

4. Data integrity — one clear contradiction, several thin-coverage warnings

REJECTED — the profile's dividend field. The company profile reports lastDividend: 0.32. The cash-flow statement reports commonDividendsPaid of $0 in FY22, FY23, FY24 and FY25 — four consecutive years — and ratios_ttm.dividendYieldTTM is 0. Three of four fields agree that ATI pays no dividend. We treat lastDividend: 0.32 as stale or erroneous and state that ATI pays no dividend. Any screen that credits this name with a dividend is wrong.

FLAGGED — the far-year estimate series is non-monotonic and effectively unsupported. Consensus EPS runs $4.446 (FY26E) → $5.424 (FY27E) → $6.367 (FY28E) → $7.317 (FY29E) → $7.050 (FY30E). Earnings decline between FY29 and FY30, while consensus net income also falls ($1,037.5M → $999.7M) on rising revenue ($6.221B → $6.422B). FY29 and FY30 each rest on 2 revenue analysts and 1 EPS analyst. These rows are noise and no valuation work in this dive uses them. FY28 (4 EPS analysts) is the last usable year, and even that is thin.

FLAGGED — the segment disclosure does not foot, as set out in section 1: segments sum to $5.011B against $4.587B of consolidated revenue.

FLAGGED — a share-count discontinuity in the 2024 quarters. Basic weighted shares run 142,958,481 (Q1 2024) → 124,400,000 (Q2) → 128,700,000 (Q3) → 142,000,000 (Q4) — a 13% drop and full recovery within one year, while diluted shares stay steady at 146-148M. That basic-share pattern is not economically plausible and the 2024 quarterly basic EPS figures inherit the problem. It does not affect FY25 or Q1 2026, which are internally consistent, and no valuation work here uses the 2024 quarterly basic series.

CHECKED AND CLEAN — the most recent quarter. inc_q[0] (Q1 2026, 2026-03-29) passes the standard screen: net income $118.2M against 139.4M diluted shares gives $0.848 versus the reported $0.85; share counts (135.9M basic / 139.4M diluted) do not duplicate the prior quarter's — well, in fact they do duplicate Q4 2025 exactly (135.9M / 139.4M). We note this as a minor duplication flag. Because the EPS reconciles correctly against net income at that count and the count is a plausible continuation of the declining trend, we use the row, but we disclose the duplication rather than ignoring it.

CHECKED AND CLEAN — the key-metrics block. enterpriseValueTTM of $27.781B against a market cap of $26.355B implies $1.426B of net debt, close to my computed $1.531B (the difference is timing). Return on equity of 24.2% and return on invested capital of 12.9% are consistent with the statements. Unlike several files in this batch, ATI's estimate EBITDA series is plausible — FY26E EBITDA of $808.8M against FY25 actual of $810.4M is a sensible flat forecast, not the 76%-to-infinite overstatements seen elsewhere. Credit where due.

5. Knowledge base — zero claims, and an adjacent lane that partly cuts the other way

ATI Inc. has ZERO tagged claims in the Synthos knowledge base. No tracked source has named this company. kb_claim_count: 0. This dive is explicitly fundamentals- and arithmetic-driven; the Avoid verdict rests on valuation and the operating record, not on borrowed expert conviction.

Four theme-adjacent claims describe the titanium and critical-minerals environment. Two support the demand backdrop:

• brent_johnson, 2025-11-09 (bullish, conviction 85): "Critical minerals and their refining are the focus because China dominates them; US self-sufficiency drive makes domestic producers of these materials prime beneficiaries."
• macrovoices, 2026-01-22 (bullish, conviction 76, entities: US Government, Department of War, Department of Energy): "Follow the state subsidies — DoW/DoE funding of scandium, gallium, rare-earth and titanium plays flags the companies that will flourish; invest in scarcity not reshoring itself."

One cuts the other way, and it deserves prominence because the bulls do not mention it:

• macrovoices, 2026-01-22 (bullish, conviction 75, entity: Iperion X): "IperionX makes low-cost titanium in Virginia using Zack Fang's University of Utah powder process at ~80% less cost — a reshoring winner."

That is a tracked source, at conviction 75, on the same date, arguing that a domestic competitor can produce titanium at roughly 80% lower cost via a fundamentally different process. If correct at commercial scale, it is a threat to the economics of every incumbent titanium melter, ATI included. We do not know the timeline or the scale, and we do not overstate it — but a knowledge-base lane that contains a specific, high-conviction claim about a low-cost substitute for your product is not a supportive lane, and presenting only the two favourable claims would be selective.

A fourth cluster (crux_investor, 2026-06-30, four claims on a large Australian titanium orebody, conviction 55-82) concerns titanium pigment feedstock rather than aerospace alloy and is only loosely adjacent; it does describe an 8-9 million tonne titanium dioxide market growing 3.4% annually needing 250-400kt of new supply per year, which is context rather than a claim about ATI.

Search-hygiene note. Matching on the bare three-letter string "ATI" against the knowledge base is worthless — it appears inside dozens of ordinary words. The claims above were located by matching on the company's business (titanium, specialty alloys, critical minerals) and on its former name, and every one of them was individually read before inclusion. The honest count of claims naming this company is zero.

6. Valuation — priced in or room?

At $193.12 (market cap $26.355B, net debt $1.531B, EV $27.886B):

Trailing P/E is computed on the sum of the last four reported diluted quarters ($0.70 + $0.78 + $0.69 + $0.85 = $3.02), giving 63.9x, which matches the vendor's 62.3x within rounding. FY29-FY30 estimates are rejected (section 4).

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

At $193.12 — 43.4x FY26E and 35.6x FY27E — the price embeds the following falsifiable claims:

• Revenue reaccelerates from zero to +8.5% in FY26 and +9.7% in FY27 (consensus, 6 analysts). This is the load-bearing assumption and nothing in the last five quarters of actuals supports it. Q1 2026 grew 0.6% and missed its estimate; to hit the FY26 consensus of $4.978B, the remaining three quarters must average $1.276B — 10.8% above the $1.152B just delivered. That is a large step-change, required immediately.
• EPS reaches ~$6.37 by FY28 (consensus, 4 analysts), i.e. 123% growth from FY25's $2.85 in three years, on revenue growth of 27% over the same period. The gap must be filled by margin expansion and buybacks.
• Operating margin expands beyond 14.8% — the Q1 2026 level, itself the high point of a five-quarter expansion.
• The buyback continues at scale. FY25's $504.1M repurchase exceeded free cash flow by 51% and was funded from the cash balance, which fell 42% to $416.7M. Repeating it requires either higher free cash flow or more debt against an already 1.89x-levered balance sheet.
• The market keeps paying above 35x forward earnings for a cyclical specialty-metals fabricator — despite spot already sitting above the consensus target, and despite Carpenter Technology offering higher margins and one-fifth the leverage at a comparable multiple.

The single most fragile assumption: the revenue reacceleration. Everything else follows from it. The 2026-08-06 print is the first direct test — consensus wants $1.216B, which would be +6.6% year over year.

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

Expected return ≈ EPS growth + multiple drift + shareholder yield.

• EPS growth: consensus +56% (FY26E) and +22% (FY27E). Strong on paper, contingent on the revenue reacceleration above.
• Multiple drift: our base case assumes meaningful COMPRESSION — 28x applied to FY27E against the 43.4x the market pays on FY26E today. The reasons are specific and we state them rather than asserting a number: (i) spot already trades above the sell-side's own mean and median targets, so the marginal buyer at this price is buying past where the informed consensus stops; (ii) a specialty-metals fabricator with 13.8% operating margins and 1.89x leverage has historically not commanded a 35x+ multiple, and this company's operating margin was 4.2% as recently as FY21 and its net income was -$1.573B in FY20; (iii) the direct comparable earns 63% higher operating margins at one-fifth the leverage and trades at a similar multiple, which implies one of the two is mispriced; (iv) a 2.1% free-cash-flow yield gives the multiple no cash support. The roll from 43.4x FY26E to 35.6x FY27E at a constant price is mechanical, not re-rating, and we do not count it.
• Shareholder yield: no dividend (see section 4). Buybacks of $504.1M are 1.9% of market cap — real, but funded partly from the cash balance rather than from free cash flow.

The bridge reads: strong forecast earnings growth, more than offset by assumed multiple compression, with a ~1.9% buyback yield — netting to approximately -21% over the horizon. That is our base case. If instead the multiple simply held at 43x FY26E, fair value would be ~$236 and this would be a Buy — we disclose that sensitivity explicitly, because it is the honest statement of what separates our view from the bull case. We do not believe a flat-revenue metals fabricator holds a 43x multiple, and we say why.

Scenario bands (stated arithmetic, no DCF):

• Bear ~$108 — 20x FY27E EPS of $5.424. The revenue reacceleration does not arrive, aerospace build rates disappoint, margin expansion stalls, and the multiple reverts toward a normal specialty-metals range. -44% from spot. Note this would still leave the stock 51% above its 52-week low.
• Base ~$152 — 28x FY27E EPS of $5.424. Revenue grows in the high single digits roughly as consensus expects, margins hold near 14.8%, and the multiple compresses toward something defensible for the asset class. -21% from spot.
• Bull ~$204 — 32x FY28E EPS of $6.367. Aerospace demand inflects hard, revenue growth exceeds 10%, operating margin pushes past 16%, and the market maintains a premium multiple. +6% from spot. This is roughly the all-time high and just below the $215 street high.

Base sits 21% BELOW spot; the bull case offers +6%. A bull case that barely clears the current price is the clearest possible statement that the risk/reward is inverted.

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

• We differ from the street on the multiple, not on the forecast. Our revenue and EPS path is consensus. Our variant perception is that 43x forward earnings is the wrong multiple for a business with 13.8% operating margins, 1.89x leverage, a 2.1% free-cash-flow yield and five quarters of flat revenue — and that the appropriate comparison is not to itself twelve months ago but to Carpenter Technology, which delivers materially better economics at a similar price. This is a relative-value call inside a single theme, and it is the sharpest one available in this batch.
• We also weight the insider tape, which the street does not. Twenty buy ratings, a $191.00 target — and the Chair, President and Chief Executive reduced her disclosed holding by roughly 36% in three weeks (section 7). We do not treat that as decisive on its own, but combined with a price above the consensus target it is the second independent signal that the informed marginal seller is at these levels.
• Positive surprise that would force a repricing (and would flip our verdict): the 2026-08-06 print delivering revenue above $1.24B (+8.7% year over year) — comfortably ahead of the $1.216B consensus and decisively breaking the five-quarter stall — with operating margin above 15% and management raising full-year guidance. That would validate the reacceleration the price already assumes and would justify a materially higher fair value. It is watchable, dated, and two days away.
• Negative surprise: revenue at or below $1.16B (confirming the stall); a second consecutive revenue miss; free cash flow declining as capex rises further; the buyback slowing (removing the mechanism currently generating much of the EPS growth); or further large chief-executive selling.

7. Insider activity — the chief executive sold roughly a third of her stake

The Form 4 record is eight consecutive sales by one person, with zero purchases by anyone:

Totals: approximately 70,769 shares for roughly $13.2M across three weeks. The disclosed holding falls from about 197,321 shares (before the first listed sale) to 125,564 — a reduction of approximately 36%.

How to read it, fairly. The 7 July sequence — seven tranches at incrementally rising prices within a single session — has the appearance of pre-arranged plan execution rather than a discretionary decision. Executives diversify, and Fields retains 125,564 shares worth roughly $24.2M at spot, so she remains meaningfully exposed.

Why it still matters here. Three things distinguish this from routine selling. First, the magnitude: roughly 36% of a disclosed holding in three weeks is not incidental trimming. Second, the 28 July transaction of 31,757 shares at $190.99 is a separate, later, single-block sale at the highest price in the sequence — three weeks after the plan tranches and eight days before an earnings print. Third, there is not a single purchase anywhere in the record, by Fields or by any other officer or director, in a year when the stock rose 151%.

Read: a genuine negative signal, and one of the three legs of the Avoid verdict. It is not, on its own, a reason to sell a good business. Combined with a price above the sell-side's own target and five quarters of flat revenue, it is the third independent indication that the marginal informed participant is a seller at these levels.

8. Moat and competitive position

ATI's moat is the same structural one Carpenter Technology enjoys: application-specific qualification. Alloys used in rotating jet-engine components and critical airframe structures must be individually approved for each part, a process measured in years and involving the engine manufacturer and the regulator. Once qualified, a supplier is embedded for the life of the programme. That is why a metals fabricator can earn a 13.8% operating margin and a 24.2% return on equity.

Evidence the moat works: operating margin expanded from 4.2% (FY21) to 14.8% (Q1 2026); return on invested capital of 12.9%; a High Performance Materials segment that has grown 46.6% since FY21.

Three things the moat does not do. It does not create demand — qualification guarantees ATI's share of a programme, not the programme's size, which is exactly why revenue has been flat for five quarters while margins expanded. It does not protect against a process substitution — the macrovoices claim about a domestic producer making titanium at roughly 80% lower cost via a powder process is precisely the kind of development qualification does not defend against, because a genuinely superior cost structure eventually gets qualified too. And it did not prevent a $1.573B net loss in FY20, five years ago, which remains the clearest available statement of what this business looks like when the aerospace cycle turns.

Assessment: a real qualification moat on a cyclical demand base, currently priced as though the cycle does not exist.

9. Verdict, kill-criteria and flip conditions

Avoid at this price. To be explicit about what this verdict is and is not: it is a judgement on the stock, not a criticism of the company. ATI runs 13.8% operating margins, earns a 24.2% return on equity, generated $333.7M of free cash flow, holds a genuine qualification moat in aerospace alloys, and has expanded operating margin in five consecutive quarters. It is a sound business.

It is also a stock that has risen 151.0% in twelve months against revenue that grew 0.6% over the same five quarters, that trades at 43.4x FY26E and 35.6x FY27E earnings on a 2.1% free-cash-flow yield with 1.89x net-debt/EBITDA, that sits above both the mean ($191.00) and the median ($187.00) sell-side price target, and whose chief executive reduced her disclosed holding by roughly 36% in the three weeks before this dive. Our base fair value of $152 is 21% below spot, and our bull case of $204 offers only +6%.

When the bull case is +6% and the bear case is -44%, the answer is not to own it.

Pre-registered FLIP conditions (any TWO would move this from Avoid to Watch; all three to Stage-In):

• Revenue above $1.24B on 2026-08-06 (+8.7% year over year), decisively breaking the five-quarter stall, with full-year guidance raised.
• A price below ~$150 (roughly 28x FY27E, our base case) absent fundamental deterioration — at which the reacceleration is no longer pre-paid.
• Insider PURCHASES, or at minimum a complete cessation of chief-executive selling for two full quarters.

Pre-registered further-DOWNGRADE criteria (would move Avoid to an explicit warning for holders):

• A second consecutive revenue miss on 2026-08-06.
• Operating margin contracting below 14% — removing the mechanism that has produced all of the earnings growth.
• The buyback slowing or stopping, which would expose the flat revenue line directly in EPS.
• Net-debt/EBITDA rising above 2.25x, particularly if driven by debt-funded buybacks at 43x earnings.
• Any evidence that low-cost titanium substitution (the macrovoices claim) is reaching commercial qualification.

Where ATI fits in the Synthos Framework Portfolio. It does not, at this price. If the industrial / aerospace-supply-chain sleeve is to be held, the vehicle is Carpenter Technology — higher margins (22.5% versus 13.8%), one-fifth the leverage (0.35x versus 1.89x), a higher return on capital (17.1% versus 12.9%), and 12.6% revenue growth against ATI's 0.6% — and our Carpenter dive rates even that name only a Hold on valuation grounds. If the better business in the same theme is only a Hold, the weaker one at a higher multiple is an Avoid. Logged as a tracked Synthos call (Avoid) as of 2026-08-04 at $193.12.

Single biggest risk (for anyone who owns it): that the five-quarter revenue stall is the signal and the 151% advance is the noise. At 43x forward earnings with 34% of air between spot and the 200-DMA, and no dividend and only a 2.1% free-cash-flow yield underneath, a single disappointing guide has a very long way to fall.

Most fragile assumption in the price: that revenue reaccelerates from 0.6% to 8.5% this year. To reach the FY26 consensus of $4.978B, the three remaining quarters must average $1.276B — 10.8% above the quarter just delivered, which itself missed its estimate. Everything in the bull case is downstream of a step-change that has not yet appeared in a single reported quarter.

Provenance & disclosures

• Traceability: ZERO tagged KB claims name ATI Inc. (kb_claim_count: 0, breadth 0). This dive is explicitly fundamentals- and arithmetic-driven — an accepted, precedented outcome, and never a reason to invent conviction. Four theme-adjacent claims are cited and labelled as adjacent: brent_johnson 2025-11-09 (conviction 85, critical-minerals self-sufficiency, supportive); macrovoices 2026-01-22 (conviction 76, following state subsidies into titanium, supportive); macrovoices 2026-01-22 (conviction 75, a domestic competitor producing titanium at roughly 80% lower cost — a claim that cuts AGAINST incumbents and is presented as such); and a crux_investor cluster on titanium-dioxide feedstock supply, only loosely relevant. A bare-string search on "ATI" is worthless (the sequence appears inside dozens of ordinary words); claims were located by business description and read individually.
• Data as-of: fundamentals 2026-03-29 (Q1 2026; fiscal year ends late December) · estimates 2026-08-04 · prices 2026-08-04 ($193.12, 50-DMA $187.73, 200-DMA $143.83, RSI 52.7) · KB claims 2026-08-04. Sole market-data source: the Synthos vendor file for this ticker.
• Data REJECTED: (1) the profile's lastDividend: 0.32 — contradicted by four consecutive years of $0 dividends paid in the cash-flow statement and by a TTM dividend yield of 0. ATI pays no dividend. (2) The FY29 and FY30 estimate rows — EPS declines from $7.317 to $7.050 on rising revenue, and each row rests on 2 revenue analysts and 1 EPS analyst. No valuation work uses them.
• Data FLAGGED (used with caveats): (1) the segment disclosure does not foot — the two segments sum to $5.011B against $4.587B of consolidated revenue, a $423.2M undisclosed elimination; segment shares are directional only. (2) The 2024 quarterly basic share counts swing from 143.0M to 124.4M to 128.7M to 142.0M within one year while diluted stays at 146-148M — not economically plausible; no work here uses the 2024 quarterly basic series. (3) Q1 2026 basic and diluted share counts (135.9M / 139.4M) exactly duplicate Q4 2025's — a minor duplication flag, disclosed; the row is used because EPS reconciles correctly against net income at that count and the count continues a plausible declining trend.
• Data VALIDATED: the key-metrics block is internally consistent (enterprise value ≈ market cap + net debt; return on equity 24.2% and return on invested capital 12.9% tie to the statements); the geographic disclosure foots exactly to $4.587B; and, unlike several other files in this batch, the consensus EBITDA series is plausible (FY26E $808.8M against FY25 actual $810.4M) rather than grossly overstated.
• Fair-value caveat: the $108 / $152 / $204 anchors are scenario multiples applied to consensus EPS, not a discounted cash-flow model. The base case explicitly assumes multiple compression from 43.4x FY26E to 28x FY27E, for the four reasons set out in section 6b. If the multiple instead simply held at 43x FY26E, fair value would be approximately $236 and the verdict would be Buy — that sensitivity is disclosed rather than hidden, because it is precisely what separates this view from the bull case.
• Comparison basis: the ATI-versus-Carpenter Technology table in section 3 is constructed entirely from the two companies' own vendor files in this same batch, on identical definitions, as of the same date.
• Not investment advice. Independent research, educational and informational only, never personalised.
• Version: 2026-08-04-full.