荏原製作所 Ebara Corporation (6361)

Signal panel

Investment thesis

1. Orders are running well ahead of sales. H1 FY2026 orders were ¥636.3bn (+41.0% YoY) against sales of ¥490.8bn. Full-year order guidance was raised on 2026-08-14 from ¥1.07tn to ¥1.245tn (+16.4%) (Q2 短信 p.4, p.9; reported fact and company guidance). Precision orders guidance was raised to ¥550bn against sales of ¥425bn, a book-to-bill of 1.29. Equipment orders rose 116% in H1, driven by CMP for logic and foundry and by memory demand for AI (Q2 短信 p.24).
2. The backlog supports 2027. Group backlog was ¥1,059.2bn at 2Q (+17.6% YoY), and Precision backlog was ¥233.5bn (+31.5%) (Q2 短信 p.23). 2027 consensus (IFIS, 2026-10-16) is OP ¥172.6bn (+38% on 2026 guidance) and NP ¥117.2bn. The company itself now expects WFE growth of 18% or more in 2026, up from 10% or more (Q2 説明会 p.21).
3. Profit is not yet keeping pace. Q2 standalone OP fell 12.9% YoY to ¥23.9bn, and OPM fell to 9.8% from 11.6% (zaimu_4q_all). Energy lost ¥0.8bn in H1 against a ¥11.2bn profit a year earlier, because of Middle East shipment delays, hydrogen costs and operating losses at the Jeannette (US) plant (Q2 説明会 p.13, p.22). Full-year OP guidance stayed at ¥125.0bn despite the higher sales. The flat OP also absorbs a pension-plan settlement loss of undisclosed size (Q2 短信 p.9).
4. Headline EPS is flattered. The 2026 NP guidance rose from ¥86.6bn to ¥99.5bn on 2026-05-15 because of the 水ing sale. That sale brings a ¥17.2bn gain on sale plus ¥3.4bn of financial income, both booked in Q3 (Q2 短信 p.20). Underlying 2026 EPS is about ¥190 (the initial guidance EPS of ¥189.67; FY 説明会 p.19). On that basis the PER is 22.1x, not the 19.3x on the site (inference).
5. The price already embeds the sector scare. On 2026-10-15 the stock fell 10.6% (¥4,626 to ¥4,138), when the Nikkei fell 3,318 yen on AI and semiconductor selling (jiji.com, 2026-10-15). Before that, the stock fell about 9% after the Q2 results, on doubts that the orders would turn into margin. At ¥4,200 the price equals band-mean PER (21.2x) times EPS of ¥198, which is roughly zero growth on underlying 2026 EPS (inference).

Business and earnings

Ebara sells pumps, compressors, turbines, waste-incineration plants, and semiconductor vacuum pumps, CMP and plating tools. Customers are chip makers (Precision & Electronics), oil, gas, LNG and power operators (Energy), building and industrial contractors (Building & Industrial), public water and flood-control agencies (Infrastructure), and Japanese municipalities (Environmental). No single customer is 10% or more of sales (有報 p.142).

Profit engine (FY2025, segment profit and margin; 有報 p.18–19, factsheet):

Precision is 36% of sales but 51% of FY2025 segment OP and 62% of the 2026 plan. It is a cyclical business tied to WFE. About 65% of its H1 sales go to Asia outside Japan, and China is held "at last year's level" (Q2 説明会 p.10, p.34). Service and support provides a cushion: S&S is 29% of Precision sales and 55% of Energy sales (p.35). The installed base gives Ebara switching costs in dry vacuum pumps, and it is adding overhaul sites in Taiwan and Korea in 2026 (p.25).

H1 FY2026 OP bridge (¥bn, Q2 説明会 p.7): ¥50.0 to ¥50.7. Volume +10.7, profitability +10.2, fixed costs −23.8 (wages, depreciation up 28% to ¥20.4bn), FX +2.8, other +0.9. Fixed-cost growth is absorbing almost all of the gross gain. This follows the investment cycle: capex rose from ¥58.6bn in FY2024 to ¥100.7bn in FY2025, 2.9x depreciation (zaimu_ive).

Regions (H1 by customer location, Q2 短信 p.25): Taiwan, Korea and other Asia +45.8% to ¥124.9bn, Japan +9.8%, China −2.7% to ¥73.8bn (15%), North America −16.2%, Middle East −26.0%. Overseas sales are 66% of the total. FX assumption: ¥145/USD against an H1 actual of ¥159.3 (Q2 説明会 p.5). Guidance therefore carries an FX cushion if the yen stays weak (inference; the company gives no sensitivity).

Forecast test

• Seasonality, like-for-like. The fact sheet shows OP progress of 40.5% against a 40.0% median, a delta of +0.5 pts. That median divides past H1 OP by the final actual. This year's figure divides by current guidance. Ebara beats guidance most years, so the comparison is not like-for-like. Dividing past H1 OP by the guidance in force after each Q2 gives 2022 40.3%, 2023 47.6%, 2024 45.9% and 2025 48.8%, a median of 46.8%. 2026 at 40.5% is therefore −6.2 pts behind. Sales are 46.6% against a 48.2% median, −1.6 pts. Pretax and NP progress (−6.0 and −4.6 pts in the fact sheet) are distorted by the ¥20.6bn 水ing gain, which lands in Q3, so ignore them.
• Track record. Actual OP beat initial guidance in all of 2020–2025: by +39%, +43%, +7%, +21%, +13% and +12%. The beat is a habit, so the market expects it. The closest analogue to 2026 is 2022: H1 progress was 40.3%, and the final beat over Q2 guidance was only 4.6% (¥67.5bn to ¥70.6bn).
• H2 hurdle. Guidance implies H2 OP of ¥74.3bn, +16% YoY. That requires Precision at ¥44.4bn (+29%) and Energy at ¥18.7bn, against ¥14.8bn in H2 FY2025 and a ¥0.7bn loss in H1 FY2026 (Q2 説明会 p.29). The company points to Middle East S&S shipments that slipped into H2, the Jeannette losses ending, and backlog (p.22). The Energy recovery is the main risk to guidance.
• Leading indicators. Precision orders ran ¥151.2bn in Q1 and ¥116.2bn in Q2. Equipment orders were ¥114.6bn and then ¥71.5bn (p.11). The pace slowed from Q1, but Q2 is still almost 2x Q2 FY2025.
• Read. Expect a sales and orders beat (FX at ¥145, plus backlog) and OP roughly in line to +5%, which is about what consensus at ¥130.7bn already has (inference). An OP revision is unlikely before Q3. Revision probability rises if Q3 Precision OPM holds at 18% or more and Energy turns profitable.

Financial strength and capital allocation

• Balance sheet (2Q FY2026, Q2 説明会 p.36). Interest-bearing debt was ¥228.2bn and cash ¥153.0bn, so net debt was ¥75bn. D/E was 0.43, against a policy band of 0.4–0.5 with an "A" rating as a precondition (有報 p.22). Owners' equity ratio was 48.3%. The FY2025 current ratio was 1.60 (factsheet). The 水ing proceeds (¥25.0bn, plus a ¥5.3bn special dividend) arrived on 2026-07-01 (Q2 短信 p.20).
• Cash conversion. FY2025 OCF was only ¥40.8bn (OCF/NP 0.53), held back by receivables (+24% against sales +11%, an amber flag), lower contract liabilities and lower payables (FY 短信 p.10). FCF was −¥50.5bn, after ¥92.2bn of capex. The swing reverses in H1 FY2026: OCF ¥75.5bn against ¥54.7bn a year earlier, and FCF ¥30.7bn. The FY2026 plan is OCF ¥100bn and FCF ¥45bn (Q2 説明会 p.37). OCF was positive in every year from 2020 to 2025, so the gate passes.
• Capital efficiency. ROE was 14.5%, 15.0%, 15.7%, 16.2% and 15.6% in 2021–2025. Decomposed on year-end figures, FY2025 is net margin 8.0% × asset turnover 0.89 × leverage 2.13. FY2021 was 7.2% × 0.84 × 2.30. The gains came from margin and turnover, not leverage. Company ROIC was 11.9% in FY2025, against a company WACC of 5–6% (有報 p.22). Against an analyst cost of equity of 8–9%, ROE clears by about 7 pts.
• E-Plan2028 targets (有報 p.22–23). By FY2028: ROIC ≥13% (company WACC assumption 8–9%), ROE ≥18%, OPM ≥14.5%, sales of about ¥1.2tn, Precision OPM ≥20%. Management told MONOist (2026-08-19) that the ¥1.2tn sales target is still uncertain. Executive share awards are tied to FY2028 ROIC (有報 p.111).
• Peers (latest FY to March 2026, IRBANK; multiples from Yahoo! Finance, 2026-10-16):

Ebara sits between the pure semiconductor names and the pump makers on returns and on multiples. That fits a mixed portfolio with about 60% of profit from semiconductors.

• Returns. The policy is a payout ratio of 35% or more, plus buybacks so that 3-year shareholder returns equal 100% or more of FCF excluding asset sales (有報 p.22). DPS has risen 6 years in a row, to ¥66 planned for 2026 (yield 1.6%). The FY2025 buyback was ¥20.0bn (FY 説明会 p.22; FY 短信 p.10), so the FY2025 total payout was 61%. The fact sheet shows 74.5% because the IFIS feed puts ¥30.0bn in FY2025. The FY2026 program of up to ¥10bn ran to 2026-07-31, and ¥5.7bn had been executed by 2026-06-30 (Q2 短信 p.17). 5.0M treasury shares were cancelled on 2026-02-27.
• Cross-shareholdings. None. Ebara holds no listed 特定投資株式, and only ¥1.2bn of unlisted equity (有報 p.116).
• Gates. The accounting-integrity gate passes: the audit KAM is the routine cost estimate on long-term contracts (有報 p.219). Liquidity passes, and there is no going-concern note (Q2 短信 p.18). OCF passes. The AI exposure claim passes, because it is backed by order figures. Contingent claims to watch: the Naphtachimie fire suit (no provision possible), the Kirloskar arbitration in India (no provision possible), and the Gifu case, sent back to the High Court on 2026-07-07 (described as immaterial) (Q2 短信 p.20–21). The JFTC issued a 下請法 recommendation on 2025-02-20 (有報 p.33). That is a compliance matter, not an accounting one.

IR and management credibility

• 2026-02-13. The company published E-Vision2035 and E-Plan2028, with a WACC disclosed by segment. This exceeds the TSE cost-of-capital request (有報 p.18–24).
• 2026-07-14. IR Day held on the Precision, Energy and Building strategies (Q2 説明会 p.40).
• 2026-08-14. The Middle East impact was quantified by segment (−¥2.2bn in H1, −¥2.3bn for the full year), and the Kumamoto quake (2026-07-28) was flagged as not yet assessed (Q2 説明会 p.3, p.28).
• Promises against outcomes. E-Plan2025 was met at the group level: OPM 11.9% against ≥10%, ROIC 11.9% against ≥10%, ROE 15.6% against ≥15%, sales CAGR 12.1% against ≥7%. It was narrowly missed in Precision (OPM 16.9% against ≥17%) and Building (6.3% against ≥7%) (有報 p.18–19).
• KPI changes to watch. From Q1 FY2026, the separate CMP sub-segment was merged into "装置 (equipment)", and part of hydrogen moved into Energy (Q2 短信 p.19, p.24). The CMP line can no longer be tracked on its own. The past series were restated, so this is not a red flag, but it reduces visibility.
• The 説明会 was corrected on 2024-06-21 and 2025-03-31, and the 有報 on 2024-04-05, 2024-06-27 and 2025-04-03 (IFIS 決算書 matrix). The contents were not reviewed. None led to a restatement of results that this review found.

Valuation and scenarios

FY2027 EPS uses 456.1M shares (Q2 短信 p.2).

• Bands. Forward PER is 16.9x against a one-year mean of 21.2x, z −1.47. PBR is 3.60x against a mean of 4.48x, z −1.40 (factsheet). EV/EBITDA is 12.3x on forecast EBITDA (stock_chart_eveb). IFIS's theoretical price is ¥5,263 at mean PER. That is a reference point, not a target.
• What the price implies. ¥4,200 is 22x underlying 2026 EPS and 16.3x 2027 consensus. At the band-mean PER of 21.2x, it needs EPS of only ¥198. The market is therefore pricing roughly no growth beyond 2026, or a de-rating of the semiconductor cycle (inference).
• Why the market may be right. Consensus 2026 OP has fallen 3.4% in three months. Two brokers moved to neutral, on 2026-08-26 (target ¥5,410) and 2026-09-08 (target ¥4,500) (IFIS report_topix). Q2 OPM fell, and fixed costs are rising faster than gross profit. AI capex sentiment broke on 2026-10-15. Weighting the cases 25/50/25 gives about +9%, which is not enough for a Buy given the pending Q3 test.
• Consensus target. ¥6,426, rating 4.55 out of 5 (IFIS, 2026-10-16). This is an external estimate, not fair value.

Bull case

• Precision book-to-bill of 1.29 and backlog of ¥233.5bn (+31.5%) lift 2027 Precision sales above 20% growth, with operating leverage on capacity added in 2025 (Q2 短信 p.23).
• Energy H2 recovers to ¥18.7bn as Middle East S&S is booked and the Jeannette losses end. That removes the main drag on the group margin (Q2 説明会 p.22).
• FX: the H1 average was ¥159/USD against a ¥145 assumption, which leaves room for a sales and OP beat in Q3 and Q4.
• After the gain on sale, ¥25bn of cash supports buybacks toward the "100% of FCF" policy.

Bear case

• AI capex digestion. WFE growth guidance moved from 10%+ to 18%+ in 2026, which puts 2027 at risk of a peak-year comparison. Q2 Precision orders already fell from ¥151bn to ¥116bn quarter on quarter (Q2 説明会 p.11).
• The fixed-cost build (D&A +28% in H1, R&D ¥25bn) compresses margins if volumes stall. Q2 OPM was 9.8%.
• Energy: hydrogen costs, Middle East exposure (¥14.1bn of H1 Energy sales) and US plant execution.
• Taiwan, Korea and other Asia is 26% of H1 sales and China 15%. Export controls and customer concentration in leading-edge foundry are uncertainties (有報 p.36, inference).

Catalysts

What would change the rating

• Upgrade to Buy if Q3 shows Precision OPM of 18% or more, Energy back in profit, nine-month OP progress of 66% or more of ¥125bn (the 2025 nine-month figure was 61% of actual), and Q3 orders still at or above ¥250bn.
• Downgrade to Sell if Q3 orders fall below the year-earlier level, OP guidance is cut, or 2027 consensus OP falls below ¥155bn.
• Thesis invalidated if Precision orders turn negative YoY, which would mean the backlog build was pull-forward rather than a cycle.

Monitoring

Scorecard

Fact-sheet flags

Computed by factsheet.mjs. Each flag is a prompt to investigate, not a conclusion.

• ▲amberreceivables-outpacing-sales receivables grew 24.3% in 2025/12 (171,164 to 212,775) against sales growth of 10.6%
• ▲amberprogress-behind-rp 2026/12 Q2 rp progress 35.0% vs median 46.7% (-11.6 pts): behind seasonal norm
• iinfoguidance-revised-up 2026/12 guidance raised on 2026-08-14: sales 1,020,000 to 1,053,000
• ▲ambernon-operating-in-guidance 2026/12 guidance pretax profit (税引前利益) 142,000 exceeds OP 125,000 by 13.6% of OP: RP, NP, EPS, ROE and forecast PER include non-operating items; check the 短信 for one-offs
• ▲amberfcf-negative Free cash flow -50,477 in 2025/12 (OCF 40,755, ICF -91,232, capex 100,735)
• iinfocapex-heavy Capex 100,735 is 2.89x depreciation 34,804 in 2025/12
• ▲amberconsensus-falling 2026/12 consensus OP changed -3.4% over 3 months to 130,673
• iinfoper-cheap Forward PER 16.91 vs 1-year mean 21.18 (+1σ 24.09), z -1.47: cheap vs own history
• iinfofar-below-52w-high Price 4,200 is -39.0% from the 52-week high 6,890

Sources and limitations

• filings/2025-12_FY_短信_20260213.pdf p.1, p.10 (2026-02-13)
• filings/2025-12_FY_説明会_20260213.pdf p.3, p.5, p.19–22 (2026-02-13)
• filings/2025-12_FY_有報_20260323.pdf p.18–24 (E-Vision2035 and E-Plan2028), p.33–36 (risks), p.111, p.116, p.142, p.219 (2026-03-23)
• filings/2026-12_Q2_短信_20260814.pdf p.1–2, p.4–10, p.17–25 (2026-08-14)
• filings/2026-12_Q2_説明会_20260814.pdf p.3–38 (2026-08-14)
• filings/2026-12_Q2_半期_20260814.pdf p.3 (no new risks) (2026-08-14)
• IFIS snapshot data/6361/2026-10-17/ feeds (zaimu_fy_all, zaimu_4q_all, zaimu_4q_acc, act_est_hist_n, cons_chart, stock_price, shareholder_return), pages/report_topix.html, and the fact sheet data/6361/analysis/2026-10-17/factsheet.md
• MONOist, Ebara H1 results article, 2026-08-19: https://monoist.itmedia.co.jp/mn/articles/2608/19/news023.html
• 時事ドットコム on the 2026-10-15 Tokyo semiconductor sell-off: https://www.jiji.com/jc/article?k=2026101500547&g=eco
• IRBANK fy-data-all.csv for 8035, 7735, 6363 and 6351 (FY to 2026/03), and Yahoo! Finance quotes for 8035, 7735, 6363, 6351 and 6361 (2026-10-16)
• Limitations. TDnet and the Ebara IR site returned HTTP 403. The IFIS 適時開示 list shows no Ebara disclosure after 2026-08-14, and none was found by search, but this is not confirmed against TDnet. The Q3 date is estimated. The pension-settlement loss amount is not disclosed. Peer PERs mix bases (TEL trailing, the others company forecast). The IFIS FY2025 buyback (¥30.0bn) conflicts with the filings (¥20.0bn), and the filings were used.
• This is research, not individualised investment advice.