RWS Holdings: the price needs 78% cash conversion, the company aims for 65%

RWS made £26.6 million of cash for its owners in the year to September 2025, after capital spending, interest, leases and stock pay. The number the company headlines, operational free cash flow, was £80.1 million. The market value is about £563 million, and the share price has risen about 64% from the September 2025 year-end close of 92.3p. A screen shows the stock at 9.5 times next year's adjusted earnings, which is cheap if those earnings are cash and not cheap if they are not.

At 151.77p, a delayed trade from 12:31 London time on October 19, 2026, the shares trade 19.1% above my triangulated Base value of 122.76p. The Bear value of 68.58p is 54.8% below the price and the Bull value of 199.49p is 31.4% above it. Confidence is grade C (13/25), the required margin of safety (MOS) is 45%, and the maximum buy price is 67.52p. The latest verified statements are the unaudited half-year results to March 31, 2026, published June 11. The full-year results to September 30, 2026 had not been published when I checked, so the 2026 figures here are my estimates.

Valuation route: technology-enabled language, content and IP services in transition. Primary method: owner cash at a 9 times multiple in September 2028 with two forward years of cash, discounted at 11%. Secondary method: a sum of the parts built on cash-adjusted operating profit for each business unit. Classification: standard company, 4 of 12 on the scorecard. Share prices and per-share values are in pence, company figures in pounds sterling millions. Values are present fair value estimates, not 12 month price targets.

What RWS is

RWS translates, localizes and protects content for large companies. It also sells software for managing that work, and since 2026 it reports three units. In the half year to March 2026:

Transform is still 58% of sales and fell 4.6% on an organic constant currency basis. Generate grew 52.0%, but the company says the figure includes an exceptional product-launch program for one strategic client, so it is not yet a repeating run rate. AI-related revenue was 32% of the group against 26% a year earlier, and net repeat revenue from the top 100 services clients was 109%. That last number covers a selected cohort of services clients. It is not software subscription retention, and I do not treat it as such. No single customer exceeded 10% of sales in fiscal 2025.

In 2026 the company bought two things. Obviously, a brand and IP protection platform, was acquired on May 5 for £16.5 million in cash plus up to £23.5 million of deferred payments tied to profit targets. Acogroup, the owner of the Acolad localization business, was agreed on August 3 for £40.2 million, of which about £17.8 million is cash on its balance sheet, an enterprise value of £22.4 million. The company expects about £155 million of revenue and £11 million of adjusted EBITDA from it in the year to September 2027, and expects to complete in the first half of that year, subject to French regulatory clearance. I assume completion on January 1, 2027.

The latest results

Revenue grew 5% in the half year and adjusted profit improved. The year before was much weaker: statutory results fell into a loss of £99.7 million after an £88.0 million goodwill impairment, and the total dividend was cut by 43%. The company says it continues to trade in line with expectations for 2026, with mid-single-digit organic revenue growth and improving profitability. Operational free cash flow in the half year was £30.4 million, 67% of adjusted EBITDA.

Which cash to value

RWS publishes two different cash measures, and they should not be mixed up. Operational free cash flow (£80.1 million in 2025) starts from adjusted EBITDA and is before interest, tax and the company's own exceptional costs. Its adjusted free cash flow (£56.1 million) adds back £25.7 million of exceptional cash costs. I value what I call owner cash: operating cash flow after tax, less all capital spending including capitalized software, less net interest, less all lease payments, less the stock pay that the cash flow statement adds back.

Owner cash was 47.4% of the company's adjusted free cash flow and 33.2% of operational free cash flow. The same number comes out from the other side: 56.1 - 25.7 - 3.8 = 26.6.

The FY2022 charges include a £1.5 million refinancing fee. Net interest ran from £0.6 million to £4.2 million, lease payments from £8.6 million to £13.1 million and stock pay from £1.4 million to £3.8 million. The three-year average of owner cash is £31.4 million and the five-year average £44.0 million. I treat these as cross-checks, not as a level that must return. In the half year to March 2026, owner cash was only £7.9 million, 26% of operational free cash flow, held back by the timing of working capital (after refinancing fees my estimate is £6.7 to £7.1 million). Trailing twelve month owner cash is about £17.3 to £17.7 million, a 3.1% yield on the market value.

Capital spending and leases together were 7.3% of sales in fiscal 2024 and 5.0% in fiscal 2025. If I treated the £9.7 million of purchased technology as non-recurring, 2025 owner cash would be £36.3 million. I do not use that figure, because a technology transition needs product spending to keep going. No growth is valued while its software spending is removed.

Debt, shares and other claims

At March 31, 2026, RWS had £33.7 million of cash and £68.0 million of borrowings before issue costs, net principal debt of £34.3 million, plus £19.8 million of lease liabilities. The bank facility is US$285 million, maturing in October 2029. Rolling the balance forward to October 19 gives my estimate of £33.0 million of net debt.

There is a disclosure conflict about the facility. The half-year report says US$91 million was drawn, leaving US$194 million of headroom, but it also gives total liquidity of £290.1 million and calls US$91 million undrawn in the going concern note. At the implied exchange rate, US$194 million plus cash is about £179 million, not £290 million. I use the principal loans and treat the exact headroom as unknown. This looks like a drafting inconsistency, not a sign of distress, and the board's going concern review runs to September 2027.

I keep the £33.7 million of cash as operating liquidity and give it no credit as surplus, so I subtract gross debt, not net debt, from enterprise value in the sum of the parts. Lease cash is already in earnings, so I do not subtract the lease liability as well. Acogroup is not yet owned, so its price and its acquired cash enter together in fiscal 2027: £40.2 million gross less £9.8 million of usable acquired cash gives £30.4 million of net investment, after keeping £8 million of its cash as operating liquidity.

The market value is £563.3 million on the issued shares. The diluted average share count fell about 0.6% a year from FY2021 to FY2025. Future stock pay is charged as a cash cost (£5.5 million and £6.0 million in Base), so I do not also add new shares: that would charge the same cost twice. In FY2025 the stock pay of £3.8 million was 14.3% of owner cash. At September 2025 there were about 12.5 million awards and options outstanding, and 10.6 million were granted in the half year, so the allowance is a judgment, not a measurement.

Operating build behind the Base

The 2026 working figures are about £709.5 million of revenue and £110 million of adjusted EBITDA, which are my numbers and not company guidance. Revenue grows at a 10.1% rate to 2028, but a good part of that is purchased: Acogroup and Obviously together contribute £161 million of sales and £16.5 million of EBITDA in 2028, including synergies. Of the Base EBITDA, 65% turns into operating cash, which is the company's own medium-term aim. The remaining 35% has to pay for all leases, all capital spending and working capital, which in 2028 is a basket of about £49.5 million.

Primary valuation: owner cash with a two year transition

Normal cash is two years away, so multiplying it by a multiple and calling the result today's value would be a trick. The model runs forward to September 30, 2028, 1.95 years from the valuation date. Cash earned before then is valued in the year it arrives, the 2028 owner cash is capitalized at 9 times, and nothing is added for cash piled up in between. I use 9 times because it sits in the 8 to 10 times range I use for variable cash flows, and a premium multiple of 12 to 15 times needs a stable record that RWS does not have. Bear uses 8 times and Bull 11 times.

The 65% conversion already includes capital spending, working capital and leases, so they are not deducted a second time. Cash tax is 25% of EBITDA less leases, depreciation of non-acquired assets, stock pay, recurring cash costs and interest, which is a proxy and not a claim about taxable income. A last £3 million Obviously payment falls after the exit and is subtracted as a tail.

Because the multiple is applied to cash after interest, there is no further debt subtraction here. The 11% discount rate is my assumption for timing and execution risk, tested at 9% and 13%.

Secondary valuation: sum of the parts

The second method values each unit on its cash-adjusted operating profit (EBITDA less lease cash, ordinary depreciation, stock pay and recurring cash costs, before acquired amortization), then subtracts gross debt. The multiples are my calibration, not peer medians: I could not build a clean peer set of quoted language and software companies with comparable cash definitions, so no peer multiple enters the value.

Less £66.7 million of gross debt leaves £561.3 million of equity, or 150.89p at September 2028. Adding the 2027 and 2028 cash flows and subtracting the Obviously tail gives 134.92p today. Bear is 66.59p and Bull 219.87p.

At an 11% return and 2% long-run growth, an 8 times multiple on after-tax profit needs about 96% of that profit to turn into unlevered cash. Transform at 5 times with no growth needs about 73%. So the higher segment value assumes a more efficient mature business than the primary method's near-term cash basket. It does not prove that efficiency will appear.

• Shared by both methods: the recovery in revenue and profit, the acquisitions, the share count, the timing and the Obviously payments. More than half of each value rests on the same recovery.
• Different in each: the first capitalizes group owner cash at one multiple, the second values units separately and subtracts debt at the end.
• What agreement would prove: little. The two methods use the same forecasts, so close values cannot confirm that the forecasts are right, and I did not raise Confidence because they agree.
• Cross-check on the Acogroup price: an enterprise value of £22.4 million is 2.0 times its annualized EBITDA of £11 million. That is a declining, low-margin business, so it says little about what RWS itself is worth.

History, peers and growth diagnostics

I could not build a reliable peer median, and I did not make one up. For history I use the share price printed in each annual report divided by owner cash per share, because the third-party price series I checked disagrees with the annual reports (for example 86p against 92.3p for FY2025).

Five snapshots do not make a distribution, and the high 2024 multiple reflects depressed cash. The median is 18.1 times. Applied to the three-year average of £31.4 million, it gives 152.7p, almost exactly today's price. Applying my 9 times to the same average gives 76.0p. That is the strongest argument against my Base, and it is a diagnostic only: it imports past confidence and an acquisition-heavy past into a changed business. Still, my Base assumes owner cash of £52.1 million in 2028, well above either historical average.

Third-party rounded estimates put adjusted earnings per share at about 13p for 2026 and 16p for 2027. That is 9.5 times next year's adjusted earnings and growth of about 23%, which looks cheap. But those earnings exclude stock pay, acquired amortization and exceptional costs. My owner cash for 2027, before acquisition spending, is £43.2 million, against about £59.5 million of earnings implied by 16p on 372 million shares, a 73% conversion. The growth is also from a depressed base and includes bought revenue, so I give the ratio no weight in the value.

These growth rates are an acquisition-affected recovery from a low base, not organic compounding. Of the £25.5 million increase in annual cash, £21.2 million comes from lower exceptional and integration cash costs, the rest from higher EBITDA offset by a bigger reinvestment basket and more tax. The reliability of these estimates is low: I found no consensus for owner cash and no revision history.

Valuation triangulation

The two Base values are 16.2% apart, below my 30% line for a serious divergence. I gave the first method 60% because cash after all operating investment addresses the gap between adjusted profit and the cash that reaches owners, and the second 40% because it allows for a better mix in Generate and Protect. Equal weights give 124.78p, and the sum of the parts alone gives 134.92p. Neither reaches the price, so the choice of weights does not create a buy signal. About 90% of the value sits in the 2028 exit.

What is the market already pricing in?

These are alternative answers, not one forecast. Each says that the price needs something better than my Base, and my Base is already a recovery. A durable 78% conversion is close to what FY2025 showed with the help of working capital and above the 67% of the latest half. It is possible, but it has not been shown on a larger business after acquisitions. A modeled owner-cash return from today's price is -0.6% a year to the 2028 exit in Base, -26.2% in Bear and +28.0% in Bull (an information measure, not a forecast of dividends).

Bear, base and bull scenarios

Bear: legacy volumes and acquired profitability disappoint, EBITDA stays near £100 million, and the lower multiple reflects continuing machine-translation pressure beyond 2028. Base: the recovery continues, Acogroup is integrated, and leakage of exceptional cash falls from £28 million to £7 million. Bull: Generate and Protect grow faster, Acogroup earns more and the company pays the full £23.5 million to Obviously, so I do not take the revenue without the bill. I assign no probabilities, and the Base is not an expected price.

The range is wide, from 54.8% below the price to 31.4% above it, and about 90% of each value is in the 2028 exit. That limits how much precision any single figure deserves.

The grid shows triangulated value today in pence, rounded, changing only the two stated inputs. No cell is above today's price. The nearest are 75% conversion at 11 times, 163p, and 75% at 10 times, 154p.

How the buy price is set

A Confidence score of 13 is grade C, whose base margin of safety runs from 35% to 50%. I use 45%, near the upper end, because about 90% of value is in the exit, the key cash inputs are my estimates and the full-year cash is not yet reported. I add nothing on top. The maximum buy price is 122.76p times 0.55, or 67.52p, which is 55.5% below the price. Because the cash conversion alone moves the ceiling a lot, I show it as a review zone: at 55% conversion the ceiling is about 55p and at 75% about 80p. The upper end is not a standing order. It needs verified cash conversion first.

Cash conversion is the one input that moves the verdict, and the exit multiple alone is the second. A joint best case, with 9% discount rate, 75% conversion, a 12 times multiple and £13.7 million of surplus cash, gives about 180p, 18 to 20% above the price. Even that does not offer a 45% margin of safety at the market.

I keep four numbers apart. Fair value today is 122.76p. The modeled return at today's price is -0.6% a year. An entry price for an illustrative 11% return is 122.76p before any margin of safety. The margin of safety against model error gives 67.52p. Where each risk is priced:

• AI substitution and falling localization prices: in the volume and margin scenarios and in the 9 times multiple. No extra margin of safety.
• Capital spending, working capital and restructuring: in the 65% conversion and the exceptional cash reserve of £10 million then £7 million.
• Acquisition price and earn-outs: in the £30.4 million net investment and the Obviously reserves.
• Debt and the 2029 refinancing: in the interest and the gross-debt deduction. Leverage is moderate, so no add-on.
• Shared forecasts, high terminal share and unreported full-year cash: in Confidence grade C and the 45% margin of safety.

Dividend

The dividend is not a reason to buy. The total dividend was cut 43% to 7.05p for fiscal 2025, and the 2026 interim was 1.75p against 2.45p. The latest paid pair, 4.60p in February and 1.75p in July, is 6.35p, a 4.2% yield, but the 2026 final has not been declared. In fiscal 2025 the cash paid in dividends was £45.9 million, 172.6% of owner cash. At 6.35p the annual cost is about £23.6 million, which owner cash of £26.6 million covered 1.13 times, an 88.8% payout with little room for acquisitions. In Base 2028 owner cash covers it 2.2 times. Valued as a dividend that grows at 2% forever and discounted at 11%, 6.35p is worth 72.0p. That is a diagnostic with no weight in the value, but it shows that the current payout alone does not support the price.

Reasons to own RWS

• A specialist moat in costly-error work. Regulated content and IP administration are less exposed to generic machine translation than general localization.
• Evidence of an AI opportunity. Generate grew 52% in the half year and AI-related revenue is 32% of the group.
• A management reset. Cost cuts, a rebased dividend and a low Acogroup price could improve capital allocation if integration costs stay under control.
• Scale and client access. The existing client base gives a way to cross-sell content platforms and brand tools without building a new sales force.
• The Bull case is real. It reaches 199.49p, 31.4% above the price, if conversion and the multiple both improve.

These justify watching the stock. They do not remove the gap between adjusted profit and cash that reaches owners.

What could go wrong?

• AI replaces work faster than new products pay. Transform is 58% of sales and fell 4.6%. Lower prices or client insourcing could outweigh the growth in Generate, especially if that growth is a one-time launch program.
• Adjusted profit overstates cash. Capitalized software, recurring exceptional costs, working capital and stock pay kept owner cash at 26% of operational free cash flow in the latest half year.
• Acquisitions. Acogroup needs integration cash and French regulatory clearance, Obviously is small and was loss-making when bought, and larger profit can trigger larger earn-out payments. An £88.0 million goodwill impairment in 2025 shows the record on earlier deals.
• Funding and the forecast. Full-year cash is not yet reported, the facility disclosure conflicts, and the facility matures in October 2029, shortly after my exit date.

A further risk is concentration inside business lines: no client exceeded 10% of group sales, but a single program can dominate a unit. None of this is an immediate solvency concern.

Management execution

This scores the company's record, not only the current CEO, who joined in January 2025. Patrick De Smedt became independent Chair on October 8, 2026. The £50 million buyback used £19.4 million in FY2023 and £30.4 million in FY2024, and fewer shares are good, but owner cash did not rise afterward. The Obviously earn-out depends on EBITDA and the new long-term awards on total return and earnings per share, so none directly pays for strong owner cash after software spending. Assessment: 7/20, with credit for the cost action and the half-year improvement, and a cautious view of capital allocation until the acquisitions show returns.

Stock Analyza scorecard

The gap between the company's moat story and its cash record is what keeps it in the standard class. The next step up needs sustained high returns on all capital and cash per share that grows without acquisitions.

Business quality is 48/100. The moat is stronger where mistakes are expensive and weaker in general localization. Gross margin fell from 46.9% to 43.4% in FY2025 according to the annual reports as I read them. The balance sheet offers little liquidation support: equity of £763.2 million less £485.9 million of goodwill and £276.7 million of other intangibles leaves £0.6 million.

The valuation score is 43/100. The price is above Base, which gives no discount at all, and the points that remain come from the recovery range. A Bear value 54.8% below the price earns almost nothing for protection.

Confidence score

The key inputs are my estimates, which already caps Confidence at B, and the score of 13 is lower. The 16.2% gap between the methods is not enough for a downgrade of its own, and agreement between them would not have raised the score, because they share the same forecasts.

What I would watch from here

The next review point is the full-year 2026 results and any Acogroup completion or delay. I did not find a confirmed date for the results, so I give none. Monthly voting rights notices update the share count. I would refresh the valuation after any guidance change, material acquisition, large issuance or buyback, dividend change, financing news, major legal or regulatory event or leadership change.

The three most important thesis breakers

1. Transform erodes faster than new AI cash replaces it.
2. Owner cash cannot rise above its recent weak levels after recurring investment.
3. Acquisitions add consideration, dilution or leverage without the modeled earnings and cash.

Stress test and independent check

I rebuilt the model in a separate script: the owner cash for each case, the primary and sum-of-the-parts values, the triangulation, the returns, the reverse solutions, the conversion and multiple grid, and the one-input tests. The values reproduced to within about 0.02p, except the conversion tests, which differ from the source model by up to about 0.2p, and the joint best case, which came out near 179p against 181.5p. None of these differences touches the verdict. I checked the half-year figures, the FY2025 figures, the Acogroup and Obviously terms, the share count and the facility facts against the company's reports and announcements.

The best case needs four things at once: stronger conversion, a rerating, a lower required return and credit for cash. Even then it does not give the 45% margin of safety at today's price.

The strongest case against my Reduce is that adjusted profit really is recovering, that AI-related revenue and client repeat rates are improving, and that a 9 times multiple on cash may be too low if conversion holds near 75%. In that case the stock is roughly fairly valued, which would move me toward Hold. I take it seriously, and that is why the call is conditional and not a blanket sell. The strongest case against my Base is the opposite one: that stock pay, capital spending and exceptional costs keep owner cash near its 2025 level of £26.6 million, and then 9 times that cash is worth only about 64p at the exit.

Main data limits: full-year 2026 cash, the cash and debt at October 19, the Acogroup opening balance sheet, the exact facility headroom and the Obviously targets are not verified. The 2026 to 2028 revenue and profit paths, the multiples and the unit values are my estimates. A weighted peer median is not available, estimates for cash are not available, and the third-party EPS figures are rounded. The delayed price of 151.77p could not be matched with a second source: a quote page I checked later in the session showed a bid and offer of 145.60p and 145.90p, a previous close of 145.30p and a day high of 149.40p. At 145.50p the Base is 15.6% below the price and the conclusion is the same.

Final verdict: reduce at 151.77p

RWS has real operating progress and a plausible AI route, and my Base already counts it. But the price needs about 78% of EBITDA to become operating cash, while the company aims for 65% and the owner cash it actually delivered was a third of its headline figure. Until full-year cash shows that conversion, 151.77p pays for a result I cannot find in the numbers.

Verdict: REDUCE at 151.77p, watchlist for new money. Triangulated Base value 122.76p, downside 19.1%, Confidence C (13/25), required margin of safety 45%, maximum buy price 67.52p, review zone about 55p to 80p.

For current holders this is a reduce call, not a panic call: the successful-transition case is real and I show what it takes. For new money, the shares belong on the watchlist until the price approaches the review zone or the full-year results show owner cash on the way to £43 million.

Sources

Primary sources include RWS's 2025 annual report, the FY2025 final results announcement of December 11, 2025, the half-year results statement of June 11, 2026, the total voting rights notice of October 1, 2026, the Obviously announcement of May 5, 2026, the Acogroup announcement of August 3, 2026, the 2024, 2023, 2022 and 2021 annual reports for cash flows, share counts and year-end prices, and the half-year 2025 statement for earlier guidance. The reference price comes from the AJ Bell market page (delayed at least 15 minutes). Rounded earnings estimates come from a third-party forecast page. Broker price targets and ratings are not used.

The reference market price is 151.768p, a delayed trade at 12:31 London time on October 19, 2026. Values per share use 372.0 million economic shares. Net debt of £33.0 million is £66.7 million of gross loans less £33.7 million of cash, both estimates for October 19. The earnings path, multiples, scenario values and tests are analytical estimates, not company guidance.