Radware
An Israeli company that protects company websites and apps from attacks that try to knock them offline or break in.
The scores
Growth3/4 ↑
Revenue grew 10% in 2025 and 12% in Q3 2026, with cloud recurring revenue up 21%; not the fast-compounding growth a 4 needs.
Business quality2/4 →
Sticky appliances and no large customer, but a sliver of the web's protection and an operating margin a fraction of its rivals'.
Financial position4/4 →
$421.5M of cash and securities, no debt, positive cash from operations and $207.3M paid in advance.
Execution record3/4 ↑
Turned the mix toward cloud subscriptions and has met its recent guides, but five years of growth bought no operating profit.
Per share capture3/4 →
Buybacks of $61.8M in nine months outrun stock pay, but stock pay rose to $8.9M a quarter and absorbs much of the gain.
Potential2/4 →
A large, growing need for attack protection, but Radware holds a tiny share and its AI security products have no reported revenue.
Risk profile3/4 →
No customer above 10% and no debt, against Israeli operations, two contract manufacturers and a shekel that moves profit.
Valuation3/4 ↓
About 3.35 times sales net of cash; the price needs about 10% yearly growth for ten years, near what it grows now.
Evidence quality3/4 →
Filed figures carry the core business, but cloud recurring revenue is management's measure and AI security's scale is undisclosed.
If it works 1.1× · if it fails you keep about 100%.
The Verdict
We rate Radware neutral. The business is solid: sales grow about 12% a year, cloud subscriptions faster, and it has plenty of cash and no debt, but profit has stayed flat while costs rose. The price is fair. At about 3.35 times sales, buyers need revenue to grow about 10% a year for ten years, a little below today's pace. As a bet on its new AI security products it is defensive: the rest of the business and the cash cover the price, but even a good result adds little. The view hinges on whether profit finally starts to follow sales.
The long caseAs companies hand work to AI agents, they will pay to stop those agents being tricked or misused.
The evidence under this - 6 claims
- Recurring revenue is Radware's own measure, unaudited, and it counts maintenance contracts and booked orders rather than revenue.
- Cloud recurring revenue grew 23% at the end of 2025 and 21% now, so it is not speeding up.
- The test ignores the cash the business makes along the way, which lowers the growth needed.
- At a sale for 2 times revenue it needs about 14.8% a year.
The company
When attackers flood a website with fake traffic to knock it offline, or try to break in through its apps, Radware's products filter the bad traffic out. It sells them three ways: as a cloud service, as boxes that sit in the customer's own data centre, and as support for those boxes.
Its buyers are large companies and telecom carriers, mostly reached through resellers. A major global cloud provider recently chose its newest attack-protection box. Each sale keeps about 81 cents of every dollar after direct costs.
The evidence under this - 3 claims
- Neither the customer, the value nor the term was disclosed.
- A refresh cycle ends; once the base is replaced, appliance sales fall back.
How it makes money
In 2025 Radware made $301.9M: 63% from products, including cloud subscriptions, and 37% from support. Most of it now repeats: recurring revenue was $266M a year at September 2026, and the cloud share of it rose to 41%.
Customers often pay ahead. Radware held $207.3M of revenue already paid for but not yet earned, which arrives over the coming years.
The evidence under this - 4 claims
- Recurring revenue is Radware's own measure, unaudited, and it counts maintenance contracts and booked orders rather than revenue.
- Cloud recurring revenue grew 23% at the end of 2025 and 21% now, so it is not speeding up.
Radware's own unaudited measure.
Where the money comes from, stream by stream
| Stream | Share of sales (FY2025) | How it earns it |
|---|---|---|
| Products | 63% | Cloud security subscriptions, appliances and software licences; grew 22% in 2025 on cloud and the DefensePro X refresh. |
| Services | 37% | Support, maintenance, managed and professional services; fell 6% as customers move to subscriptions. |
Who runs it
Roy Zisapel founded Radware in 1996 and still runs it. He owns 5.89% of the shares, and his family's companies rent Radware its offices. His bonus pays on milestones and results against budget, and in May 2026 he was granted shares and options; nothing in his pay rewards cash flow.
Officers sold 99,638 shares on the open market in the past year, none under a pre-arranged plan, and the finance chief bought 2,000.
The evidence under this - 5 claims
Why it wins, and who could beat it
Why customers come back
Once a bank or a carrier has Radware's boxes tuned to its own traffic, swapping them out is slow and risky, so customers tend to renew and upgrade. No customer is 10% of sales, so no single loss hurts much.
Rivals take it seriously: A10 and NetScout both name Radware in their own annual reports. But it is a niche player. Among websites whose protection can be identified, Radware sits in front of under 0.1%.
The evidence under this - 5 claims
- Recurring revenue is Radware's own measure, unaudited, and it counts maintenance contracts and booked orders rather than revenue.
- Cloud recurring revenue grew 23% at the end of 2025 and 21% now, so it is not speeding up.
The threat
Cloudflare protects 85.5% of the websites whose protection can be identified, and Akamai, F5 and Amazon's cloud sell the same kind of protection. A10 and F5 earn far higher profit margins at similar growth. In the new AI security market the chief executive counts over 20 start-ups.
The evidence under this - 4 claims
- The vendor's Radware multiple nets only part of the cash, so the true discount on sales is wider.
- Radware's adjusted margin of about 14% is near A10's accounting margin, so part of the gap is stock pay and amortisation.
- No figure was given for customers, orders or revenue.
- In his own words, the agent security field is fragmented and 'yet to be seen'.
The rivals: NET, AKAM, AMZN, FFIV, ATEN, NTCT, and 2 more that are private, in-house or listed abroad.
Who else does this, rival by rival
| Ticker | Rival | Where it competes | How far along | Priced against it |
|---|---|---|---|---|
| NET | CloudflarePriced against it: One segment only | Core security: cloud protection, appliances and support | On sale | One segment only |
| AKAM | AkamaiPriced against it: One segment only | Core security: cloud protection, appliances and support | On sale | One segment only |
| AMZN | Amazon Web ServicesPriced against it: One segment only | Core security: cloud protection, appliances and support | On sale | One segment only |
| FFIV · ATEN | F5 and A10Priced against it: Yes | Core security: cloud protection, appliances and support | On sale | Yes |
| Not listed | Imperva (Thales)Priced against it: Not listed | Core security: cloud protection, appliances and support | On sale | Not listed |
| Private | AI agent security start-upsPriced against it: Private | AI security | On sale | Private |
| NTCT | NetScout SystemsPriced against it: Yes | - | On sale | Yes |
How big it could get
How big this could get
There is no reliable count of this market. One rough way to see it: the United States alone had 21,041 firms with 500 or more staff in 2022, and Cloudflare's larger customers pay about $433,000 a year each.
Radware's whole yearly revenue is a small fraction of what those firms could spend, so the room to grow is wide. How much of it Radware can win is the open question.
The evidence under this - 3 claims
Cloudflare sells far more than attack protection to these customers, so the average overstates what Radware's products would earn.
The results
The numbers
Sales grew 9.8% in the year to Dec 2025, about the same pace as the 5.2% in the year before.
Operating margin is 3.8%, up from −1.1% in the year to Dec 2022.
Running costs take 77¢ of each sales dollar, up from 75¢ in the year to Dec 2021.
Shareholders' money earned 6.1% a year.
It turned 14% of sales into free cash, $41.6M over the year.
Yearly sales growth has ranged from −11% to +15% over the last 5 years, with sales falling in 1 of them.
The share count fell 5.9% in 4.0 years, and pay in shares costs 8.0% of sales.
All the numbers
Filed figures from the company's 10-K and 10-Q, read 2026-10-30. Fiscal years end in December.
| Dec 2021 | Dec 2022 | Dec 2023 | Dec 2024 | Dec 2025 | |
|---|---|---|---|---|---|
| Size and growth | |||||
| Sales | $286M | $293M | $261M | $275M | $302M |
| Sales growth | +15% | +2.4% | −11% | +5.2% | +9.8% |
| Profit | |||||
| Gross margin | 82% | 82% | 80% | 81% | 81% |
| Operating margin | 6.4% | −1.1% | −12% | −1.4% | 3.8% |
| Net margin | 2.7% | −0.1% | −8.3% | 2.2% | 6.7% |
| What growth costs, per sales dollar | |||||
| Research and development | 26% | 30% | 32% | 27% | 26% |
| Sales and marketing | 42% | 43% | 48% | 45% | 42% |
| General and admin | 7.6% | 10% | 12% | 10% | 8.5% |
| Pay in shares | 6.1% | 9.3% | 13% | 9.5% | 8.0% |
| Capital spending | 2.0% | 3.0% | 2.1% | 1.9% | 2.8% |
| Cash | |||||
| Cash from operations | $71.8M | $32.1M | −$3.50M | $71.6M | $50.1M |
| Free cash flow | $66.2M | $23.3M | −$8.93M | $66.3M | $41.6M |
| Balance sheet | |||||
| Cash and short-term investments | $132M | $90.4M | $157M | $172M | $121M |
| Returns and shares | |||||
| Return on capital | - | - | - | - | 2.4% |
| Shares (diluted) | 47.5M | 44.9M | 42.9M | 43.4M | 44.7M |
What growth costs
Growth has not paid off yet. From 2021 to 2025 revenue rose from $286.5M to $301.9M while official operating income fell from $18.2M to $11.4M, because sales, marketing and research spending grew about as fast as sales. Pay in shares has since climbed to 8.8% of revenue.
The evidence under this - 3 claims
The last report (28 Oct)
Quarter ended September 2026, filed figures.
In the third quarter of 2026 sales rose 12% to $84.4M, every region up 12%, and cloud subscriptions reached $107M a year, up 21%.
On the surface it was steady growth. The problem was profit. Adjusted operating income was $11.9M, the same as a year earlier. Management blames the stronger shekel, which raised its Israeli costs, but research spending also rose 31% and pay in shares jumped to $8.9M from $5.3M.
For the fourth quarter it expects sales of $88M to $90M.
What it did to the thesisIt confirmed the growth half of the argument and left the profit half open: sales and subscriptions kept growing, but profit stayed flat for another quarter, so the case that growth reaches shareholders is still unproven.
The quarter in detail
Sales growth from the previous quarter, filed figures.
Earnings per share $0.28 against $0.29 expected.
The evidence under this - 6 claims
- Recurring revenue is Radware's own measure, unaudited, and it counts maintenance contracts and booked orders rather than revenue.
- Cloud recurring revenue grew 23% at the end of 2025 and 21% now, so it is not speeding up.
- The adjustment is management's own arithmetic and unaudited.
- Most of Radware's staff are paid in shekels, so the currency is a standing cost of the business rather than a one-off.
- Q4 is the seasonally strongest quarter, the 20-F says, so part of the step up is the calendar.
- The shekel cost $4.6M of Q3 net income and the guide assumes nothing about it.
What management did with the money
Radware bought back $61.8M of its own shares in the first nine months of 2026. It also shut down a money-losing side business that lost about $13M a year, and bought Pynt, a small company that tests apps for security holes, for $5.9M.
Insiders in 2026Q3: 2 purchases and 3 sales, from their Form 4 filings.
The evidence under this - 3 claims
The price, and where we disagree
In short. Fair. The price asks for growth a little below today's, and pays nothing extra for a margin that has not yet risen or for the AI products.
The shares rose 23% in five days in late October, almost 10% of it on the day Radware announced the large cloud customer. In July they had fallen 17% in a day after a weak second quarter.
So buyers have swung from selling flat profit to buying the growth story, with the profit unchanged in between. The next test is the fourth-quarter report, due by 10 February 2027: the guidance implies adjusted operating margin near 17%, up from 14.1%.
The evidence under this - 4 claims
- Neither the customer, the value nor the term was disclosed.
- A refresh cycle ends; once the base is replaced, appliance sales fall back.
Adjusted gross margin fell a point in Q3 on currency, mix and supply costs and could fall again.
How we read the multiple
Counting all its cash, the company costs about $1.09B; the tile counts less. What has to be true:
1. Revenue over the last year was $326.7M, so the price is about 3.35 times sales. 2. A buyer wanting 9% a year, a common yardstick, needs revenue near $860M in ten years, selling then at 3 times sales. 3. That is growth of about 10% a year for a decade, against 12% today.
Rivals earn several times its margin, so it looks cheap only if that margin rises.
Against its own history. Buyers pay less per dollar of sales than they usually have, which fits a company whose profit has gone nowhere for years.
The evidence under this - 4 claims
The vendor's company value of $1.45B nets only part of the cash, giving 4.45 times sales; the share count is a weighted average while buybacks continue.
- The test ignores the cash the business makes along the way, which lowers the growth needed.
- At a sale for 2 times revenue it needs about 14.8% a year.
- The vendor's Radware multiple nets only part of the cash, so the true discount on sales is wider.
- Radware's adjusted margin of about 14% is near A10's accounting margin, so part of the gap is stock pay and amortisation.
Priced against its rivals
The competitors its own annual report names, each confirmed against the rival's filing. Measured on company value for each $1 of sales, cheapest first:
| Ticker | Company | company value for each $1 of sales |
|---|---|---|
| NTCT | NetScout Systems | 2.7× |
| RDWR | Radware | 4.5× |
| FFIV | F5 | 5.6× |
| ATEN | A10 Networks | 6.6× |
The middle of the other 3 is 5.6. Radware Ltd. is 2nd cheapest of 4.
Rivals' figures come from a data provider, so read the spread of the group rather than any single pair.
Radware is cheaper than its rivals per dollar of sales, but they earn far more on each dollar, so the discount is Radware's own.
Private or foreign rivals that cannot be priced: Imperva (Thales), Citrix (Cloud Software Group).
The bet: what it pays if it works
The bet is on Radware's new AI security products, which protect the AI models and agents companies now build into their apps. They launched in 2026, have first customers and no reported revenue.
Valuing the rest of Radware fairly and adding its cash already covers today's price, so the bet costs almost nothing. But even if the products become a real business, they add only a little to the company's value: a defensive bet. If they fail, shareholders keep the core business and the cash.
| AI security revenue in two to three years | Million dollars a year needed | At the stated multiple |
|---|---|---|
| $10.00M (the big win) | about 10 at 100% kept | 1.1× |
| $25.00M | about 25 at 100% kept | 1.2× |
To pay 2× (a Fair bet) it would need about $247.32M of AI security revenue (about 247 million dollars a year); to pay 5× (a Cheap bet), about $1.07B of AI security revenue (about 1.1K million dollars a year), at the same multiple.
Each row is AI security alone, plus the rest of the company (Core security: cloud protection, appliances and support) counted at about $1.20B.
The steps it has to take
- Done AI Protect launched
- Done First production customers for MCP Server Protection
- Done Agent Trust Management launched
- Not started First disclosed AI security revenue or customer count due 2027-12-31
- Not started AI security near $10M of yearly revenue due 2028-12-31
The detail behind the label
The evidence under this - 5 claims
- No figure was given for customers, orders or revenue.
- In his own words, the agent security field is fragmented and 'yet to be seen'.
- The test ignores the cash the business makes along the way, which lowers the growth needed.
- At a sale for 2 times revenue it needs about 14.8% a year.
Where we disagree with the market
Radware is a cloud and AI security growth story, and the shares should follow the subscriptions.
The growth is real, but the price is only cheap if margins rise toward rivals', and they have not yet.
The market is counting new subscriptions and AI launches. We are counting profit, which has been flat for nine months while pay in shares rose fast.
We could be wrong: management says the shekel alone took $4.6M off quarterly profit, and if the currency turns, margins could jump quickly.
What analysts say, and the evidence
The evidence under this - 5 claims
- No figure was given for customers, orders or revenue.
- In his own words, the agent security field is fragmented and 'yet to be seen'.
- The test ignores the cash the business makes along the way, which lowers the growth needed.
- At a sale for 2 times revenue it needs about 14.8% a year.
What could go wrong, and how we would know
The risks
- Giant rivals Cloudflare and the large clouds sell similar protection to far more websites, and can bundle it cheaply.
- Israel Its headquarters, most research and final assembly are in Israel, which its annual report names as a risk from regional hostilities.
- The shekel Most costs are in shekels and sales in dollars, so a stronger shekel cuts profit, as it did by $4.6M last quarter.
- Two suppliers Two contract manufacturers made about 81% of its direct product costs in 2025.
- Pay in shares Stock pay rose to 8.8% of revenue; if it keeps outrunning buybacks, each share's slice shrinks.
- AI may not arrive The AI security products have no reported revenue, and over 20 start-ups compete for the same buyers.
The evidence under this - 6 claims
- The adjustment is management's own arithmetic and unaudited.
- Most of Radware's staff are paid in shekels, so the currency is a standing cost of the business rather than a one-off.
What we still do not know
- How much revenue the AI security products bring, and from how many customers.
- How much of the flat profit is the shekel, and how much is spending that will last.
- What happens to EdgeHawk, the other half of the side business that lost money.
The evidence under this - 3 claims
- The adjustment is management's own arithmetic and unaudited.
- Most of Radware's staff are paid in shekels, so the currency is a standing cost of the business rather than a one-off.
What would change our view
- Neutral→AttractiveA price that pays little for the growth
It needs all of:
- ○The company's value falls to 2.5 times its yearly sales, with the business unchanged.about ≈$27.34 today, 25% below the last close
- ○Cloud subscriptions still grow 18% or more a year.checked at the next results, due by 10 Feb
- Neutral→AttractiveProof that growth reaches profit
It needs all of:
- ○Adjusted operating margin reaches 18% in a quarter.checked at the next results, due by 10 Feb
- ○Sales grow 10% or more a year in that same quarter.checked at the next results, due by 10 Feb
We would also be wrong if
- Cloud subscriptions grow less than 15% a year for two quarters running.
- Pay in shares costs more than the buybacks return for a full year.
What is coming
- 10 February 2027 Q4 2026 results: Revenue against the $88M to $90M guide, whether adjusted operating margin reaches about 17% as the guide implies, and cloud recurring revenue growth.
Every dated event ahead
- 10 February 2027 Q4 2026 results: Revenue against the $88M to $90M guide, whether adjusted operating margin reaches about 17% as the guide implies, and cloud recurring revenue growth.
- May 2027 Q1 2027 results: Whether the first AI security revenue or customer figure is given, and whether stock pay keeps rising faster than sales.
Since we wrote this
- 30 OctThesis writtenNeutral, medium convictionWritten two days after Q3 2026 results, when the shares had risen 23% in a week on a cloud provider win and steady growth.
Sources 17
Sources - 17
Every claim above cites one of these, or states the reasoning that derived it. The quote is frozen at the date it was read: a link can rot, and an agent fetch returns a reading of a page rather than the page. Backs counts the whole claim ledger, so it can name a claim this page does not surface.
- S01primarycompany filing2026-03-31business description, segments, the Hawks, customers and concentration, channels, suppliers, competition, risk factors, directors and officers, pay, ownershipbacksC01C02C03C12C16C18C26C27C35C36C38C42
- S02primarycompany filing2026-10-28Q3 and nine-month statements, GAAP to non-GAAP reconciliation, balance sheet, cash flow, buybacks, the currency effect
Record quarterly revenue of $84.4 million, an increase of 12% year-over-year. Cloud ARR of $107 million, an increase of 21% year-over-year
backsC04C05C06C07C08C09C10C11C12C16C37C41C42 - S03primarytranscript2026-10-28Radware Q3 2026 earnings call transcript (FMP)data/cache/transcripts/RDWR-2026Q3.jsontotal and cloud ARR, the Q4 guide, the currency effect on operating income, AI security products and customers, the cloud provider win, the refresh cycle; questions from Jefferies, Wolfe Research and Needham
Total ARR reached approximately $266 million, up 11% year-over-year. ... For the fourth quarter of 2026, we expect total revenue to be in the range of $88 million-$90 million. ... I do think it is a 2027 opportunity, but we are definitely seeing a rise there.
backsC05C07C13C15C16C17C23C41 - S04primarycompany filing2026-02-112025 results with the Hawks still consolidated, ARR at year end, cash, stock pay, cash flow
Cloud ARR of $95.2 million, an increase of 23% year-over-year. Total ARR of $251.0 million, an increase of 11% year-over-year
backsC02C40C44 - S05primarycompany filing2026-07-29Q2 results by region, earnings per share against a year earlier, cash
Revenue in the Europe, Middle East, and Africa (“EMEA”) region was $27.3 million for the second quarter of 2026, a decrease of 2% from $27.8 million in the second quarter of 2025.
backsC32 - S06primarycompany filing2026-02-25a rival naming Radware among DDoS protection vendorsbacksC19
- S07primarycompany filing2026-05-14a rival naming Radware first among the vendors its Arbor DDoS line competes withbacksC19
- S08primaryregulator2025-10-01ENISA, 'EU consistently targeted by diverse yet convergent threat groups', ENISA Threat Landscape 2025 European Union Agency for Cybersecurity (ENISA)how common denial-of-service attacks are against European organisations, and how rarely they disrupt service, from the EU's cybersecurity agency
DDoS attacks was the dominant incident type and accounted for 77% of reported incidents, the greater part of which were deployed by hacktivists while cybercriminals represent only a minor portion. ... with only 2% of hacktivism incidents resulting in service disruption.
backsC21 - S09primaryindustry data2025-07-30what a breach costs the buyer, and the share of organisations reporting breaches of AI models or applications (the report's headline); frozen by this repo on 2026-10-26, the page now answers 404
The global average cost of a data breach fell to $4.44 million, the first decline in five years, while the average U.S. cost of a breach reached a record $10.22 million.
backsC22C45 - S10primaryindustry data2026-10-30W3Techs, Usage statistics and market share of Radware as reverse proxy service W3Techs (Q-Success)Radware's reach among websites, measured outside the company
Radware is used by less than 0.1% of all the websites whose reverse proxy service we know.
backsC20 - S11primaryindustry data2026-10-10W3Techs, Usage statistics of Cloudflare as reverse proxy service W3Techs (Q-Success)the leading rival's share of websites on the same survey
Cloudflare is used by 85.5% of all the websites whose reverse proxy service we know. This is 26.8% of all websites .
backsC20 - S12primarygovernment data2025-04-10U.S. Census Bureau, Statistics of U.S. Businesses 2022, detailed enterprise employment size table U.S. Census Bureaufirms by size; 500 and over is the total less the under-500 line
The Number of Firms and Establishments, Employment, Annual Payroll, and Receipts by State, Industry, and Enterprise Employment Size: 2022 ... United States, Total, 01: Total, 6395635 firms, 135748407 employment ... 19: <500 employees, 6374594 firms, 62250556 employment
backsC24 - S13primarycompany filing2026-08-064,698 customers paying over $100,000 a year, 73% of Q2 revenue of $696.1M, from a rival's own filing and callread, but no claim rests on it
- S14secondaryvendor data2026-10-30Research pack (FMP market tier and FINRA short interest), frozen as content/RDWR/research-pack.jsoncontent/RDWR/research-pack.jsonprice and returns, the analyst split and its history, short interest, the filed cohort's multiples and marginsbacksC30C31C33C43
- S15primarycompany filing2026-10-29every open-market trade and grant in the year, each with its Rule 10b5-1 flagbacksC34C37
- S16primarycompany filing2026-10-30six years of filed revenue, costs, stock pay, capital spending and share countsbacksC39
- S17primarycompany filing2025-11-14a rival selling application delivery and web application, API and DDoS protection to the same buyersread, but no claim rests on it