A10 Networks
A10 Networks sells the boxes and software that spread internet traffic across servers and block flooding attacks, to phone companies, cloud operators and large businesses.
Our viewUnattractive Medium conviction
The scores
The profile · 1-4
Growth2/4 →
Revenue grew 11.0% in 2025 and 14.5% in H1 2026, all from one customer's rollout; sales to everyone else fell about 3% and 8%.
Business quality2/4 →
Installed boxes keep an 80% gross margin, but the largest customer runs its own version and Cloudflare sells protection with no box at all.
Financial position4/4 →
$357.3M of cash and securities against a $225.0M convertible due 2030 with no covenants, and about $61M of free cash flow a year.
Execution record3/4 ↑
Eight straight quarterly beats and a raised 2026 guide, but revenue fell 10.2% in 2023 and today's growth hangs on one buyer.
Per share capture2/4 ↓
Buybacks cut shares to 72.6M, but the convertible sits above its $23.46 price and Microsoft's warrant adds up to 800,000 shares.
Potential2/4 →
Growth needs Microsoft to keep buying past secret thresholds and carriers to recover while IPv4 demand slowly shrinks; nothing is contracted.
Risk profile2/4 ↓
One customer is 38% of revenue and one distributor 40%; Microsoft's ordering decision is external and resolves in a single step.
Valuation2/4 →
About 35 times trailing operating profit: the price needs that profit 2.6 times larger, which sales outside Microsoft are not growing toward.
Evidence quality3/4 →
The customer table, warrant and statements are filed; that Customer A is Microsoft, and its future orders, rest on inference and management.
The Verdict
We rate A10 Networks unattractive. It sells equipment and software that keep networks running under heavy traffic and attack, and its sales are growing again, but the growth is one customer, almost certainly Microsoft, while sales to everyone else fall, so the business is mixed. Microsoft already writes its own version of the product, and A10 now pays it in shares for buying more. The price assumes that one relationship keeps growing for years, so it is full, not cheap. The view would change if sales to other customers grow again, or the big customer's share falls while growth holds.
The long caseMuch of the internet still runs on an old address system, so carriers keep buying A10's address-sharing boxes; the slow move to the newer system must stay slow.
The evidence under this - 9 claims
C02facthigh confidence
Q2 2026 revenue was $80.1M, up 15.5%: products $49.0M, up 25.1%, and services $31.1M, up 3.0%. First-half revenue was $155.1M, up 14.5%.
C04facthigh confidence
One end-customer, Customer A, was 14% of revenue in 2023, 15% in 2024, 26% in 2025, 37% in Q1 2026 and 38% in Q2 2026 and in the first half.
C05inferencemedium confidence
Customer A is almost certainly Microsoft: analysts on the Q2 2026 call described the 37% customer and asked about Microsoft being so large a share, and the CEO answered about Microsoft without correcting either; Microsoft received the purchase-linked warrant two days earlier.
derivedThe Q2 2026 call (S05): one analyst linked the 37% customer disclosed in the Q1 10-Q to Microsoft, another asked how the company manages 'Microsoft being so large of a percentage of revenue', and the CEO answered both about Microsoft. The warrant (C07) vests only on Microsoft's purchases.
against this
- A10 never names Customer A in a filing; the 10-Q calls it only 'Customer A'.
- The CEO said Microsoft's newer business is booked as enterprise and its older business as service provider, so it could in principle be two customers of the same size rather than one.
C06estimatemedium confidence
Revenue from everyone other than Customer A was about $216M in 2023, $222M in 2024 and $215M in 2025, down about 3%; about $96M in the first half of 2026 against about $104M a year earlier, down about 8%; and about $49.7M in Q2 2026 against $48.6M, up about 2%. Customer A's revenue rose from about $39M in 2024 to about $76M in 2025 and about $59M in the first half of 2026.
derivedFiled revenue (C11, C02) times one minus Customer A's filed share (C04): 2023 $251.7M x 86%, 2024 $261.7M x 85%, 2025 $290.6M x 74%, H1 2025 $135.5M x 77%, H1 2026 $155.1M x 62%, Q2 2025 $69.4M x 70%, Q2 2026 $80.1M x 62%. The shares are rounded to whole percents, so each figure carries about half a point of revenue either way.
against this
- The CEO said on the Q2 2026 call that the business excluding Microsoft and one or two countries with macro problems is 'growing close to double digits'.
- Q2 2026 alone shows sales outside Customer A up about 2%, so the fall is concentrated in Q1 2026, when Customer A pulled a project forward.
C24facthigh confidence
Microsoft runs and sells its own layer-4 load balancer in Azure, which it says scales to millions of flows, and its own layer 3 and 4 DDoS protection service.
C07facthigh confidence
On 3 August 2026 A10 issued Microsoft a warrant for up to 800,000 shares at $0.01, in two tranches of 400,000 that vest only if Microsoft and its affiliates buy past purchase thresholds in periods ending 30 June 2027 and 30 June 2028, exercisable from 2028 and 2029. The thresholds are redacted, and A10 says it cannot yet estimate the accounting effect.
C17estimatemedium confidence
At $25.81 on 72.61M shares the market value is about $1.874B. Adding the $225.0M convertible and subtracting $357.3M of cash and securities gives about $1.742B, which is 34.8 times trailing GAAP operating income of $50.1M, 22.0 times non-GAAP operating income of $79.1M, and 28.6 times trailing free cash flow of $60.9M. The share price is 43 times trailing GAAP earnings of $0.60.
derivedClose of $25.81 on 2026-09-29 (S18) times 72.61M filed shares (C16); convertible at principal and cash and securities from C14; operating income from C12 and C13; free cash flow from C19; trailing GAAP diluted EPS of $0.17, $0.14, $0.17 and $0.12 (S04).
against this
- The vendor's enterprise value leaves out $302.7M of marketable securities and prints 40.0 times on the same line.
- Counting the convertible at its $247.5M share value rather than its $225.0M principal raises the multiple to about 35.2.
- 2026 guidance (C09) would lift operating income, so a forward multiple is lower than a trailing one.
C18estimatemedium confidence
For today's value net of cash to earn a 6% after-tax operating yield at a 21% tax rate, GAAP operating income must reach about $132M, 2.6 times the trailing $50.1M: for example revenue near $530M at a 25% margin, which is 11% a year for five years from $310M with the margin rising from 16%.
derived$1.742B (C17) x 6% / (1 - 21%) = $132M; $132M / 25% = $530M; ($530M / $310.1M) to the power of one fifth is 1.113. The 6% yield and the 25% margin are this thesis's own judgements, not the company's; 25% is the non-GAAP margin today (C13).
against this
- Management's non-GAAP margin is already 25.5% (C13); if share pay stayed at 7% of revenue, GAAP margin would reach 25% only with further operating leverage.
- At management's 12-14% growth (C09), revenue compounds faster than 11% for as long as the guide holds.
C22factmedium confidence
The vendor's rating snapshot counts 11 Buy, 10 Hold and no Sell; its dated series on a panel of 7 fell from 71.4% positive in January 2026 to 57.1% by September.
The company
When millions of people hit a website or a phone network at once, something has to decide which server answers each one, share scarce internet addresses among subscribers, and throw away floods of fake traffic sent to knock the service over. A10 sells boxes, and the same software for a customer's own servers, that do those jobs.
Its buyers are phone and internet carriers, cloud operators and large companies. The alternatives are rivals such as F5 and Radware, a router's built-in feature, or a cloud service with no box.
The evidence under this - 4 claims
C01facthigh confidence
A10 sells boxes and software, all built on its own ACOS operating system, that spread traffic across servers (load balancing), let a carrier share one public internet address among many subscribers (carrier-grade address translation), block flooding attacks (DDoS protection) and inspect encrypted traffic. It reports one segment; 96% of 2025 revenue went through distributors and resellers.
C23facthigh confidence
A10's 10-K names F5, NetScaler, NetScout's Arbor, Radware, Cisco, HPE-Juniper, Fortinet, Akamai, Cloudflare, Imperva and Amazon Web Services as competitors; Radware's 20-F and NetScout's 10-K name A10 back.
C24facthigh confidence
Microsoft runs and sells its own layer-4 load balancer in Azure, which it says scales to millions of flows, and its own layer 3 and 4 DDoS protection service.
C05inferencemedium confidence
Customer A is almost certainly Microsoft: analysts on the Q2 2026 call described the 37% customer and asked about Microsoft being so large a share, and the CEO answered about Microsoft without correcting either; Microsoft received the purchase-linked warrant two days earlier.
derivedThe Q2 2026 call (S05): one analyst linked the 37% customer disclosed in the Q1 10-Q to Microsoft, another asked how the company manages 'Microsoft being so large of a percentage of revenue', and the CEO answered both about Microsoft. The warrant (C07) vests only on Microsoft's purchases.
against this
- A10 never names Customer A in a filing; the 10-Q calls it only 'Customer A'.
- The CEO said Microsoft's newer business is booked as enterprise and its older business as service provider, so it could in principle be two customers of the same size rather than one.
How it makes money
A10 earns most of its money when a customer buys a box: products were $49.0M of the $80.1M it took in during Q2 2026. Each box comes with a maintenance contract, paid upfront and lasting about 21 months on average, which renews while the box stays in the network. That maintenance, $31.1M in the quarter, is the steady part of the business, and it grew only 3.0%.
The evidence under this - 3 claims
C02facthigh confidence
Q2 2026 revenue was $80.1M, up 15.5%: products $49.0M, up 25.1%, and services $31.1M, up 3.0%. First-half revenue was $155.1M, up 14.5%.
C28facthigh confidence
Maintenance is invoiced with the hardware, non-cancellable for its term and averages about 21 months; A10 had $154.8M of remaining performance obligations at 30 June 2026, $93.2M due within a year. Services revenue grew 3.0% in Q2 2026 and 2.9% in the first half while products grew 25.1% and 23.8%.
C01facthigh confidence
A10 sells boxes and software, all built on its own ACOS operating system, that spread traffic across servers (load balancing), let a carrier share one public internet address among many subscribers (carrier-grade address translation), block flooding attacks (DDoS protection) and inspect encrypted traffic. It reports one segment; 96% of 2025 revenue went through distributors and resellers.
Where the money comes from, stream by stream
| Stream | Share of sales (FY2025) | How it earns it |
|---|---|---|
| Products | 57.5% | Hardware appliances running A10's ACOS software under a perpetual licence, plus a small share of term and subscription licences, sold mostly through distributors. |
| Services | 42.5% | Maintenance and support contracts invoiced with each box and renewed, plus professional services, training and software-as-a-service. |
Who runs it
Dhrupad Trivedi has run A10 since late 2019 and also chairs its board; the finance chief and the sales chief both arrived in 2025. Insiders own about 1% of the company, so their fortunes ride more on pay than on shares they already hold. The yearly bonus pays half on sales and half on a profit measure that leaves out stock pay; the share awards pay when the stock price clears set levels. Nothing rewards winning customers beyond Microsoft.
The evidence under this - 3 claims
C39facthigh confidence
Dhrupad Trivedi has been president and chief executive since December 2019 and chair since September 2020, after running Belden's industrial networking and cybersecurity business from 2013 to 2019. Michelle Caron became chief financial officer in September 2025, from Danaher's Beckman Coulter Life Sciences, after Brian Becker resigned; Sheen Khoury became head of worldwide sales and marketing in June 2025, from HPE; Scott Weber has been general counsel since June 2022.
C40facthigh confidence
At 26 February 2026 the eight current directors and executive officers owned 869,373 shares, 1.21% of 71,723,964 outstanding; the chief executive's 323,209 shares were the largest insider holding, under 1%. BlackRock reported 18.38% and Vanguard 10.37%, BlackRock's figure from a Schedule 13G/A of January 2024.
C41facthigh confidence
The 2025 cash bonus paid 50% on GAAP revenue and 50% on adjusted EBITDA, which excludes stock pay, and was earned at 132.1% of target: $1,056,800 to the chief executive on an $800,000 target. Long-term awards are half RSUs vesting over three years and half PSUs that vest when the 100-day average share price reaches $20.41 and $22.26 within four years (2026 grants: $21.00 and $21.50, against $17.78 at grant). Nothing pays on per-share earnings, cash flow or customer spread.
Why it wins, and who could beat it
Why customers come back
Once one of these boxes sits in a carrier's network, loaded with thousands of rules and carrying live traffic, replacing it means re-testing everything. So customers tend to renew rather than shop around.
The evidence is that A10 keeps about 80 cents of every sales dollar after the cost of making the product, and has done so for three years even as one giant customer grew. The company does not publish how many customers renew, so the strength of this is inferred, not measured.
The evidence under this - 3 claims
C28facthigh confidence
Maintenance is invoiced with the hardware, non-cancellable for its term and averages about 21 months; A10 had $154.8M of remaining performance obligations at 30 June 2026, $93.2M due within a year. Services revenue grew 3.0% in Q2 2026 and 2.9% in the first half while products grew 25.1% and 23.8%.
C32facthigh confidence
GAAP gross margin stayed between 78.7% and 81.1% in every quarter from Q1 2024 to Q2 2026, and non-GAAP between 80.0% and 83.1% since Q1 2023, while Customer A rose from 15% to 38% of revenue.
C26facthigh confidence
The IETF's carrier-grade NAT standard exists because of the shortage of IPv4 addresses: a carrier shares one public IPv4 address among many subscribers and needs equipment in its network to translate between them.
The threat
The biggest threat is the biggest customer. Microsoft runs its own traffic-balancing service for its cloud and its own attack protection, so it buys from A10 only where A10 is better or cheaper. Cloudflare, meanwhile, sells attack protection from its own network, with no box in the customer's building.
The evidence under this - 3 claims
C24facthigh confidence
Microsoft runs and sells its own layer-4 load balancer in Azure, which it says scales to millions of flows, and its own layer 3 and 4 DDoS protection service.
C25facthigh confidence
Cloudflare sells DDoS protection for on-premises networks from its own global network, 'instead of relying on local infrastructure' such as an appliance in the customer's building.
C34inferencemedium confidence
Because Microsoft already runs its own load balancer and DDoS service, it buys A10 only for the jobs where A10 is better or cheaper, and it has bargaining power: the warrant is a price A10 paid to keep and grow the business.
derivedMicrosoft's own services (C24); the warrant tied to Microsoft's purchases (C07, C08); the CEO's description of a 'completely different product set' (C10).
against this
- Microsoft expanded its A10 purchases from about $39M to about $76M in a year while running its own services (C06), so its own build did not stop it buying.
- A multi-year warrant also ties Microsoft's roadmap to A10's, which the CEO calls mutual performance commitments.
The rivals: MSFT, NET, FFIV, RDWR, NTCT, AKAM, FTNT.
Who else does this, rival by rival
| Ticker | Rival | How far along | Priced against it |
|---|---|---|---|
| MSFT | Microsoft Azure Load Balancer and Azure DDoS ProtectionPriced against it: Different product | On sale | Different product |
| NET | Cloudflare Magic TransitPriced against it: One segment only | On sale | One segment only |
| FFIV · RDWR | F5 BIG-IP and Radware DefenseProPriced against it: Yes | On sale | Yes |
| NTCT | NetScout SystemsPriced against it: Yes | On sale | Yes |
| AKAM | Akamai TechnologiesPriced against it: One segment only | On sale | One segment only |
| FTNT | FortinetPriced against it: One segment only | On sale | One segment only |
How big it could get
How big this could get
One of A10's jobs exists because the old system of internet addresses ran short, so carriers share each address among many subscribers. About 56% of the world's internet users still depend on that old system, a share falling 2.4 to 3.5 points a year as the newer system spreads.
Nobody publicly counts how many of these boxes are bought each year, so the market cannot be sized honestly beyond that.
The evidence under this - 3 claims
C27factmedium confidence
APNIC measures 43.70% of the world's internet users as IPv6-capable on 29 September 2026, against 41.29% a year earlier, 38.82% two years earlier and 35.28% three years earlier: the share still limited to the old addresses is about 56% and falls roughly 2.4 to 3.5 points a year.
C26facthigh confidence
The IETF's carrier-grade NAT standard exists because of the shortage of IPv4 addresses: a carrier shares one public IPv4 address among many subscribers and needs equipment in its network to translate between them.
C06estimatemedium confidence
Revenue from everyone other than Customer A was about $216M in 2023, $222M in 2024 and $215M in 2025, down about 3%; about $96M in the first half of 2026 against about $104M a year earlier, down about 8%; and about $49.7M in Q2 2026 against $48.6M, up about 2%. Customer A's revenue rose from about $39M in 2024 to about $76M in 2025 and about $59M in the first half of 2026.
derivedFiled revenue (C11, C02) times one minus Customer A's filed share (C04): 2023 $251.7M x 86%, 2024 $261.7M x 85%, 2025 $290.6M x 74%, H1 2025 $135.5M x 77%, H1 2026 $155.1M x 62%, Q2 2025 $69.4M x 70%, Q2 2026 $80.1M x 62%. The shares are rounded to whole percents, so each figure carries about half a point of revenue either way.
against this
- The CEO said on the Q2 2026 call that the business excluding Microsoft and one or two countries with macro problems is 'growing close to double digits'.
- Q2 2026 alone shows sales outside Customer A up about 2%, so the fall is concentrated in Q1 2026, when Customer A pulled a project forward.
Who has signed
Microsoft holds a warrant for up to 800,000 A10 shares at a penny each. Half vests if its purchases pass a threshold in the year to June 2027, half in the year to June 2028. The thresholds are blacked out in the filing, and Microsoft is not obliged to buy anything.
The evidence under this - 1 claim
C07facthigh confidence
On 3 August 2026 A10 issued Microsoft a warrant for up to 800,000 shares at $0.01, in two tranches of 400,000 that vest only if Microsoft and its affiliates buy past purchase thresholds in periods ending 30 June 2027 and 30 June 2028, exercisable from 2028 and 2029. The thresholds are redacted, and A10 says it cannot yet estimate the accounting effect.
Where the growth is coming from, business by business
Where the growth is coming from
| Business | Share of revenue | Growth: last quarter | Last half-year | Last year |
|---|---|---|---|---|
| Product | 61% | +25.1% | +23.8% | +19.1% |
| Service | 39% | +3.0% | +2.9% | +3.4% |
| Whole company | 100% | +15.5% | +14.5% | +12.2% |
The filed split through 2026-06-30. These rates are not carried forward: the reading below says why.
How we read it: The blend. Product and Service are revenue types in one segment, not businesses; Product's 24% is Customer A's rollout, while sales to everyone else fell about 8% in the half. One buyer's orders, not a mix shift.
The results
The numbers
Is it growing?Steady
Sales grew 12% over the twelve months to Jun 2026, about the same pace as the 11% in the year to Dec 2025.
Is it profitable?Healthy
Operating margin is 16%, about where it was (15%) in the year to Dec 2023.
Does growth pay off?Costly
Running costs take 63¢ of each sales dollar, down from 65¢ in the year to Dec 2021.
Does it earn good returns?Strong
Shareholders' money earned 19% a year, down from 58% in the year to Dec 2021.
Is it making cash?Strong
It turned 20% of sales into free cash, $60.9M over the last twelve months.
Is the balance sheet safe?Net cash
It holds $357M of cash against $0 of debt, $357M more cash than it owes.
Are shares being diluted?Shrinking
The share count fell 5.5% in 4.5 years, and pay in shares costs 7.5% of sales.
All the numbers
Filed figures from the company's 10-K and 10-Q, read 2026-10-04. Fiscal years end in December.
| Dec 2021 | Dec 2022 | Dec 2023 | Dec 2024 | Dec 2025 | 12 mo to Jun 2026 | |
|---|---|---|---|---|---|---|
| Size and growth | ||||||
| Sales | $250M | $280M | $252M | $262M | $291M | $310M |
| Sales growth | +11% | +12% | −10% | +4.0% | +11% | +12% |
| Profit | ||||||
| Gross margin | 79% | 80% | 81% | 80% | 79% | 79% |
| Operating margin | 13% | 19% | 15% | 17% | 16% | 16% |
| Net margin | 38% | 17% | 16% | 19% | 15% | 14% |
| What growth costs, per sales dollar | ||||||
| Research and development | 22% | 21% | 22% | 22% | 24% | 25% |
| Sales and marketing | 34% | 32% | 34% | 32% | 29% | 28% |
| General and admin | 9.4% | 8.4% | 9.5% | 9.7% | 10% | 11% |
| Pay in shares | 5.8% | 4.8% | 5.6% | 6.5% | 6.9% | 7.5% |
| Capital spending | 2.1% | 3.9% | 4.3% | 4.7% | 6.9% | 5.1% |
| Cash | ||||||
| Cash from operations | $50.1M | $66.1M | $44.5M | $90.5M | $84.9M | $76.9M |
| Free cash flow | $44.9M | $55.3M | $33.6M | $78.2M | $64.8M | $60.9M |
| Balance sheet | ||||||
| Cash and short-term investments | $185M | $151M | $154M | $196M | $378M | $357M |
| Debt | - | - | - | $0 | $219M | $0 |
| Net cash | - | - | - | $196M | $159M | $357M |
| Returns and shares | ||||||
| Return on capital | 21% | 24% | 18% | 17% | 11% | - |
| Shares (diluted) | 80.0M | 77.8M | 75.5M | 75.3M | 73.6M | 75.7M |
What growth costs
Growing has not made A10 more profitable. It now spends less to win each sale, but more on research, administration and pay in shares, so the profit kept from each dollar of sales has barely moved in three years, and it fell in the latest quarter even as sales jumped.
The evidence under this - 2 claims
C43facthigh confidence
Per sales dollar, from 2023 to 2025 and the four quarters to June 2026, R&D rose from 21.9 to 23.8 and 24.9 cents, general and administrative from 9.5 to 10.3 and 10.7, stock pay from 5.6 to 6.9 and 7.5 and capital spending from 4.3 to 6.9 and 5.1, while sales and marketing fell from 34.2 to 29.1 and 27.7. Revenue rose from $251.7M to $310.2M; GAAP operating margin went from 15.4% to 16.8% in 2024, 16.2% in 2025 and 16.1%.
C44inferencemedium confidence
Growth has not widened A10's GAAP margin: the saving in sales and marketing per dollar was spent on R&D, administration and stock pay, so revenue 23% larger left operating margin within a point of 2023's, and in Q2 2026 margin fell on 15.5% growth.
derivedC43's cost lines per sales dollar and C12's Q2 2026 operating income
against this
- Non-GAAP operating margin, which excludes stock pay and acquisition costs, was 25.5% in Q2 2026 (C13), so part of the GAAP squeeze is share pay and TrojAI deal costs rather than the cost of selling.
- Operating margin did rise from 15.4% in 2023 to 16.8% in 2024 on 4.0% growth (C43).
The last report (5 Aug)
Quarter ended June 2026, filed figures.
15%
Sales growth
against the same quarter a year earlier
6.8%
Growth from the quarter before
against -6.7% the quarter before that
11¢
Operating profit
per dollar of sales
Q2 2026 looked like a strong quarter: revenue up 15.5% and the full-year outlook raised to 12-14% growth. The mix told a different story. Sales to large companies jumped 72.9% while sales to carriers fell 22.8%, and Customer A alone paid 38% of revenue.
Take that customer out and sales to everyone else rose only about 2% in the quarter, after falling in Q1 2026. Profit also went the wrong way: operating profit fell to $9.0M from $10.3M as share-based pay doubled.
What it did to the thesisIt confirmed the Unattractive view: the growth came almost entirely from one large customer, while sales to everyone else barely moved and carrier sales fell. Profit fell too as share pay doubled. The raised outlook depends on that customer's rollout continuing, and the company itself says its share will change.
The quarter in detail
Sales growth from the previous quarter, filed figures.
Earnings per share $0.25 against $0.24 expected; $0.12 as filed, the gap being what the company adds back.
The evidence under this - 6 claims
C02facthigh confidence
Q2 2026 revenue was $80.1M, up 15.5%: products $49.0M, up 25.1%, and services $31.1M, up 3.0%. First-half revenue was $155.1M, up 14.5%.
C03facthigh confidence
In Q2 2026 enterprise revenue was $48.0M against $27.8M a year earlier, up 72.9%, and service provider revenue $32.1M against $41.6M, down 22.8%. United States revenue rose 35%; APJ fell 7% and EMEA 16%.
C04facthigh confidence
One end-customer, Customer A, was 14% of revenue in 2023, 15% in 2024, 26% in 2025, 37% in Q1 2026 and 38% in Q2 2026 and in the first half.
C06estimatemedium confidence
Revenue from everyone other than Customer A was about $216M in 2023, $222M in 2024 and $215M in 2025, down about 3%; about $96M in the first half of 2026 against about $104M a year earlier, down about 8%; and about $49.7M in Q2 2026 against $48.6M, up about 2%. Customer A's revenue rose from about $39M in 2024 to about $76M in 2025 and about $59M in the first half of 2026.
derivedFiled revenue (C11, C02) times one minus Customer A's filed share (C04): 2023 $251.7M x 86%, 2024 $261.7M x 85%, 2025 $290.6M x 74%, H1 2025 $135.5M x 77%, H1 2026 $155.1M x 62%, Q2 2025 $69.4M x 70%, Q2 2026 $80.1M x 62%. The shares are rounded to whole percents, so each figure carries about half a point of revenue either way.
against this
- The CEO said on the Q2 2026 call that the business excluding Microsoft and one or two countries with macro problems is 'growing close to double digits'.
- Q2 2026 alone shows sales outside Customer A up about 2%, so the fall is concentrated in Q1 2026, when Customer A pulled a project forward.
C12facthigh confidence
GAAP operating income was $9.0M in Q2 2026 (11.3% of revenue) against $10.3M a year earlier, and $50.1M over the four quarters to Q2 2026. Share pay was $9.0M in the quarter against $4.4M; R&D rose 30.1% and G&A 57.4%.
C09management claimmedium confidence
Management raised its 2026 outlook on 5 August 2026 to revenue growth of 12-14% and EPS growth of 14-16%, from 10-12% and 12-14% set in February, and expects free cash flow above about $65M in 2025.
sourceQ2 2026 results release (8-K Exhibit 99.1)primaryQ2 2026 earnings call transcript (FMP), held 2026-08-05primaryQ4 and full-year 2025 results release (8-K Exhibit 99.1)primary
against this
- First-half growth was 14.5%, so 12-14% for the year implies second-half growth of about 10-13.5%, a slowdown.
- The CEO said Customer A's share is 'linked to kind of completing the rollout' and 'will change probably in the future'.
What management did with the money
A10 bought back 3.67M shares in 2025 for $68.9M and pays a quarterly dividend of 6 cents. In June 2026 it spent $34.7M in cash and $5.0M in stock on TrojAI, a small company that tests and guards AI software.
Insiders in 2026Q2: 15 purchases and 19 sales, from their Form 4 filings.
The evidence under this - 2 claims
C16facthigh confidence
72.61M shares were outstanding at 31 July 2026, against 73.69M at 31 December 2024. A10 bought back 3.67M shares for $68.9M in 2025 and $4.9M worth in the first half of 2026, pays $0.06 a share each quarter, and has $53.1M of authorisation left.
C30facthigh confidence
A10 bought TrojAI, a Canadian AI-security start-up that tests and guards AI models, on 15 June 2026 for $34.7M in cash and $5.0M in stock, recording $32.8M of goodwill; it bought the ThreatX Protect web-application business for $19.1M in February 2025. Neither is material to revenue.
The price, and where we disagree
In short. Full. The price pays for years of growth that so far comes from one buyer who could build the product itself.
Share price
$25.81
close,
Company value for each $1 of operating profit
40.0
rivals' middle 28.8
The shares nearly doubled from February to a high in July on the hope that A10 is an AI winner, then fell by a third, including 10.5% the day after a report that raised the outlook.
Two new facts arrived with that report: the Microsoft warrant, and convertible bonds that holders may now swap for shares because the price passed their trigger. About 13.7% of the shares available to trade are sold short, a sign many investors doubt the story.
The evidence under this - 4 claims
C20facthigh confidence
The shares closed at $17.52 on 4 February 2026, reached a $38.49 high on 9 July 2026 and closed at $25.81 on 29 September 2026. They fell 5.3% the day after the Q1 2026 print (28.13 to 26.63) and 10.5% the day after the Q2 2026 print (30.84 to 27.60). Over 252 sessions they rose 41.0%, of which sales per share explained 8.4% and a higher multiple 30.0%.
C07facthigh confidence
On 3 August 2026 A10 issued Microsoft a warrant for up to 800,000 shares at $0.01, in two tranches of 400,000 that vest only if Microsoft and its affiliates buy past purchase thresholds in periods ending 30 June 2027 and 30 June 2028, exercisable from 2028 and 2029. The thresholds are redacted, and A10 says it cannot yet estimate the accounting effect.
C14facthigh confidence
At 30 June 2026 A10 held $357.3M of cash and marketable securities against $225.0M of 2.75% convertible notes due April 2030, convertible at $23.46 a share; because the price stayed above 130% of that for 20 of 30 days, holders could convert in Q3 2026, so the notes were moved to current liabilities. No capped call is disclosed and the notes carry no maintenance covenants.
C21facthigh confidence
9.71M shares were sold short at the 15 September 2026 settlement, 13.7% of the float and 9.9 days of trading, up from 7.17M on 30 June 2026.
How we read the multiple
Buyers pay about 40 times A10's yearly operating profit for the business, after counting its cash and debt. Each dollar paid earns about 2 cents a year after tax.
To earn 6 cents a year after tax on each dollar, a common yardstick, A10 must:
- Grow operating profit about 2.6 times.
- Reach sales near $530M, against $310M today: about 11% growth a year for five years.
- Keep 25 cents of profit from each dollar of sales, against 16 cents today.
- Find that growth well beyond its one large buyer.
Against its own history. Buyers pay far more than their usual price for each dollar of A10's sales and profit, and that rise came from enthusiasm more than from sales.
The evidence under this - 4 claims
C17estimatemedium confidence
At $25.81 on 72.61M shares the market value is about $1.874B. Adding the $225.0M convertible and subtracting $357.3M of cash and securities gives about $1.742B, which is 34.8 times trailing GAAP operating income of $50.1M, 22.0 times non-GAAP operating income of $79.1M, and 28.6 times trailing free cash flow of $60.9M. The share price is 43 times trailing GAAP earnings of $0.60.
derivedClose of $25.81 on 2026-09-29 (S18) times 72.61M filed shares (C16); convertible at principal and cash and securities from C14; operating income from C12 and C13; free cash flow from C19; trailing GAAP diluted EPS of $0.17, $0.14, $0.17 and $0.12 (S04).
against this
- The vendor's enterprise value leaves out $302.7M of marketable securities and prints 40.0 times on the same line.
- Counting the convertible at its $247.5M share value rather than its $225.0M principal raises the multiple to about 35.2.
- 2026 guidance (C09) would lift operating income, so a forward multiple is lower than a trailing one.
C18estimatemedium confidence
For today's value net of cash to earn a 6% after-tax operating yield at a 21% tax rate, GAAP operating income must reach about $132M, 2.6 times the trailing $50.1M: for example revenue near $530M at a 25% margin, which is 11% a year for five years from $310M with the margin rising from 16%.
derived$1.742B (C17) x 6% / (1 - 21%) = $132M; $132M / 25% = $530M; ($530M / $310.1M) to the power of one fifth is 1.113. The 6% yield and the 25% margin are this thesis's own judgements, not the company's; 25% is the non-GAAP margin today (C13).
against this
- Management's non-GAAP margin is already 25.5% (C13); if share pay stayed at 7% of revenue, GAAP margin would reach 25% only with further operating leverage.
- At management's 12-14% growth (C09), revenue compounds faster than 11% for as long as the guide holds.
C12facthigh confidence
GAAP operating income was $9.0M in Q2 2026 (11.3% of revenue) against $10.3M a year earlier, and $50.1M over the four quarters to Q2 2026. Share pay was $9.0M in the quarter against $4.4M; R&D rose 30.1% and G&A 57.4%.
C19facthigh confidence
Free cash flow (operating cash flow less capital spending) was $18.1M, $16.0M, -$0.7M and $27.5M in the four quarters to Q2 2026, $60.9M in total.
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 operating profit, cheapest first:
| Ticker | Company | company value for each $1 of operating profit |
|---|---|---|
| NTCT | NetScout Systems | 17.7× |
| FFIV | F5 | 28.8× |
| ATEN | A10 Networks | 40.0× |
| RDWR | Radware | 82.7× |
The middle of the other 3 is 28.8. A10 Networks, Inc. is 2nd dearest of 4.
Rivals' figures come from a data provider, so read the spread of the group rather than any single pair.
Buyers pay more per dollar of operating profit than at F5 or NetScout, which sell similar products, and less only than at Radware; the premium is A10's own.
Private or foreign rivals that cannot be priced: NetScaler (Cloud Software Group), Imperva (Thales).
Where we disagree with the market
The market believes
A profitable, cash-returning networking company whose growth turned double-digit on AI data-centre demand.
We believe
The growth is one customer's rollout, partly paid for with a warrant, while sales to everyone else shrink.
The customer table that shows this sits in a note deep in the quarterly filing; the results announcement talks about growth from large companies, not the one buyer behind it.
We would be wrong if sales outside Customer A grew more than 5% for two quarters, or if its share fell while total growth stayed above 10%.
What analysts say, and the evidence
Analysts positive
57%
of 7, 1 Sep
Four months earlier
57%
of 7, 1 May
The evidence under this - 5 claims
C22factmedium confidence
The vendor's rating snapshot counts 11 Buy, 10 Hold and no Sell; its dated series on a panel of 7 fell from 71.4% positive in January 2026 to 57.1% by September.
C20facthigh confidence
The shares closed at $17.52 on 4 February 2026, reached a $38.49 high on 9 July 2026 and closed at $25.81 on 29 September 2026. They fell 5.3% the day after the Q1 2026 print (28.13 to 26.63) and 10.5% the day after the Q2 2026 print (30.84 to 27.60). Over 252 sessions they rose 41.0%, of which sales per share explained 8.4% and a higher multiple 30.0%.
C06estimatemedium confidence
Revenue from everyone other than Customer A was about $216M in 2023, $222M in 2024 and $215M in 2025, down about 3%; about $96M in the first half of 2026 against about $104M a year earlier, down about 8%; and about $49.7M in Q2 2026 against $48.6M, up about 2%. Customer A's revenue rose from about $39M in 2024 to about $76M in 2025 and about $59M in the first half of 2026.
derivedFiled revenue (C11, C02) times one minus Customer A's filed share (C04): 2023 $251.7M x 86%, 2024 $261.7M x 85%, 2025 $290.6M x 74%, H1 2025 $135.5M x 77%, H1 2026 $155.1M x 62%, Q2 2025 $69.4M x 70%, Q2 2026 $80.1M x 62%. The shares are rounded to whole percents, so each figure carries about half a point of revenue either way.
against this
- The CEO said on the Q2 2026 call that the business excluding Microsoft and one or two countries with macro problems is 'growing close to double digits'.
- Q2 2026 alone shows sales outside Customer A up about 2%, so the fall is concentrated in Q1 2026, when Customer A pulled a project forward.
C04facthigh confidence
One end-customer, Customer A, was 14% of revenue in 2023, 15% in 2024, 26% in 2025, 37% in Q1 2026 and 38% in Q2 2026 and in the first half.
C18estimatemedium confidence
For today's value net of cash to earn a 6% after-tax operating yield at a 21% tax rate, GAAP operating income must reach about $132M, 2.6 times the trailing $50.1M: for example revenue near $530M at a 25% margin, which is 11% a year for five years from $310M with the margin rising from 16%.
derived$1.742B (C17) x 6% / (1 - 21%) = $132M; $132M / 25% = $530M; ($530M / $310.1M) to the power of one fifth is 1.113. The 6% yield and the 25% margin are this thesis's own judgements, not the company's; 25% is the non-GAAP margin today (C13).
against this
- Management's non-GAAP margin is already 25.5% (C13); if share pay stayed at 7% of revenue, GAAP margin would reach 25% only with further operating leverage.
- At management's 12-14% growth (C09), revenue compounds faster than 11% for as long as the guide holds.
What could go wrong, and how we would know
The risks
- One customer, 38% of sales If Microsoft slows its orders once its current rollout ends, about a third of revenue is at risk in a single decision.
- The customer can build it Microsoft already runs its own traffic-balancing and attack-protection services, so it holds the bargaining power over price.
- Carriers are spending less Sales to phone and internet carriers fell 22.8% in Q2 2026, with Europe and Asia weak.
- Shareholders pay the partner The warrant and convertible bonds could add about 2.3% more shares, taking a slice of any rise.
- One distributor A single reseller carried 40% of Q2 2026 sales and 57% of money owed to A10.
The evidence under this - 5 claims
C04facthigh confidence
One end-customer, Customer A, was 14% of revenue in 2023, 15% in 2024, 26% in 2025, 37% in Q1 2026 and 38% in Q2 2026 and in the first half.
C34inferencemedium confidence
Because Microsoft already runs its own load balancer and DDoS service, it buys A10 only for the jobs where A10 is better or cheaper, and it has bargaining power: the warrant is a price A10 paid to keep and grow the business.
derivedMicrosoft's own services (C24); the warrant tied to Microsoft's purchases (C07, C08); the CEO's description of a 'completely different product set' (C10).
against this
- Microsoft expanded its A10 purchases from about $39M to about $76M in a year while running its own services (C06), so its own build did not stop it buying.
- A multi-year warrant also ties Microsoft's roadmap to A10's, which the CEO calls mutual performance commitments.
C15estimatemedium confidence
At $25.81 the convertible's 9.59M underlying shares are worth about $247.5M. If A10 pays the $225.0M principal in cash and the excess in shares, it issues about 0.87M shares; with the warrant's 0.8M, the two add about 2.3% to the 72.61M shares outstanding.
derived42.6257 shares per $1,000 on $225M (C14) is 9.59M shares; (25.81 - 23.46) / 25.81 x 9.59M is 0.87M; warrant 0.80M (C07); shares from C16; close from S18.
against this
- A10 may settle entirely in cash, and below $23.46 the notes issue no shares at all.
- Buybacks retired 3.67M shares in 2025, which would outrun this dilution at the same pace.
C03facthigh confidence
In Q2 2026 enterprise revenue was $48.0M against $27.8M a year earlier, up 72.9%, and service provider revenue $32.1M against $41.6M, down 22.8%. United States revenue rose 35%; APJ fell 7% and EMEA 16%.
C29facthigh confidence
A10's ten largest end-customers were 58% of Q2 2026 revenue against 46% a year earlier, and 40%, 38% and 33% of revenue in 2025, 2024 and 2023. One distributor carried 40% of Q2 2026 revenue and 57% of receivables at 30 June 2026.
What we still do not know
- How much Microsoft must buy for each half of its warrant to vest: the figures are blacked out.
- How the warrant will be booked, and how much it will reduce reported sales or profit.
- Whether Customer A is only Microsoft, and which of its products run on A10's equipment.
- How many of these boxes the world buys each year, and A10's share of them.
The evidence under this - 3 claims
C07facthigh confidence
On 3 August 2026 A10 issued Microsoft a warrant for up to 800,000 shares at $0.01, in two tranches of 400,000 that vest only if Microsoft and its affiliates buy past purchase thresholds in periods ending 30 June 2027 and 30 June 2028, exercisable from 2028 and 2029. The thresholds are redacted, and A10 says it cannot yet estimate the accounting effect.
C05inferencemedium confidence
Customer A is almost certainly Microsoft: analysts on the Q2 2026 call described the 37% customer and asked about Microsoft being so large a share, and the CEO answered about Microsoft without correcting either; Microsoft received the purchase-linked warrant two days earlier.
derivedThe Q2 2026 call (S05): one analyst linked the 37% customer disclosed in the Q1 10-Q to Microsoft, another asked how the company manages 'Microsoft being so large of a percentage of revenue', and the CEO answered both about Microsoft. The warrant (C07) vests only on Microsoft's purchases.
against this
- A10 never names Customer A in a filing; the 10-Q calls it only 'Customer A'.
- The CEO said Microsoft's newer business is booked as enterprise and its older business as service provider, so it could in principle be two customers of the same size rather than one.
C06estimatemedium confidence
Revenue from everyone other than Customer A was about $216M in 2023, $222M in 2024 and $215M in 2025, down about 3%; about $96M in the first half of 2026 against about $104M a year earlier, down about 8%; and about $49.7M in Q2 2026 against $48.6M, up about 2%. Customer A's revenue rose from about $39M in 2024 to about $76M in 2025 and about $59M in the first half of 2026.
derivedFiled revenue (C11, C02) times one minus Customer A's filed share (C04): 2023 $251.7M x 86%, 2024 $261.7M x 85%, 2025 $290.6M x 74%, H1 2025 $135.5M x 77%, H1 2026 $155.1M x 62%, Q2 2025 $69.4M x 70%, Q2 2026 $80.1M x 62%. The shares are rounded to whole percents, so each figure carries about half a point of revenue either way.
against this
- The CEO said on the Q2 2026 call that the business excluding Microsoft and one or two countries with macro problems is 'growing close to double digits'.
- Q2 2026 alone shows sales outside Customer A up about 2%, so the fall is concentrated in Q1 2026, when Customer A pulled a project forward.
What would change our view
- Unattractive→NeutralA cheaper price on an unchanged business
It needs:
- ○The business, after cash and debt, falls to 20 times its yearly operating profit.about ≈$15.58 today, 40% below the last close
- Unattractive→NeutralThe rest of the business grows again
It needs all of:
- ○Sales to everyone other than Customer A grow 5% or more on a year earlier.checked at the next results, due by 3 Nov
- ○Customer A falls to 30% of sales or less.checked at the next results, due by 3 Nov
We would also be wrong if
- Microsoft earns both halves of its warrant by buying more through June 2028.
- Operating profit reaches 20 cents per dollar of sales for a year while sales grow over 10%.
- Customer A's purchases fall on a year earlier in any quarter of 2027.
What is coming
- 3 November 2026 Q3 2026 results and 10-Q: Whether Customer A stays at 35% of revenue or more, whether sales outside it grow on Q3 2025, and whether the 12-14% 2026 outlook holds.
Every dated event ahead
- 3 November 2026 Q3 2026 results and 10-Q: Whether Customer A stays at 35% of revenue or more, whether sales outside it grow on Q3 2025, and whether the 12-14% 2026 outlook holds.
- February 2027 Q4 2026 results and the FY2026 10-K: Customer A's full-year share, the year's growth against the raised outlook, the first 2027 outlook, and whether buybacks resume beyond the first half's token amount.
- 30 June 2027 Microsoft warrant's first purchase period ends: Whether Microsoft bought past the redacted first threshold, vesting 400,000 shares; a vest says the rollout held, a lapse that orders fell back.
Since we wrote this
Written
30 Sep
Price the thesis read
$25.81
29 Sep
- 5 OctThesis reviewThe long case now sits under the verdictAfter the five-name pilot, the owner asked for the long case on every thesis.
- 4 OctThesis reviewOwner's checklist: revenue streams, cost of growth, management, catalystsThe owner's checklist found the method silent on who runs the company, what growth costs, where the money comes from and what is dated ahead; this pass fills those four blocks from the proxy and filings.
- 30 SepThesis writtenUnattractive, medium convictionAll of A10's growth is one customer, almost certainly Microsoft, that builds its own version; sales to everyone else shrank, and the price needs years more growth.
Also checked 1 time with nothing new to act on.
Sources 22
Where this came from
Sources - 22
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 filing2026-02-25Item 1 products, ACOS, maintenance terms, named competitors, Customer A at 26%, 15% and 14% of 2025, 2024 and 2023 revenue, ten largest customers, distribution channel shares, service provider and enterprise split, convertible terms, headcountbacksC01C04C11C14C19C23C28C29C33C38C42C43
- S02primarycompany filing2026-08-06statements, Customer A at 38% of Q2 2026 revenue, ten largest at 58%, one distributor at 40% and 57% of receivables, enterprise and service provider revenue, geography, convertible reclassified to current, TrojAI purchase, share pay, remaining performance obligations, the Microsoft warrant as a subsequent eventbacksC01C02C03C04C07C12C14C16C19C28C29C30C32C43
- S03primarycompany filing2026-08-05raised 2026 revenue and EPS growth outlook, non-GAAP reconciliation, buybacks and dividend, CEO statements on Microsoft and AI infrastructurebacksC02C09C13C33C36
- S04primarycompany filing2026-08-05quarterly revenue, margins and adjusted EBITDA back to Q1 2023, vertical and geographic split by quarter, GAAP to non-GAAP bridgesbacksC03C12C13C32
- S05primarytranscript2026-08-05Q2 2026 earnings call transcript (FMP), held 2026-08-05data/cache/transcripts/ATEN-2026Q2.jsonmanagement answers on the Microsoft agreement, pre-buys, pull-forward, the enterprise and service provider split, memory costs and the 2026 outlook; 14 analyst turns from 5 analysts, two of them on the size of the largest customer
Historically, we characterize it as service provider. We have not recategorized that portion of the business.
backsC09C10C31 - S06primarycompany filing2026-05-07Customer A at 37% of Q1 2026 revenue against 15% a year earlier; one distributor at 38%; enterprise at 56%backsC04C19
- S07primarycompany filing2026-08-05800,000 shares at $0.01, two tranches vesting on Microsoft's purchases in periods ending 30 June 2027 and 2028, exercisable from 2028 and 2029; the vesting schedule and exercise conditions in Exhibits E and F are redactedbacksC07
- S08primarycompany filing2026-08-07153,962 shares issued on 15 June 2026 as part payment for TrojAI Inc.backsC30
- S09primarycompany filing2026-02-042025 revenue and profit, buybacks of 3.67M shares for $68.9M, the first 2026 outlook of 10-12% revenue growth and 12-14% EPS growthbacksC09C11C16
- S10primarytranscript2026-04-28Q1 2026 earnings call transcript (FMP)data/cache/transcripts/ATEN-2026Q1.jsonguide held at 10-12%; an analyst question on whether large orders were pulled in ahead of memory shortages; management denies double bookingsread, but no claim rests on it
- S11primaryindustry data2013-04-01why carriers buy address-sharing equipment at all, and the behaviour a CGN must meet, from the standards body rather than a vendor
With the shortage of IPv4 addresses, it is expected that more Internet Service Providers (ISPs) may want to provide a service where a public IPv4 address would be shared by many subscribers.
backsC26 - S12primaryindustry data2026-09-29share of the world's internet users able to use IPv6: 43.70% on 2026-09-29, 41.29% on 2025-09-28, 38.82% on 2024-09-28, 35.28% on 2023-09-28
Use of IPv6 for World (XA) ... ['Date', 'IPv6 Capable', 'IPv6 Preference'], ... [new Date('2023-09-28'),35.28,34.23] ... [new Date('2024-09-28'),38.82,37.35] ... [new Date('2025-09-28'),41.29,39.95] ... [new Date('2026-09-29'),43.70,41.52]
backsC27 - S13primaryother2026-09-30Microsoft's own layer-4 load balancer, sold as a cloud service and run at scale in Azure
The service supports both inbound and outbound scenarios, provides low latency and high throughput, and scales up to millions of flows for all TCP and UDP applications.
backsC24 - S14primaryother2026-09-30DDoS protection for on-premises networks delivered from Cloudflare's network instead of a box on site
Instead of relying on local infrastructure that large DDoS attacks can overwhelm, Magic Transit uses the global Cloudflare Network to ingest and mitigate attacks
backsC25 - S15primarycompany filing2026-03-31Radware's competitor list, naming A10 in application deliverybacksC23
- S16primarycompany filing2026-05-14the Arbor DDoS competitor list, naming A10backsC23
- S17primarycompany filing2025-11-25the same application delivery and security product sold to the same buyers; F5 names no competitor by name, and warns of large cloud providers offering their own application deliveryread, but no claim rests on it
- S18secondaryvendor data2026-09-30ATEN research pack (FMP market tier, FINRA short interest, headcount), frozen 2026-09-30content/ATEN/research-pack.jsonclose of $25.81 on 2026-09-29, daily bars, the price move decomposition, analyst split and its history, forward estimates, earnings beats, short interest, cohort multiples for F5, Radware and NetScoutbacksC20C21C22C33
- S19primarycompany filing2026-09-30six open-market sales and no purchases in a year, five of them under Rule 10b5-1 plansbacksC35
- S20primaryother2026-09-30Microsoft's own network-layer DDoS protection service
protects at layer 3 and layer 4 network layers. For web applications protection at layer 7, you need to add protection at the application layer using a WAF offering.
backsC24 - S21primarycompany filing2026-08-05the warrant text: expiry, spend certification, change-of-control acceleration; Exhibits E (vesting schedule) and F (exercise conditions) redactedbacksC07
- S22primarycompany filing2026-03-10executive officers and their start dates, beneficial ownership at 26 February 2026, the 2025 cash incentive plan's metrics and payout, the 2025 and 2026 PSU price targets and RSU grants, the former CFO's resignationbacksC39C40C41