Cloudflare Q3 2026 Earnings Preview: AI Revenue, Gross Margins & Guidance

 

Margins are already improving, just not on the gross margin line. Second-quarter non-GAAP gross margin fell 3.2 percentage points year over year to 73.1%, while non-GAAP operating margin fell only 0.29 points to 13.81%, meaning 2.91 of those 3.2 points were absorbed below the gross line. That is precisely Cloudflare management's argument: AI requests carry lower gross margin but also lower acquisition cost, and upsells such as sandboxes and AI gateways offset the difference at the operating level. Third-quarter guidance pushes that argument further, implying a non-GAAP operating margin near 17.58% against 13.81% in the second quarter and 15.3% a year earlier.

 

Results arrive after the US market close on November 5, with the call at 5:00 p.m. Eastern, corresponding to 06:00 Singapore time on November 6. The company's announced call time is 2:00 p.m. Pacific.

 

 

 

Third-quarter guidance calls for revenue of $736.0 million to $737.0 million, non-GAAP operating income of $129.0 million to $130.0 million and non-GAAP earnings of $0.34 per share. Consensus straddles that range, with MarketBeat's $737.16 million sitting $160,000 above the top of guidance and Zacks' $733.3 million sitting $2.7 million below the bottom. Revenue growth at the guidance midpoint works out to 31.04% year over year, below 34% in the first quarter and 36% in the second. Year-over-year gross margin compression is already narrowing, from 4.3 points in the first quarter to 3.2 points in the second on a non-GAAP basis, with the level recovering sequentially from 72.8% to 73.1%. Full-year guidance of $2.864 billion to $2.870 billion implies fourth-quarter revenue of $791.2 million to $798.2 million, roughly 7.9% sequential growth, above the 5.81% implied by third-quarter guidance.

 

 

The timing is settled, with one detail easy to miss. US daylight saving ended on November 1, putting Eastern time at UTC-5, so 2:00 p.m. Pacific corresponds to 06:00 Singapore time on November 6. Note that MarketBeat's page lists 5:00 p.m. Eastern in its main entry while a banner on the same page shows 6:00 p.m., so the company announcement governs.

 

The three bars for the quarter can be quantified in advance.

 

 

The second row is where the interest lies. Guided non-GAAP operating income growth of 50.6% is more than double the 31.04% revenue growth, implying an operating margin near 17.58% against an actual 13.81% in the second quarter. That step is not organic efficiency improvement but restructuring savings beginning to land, unpacked below.

 

The spread in the first row also matters. One house sits above the top of guidance and the other below the bottom, $3.8 million apart, so whether the quarter counts as a beat depends on which is used as the reference. The company beat the Zacks consensus in each of the last four quarters with an average surprise of 9.30%, and on that record a meaningful beat means something above $740 million.

 

 

This is the most important comparison in the quarter and the core answer to the question in the title.

 

 

The second-quarter release showed revenue of $696.1 million, up 36% year over year. Read side by side, the margin lines are unambiguous: non-GAAP gross margin fell 3.2 points while non-GAAP operating margin fell only 0.29, a difference of 2.91 points contributed by sales, research and administrative expenses declining as a share of revenue. AI workloads did compress unit gross margin, and scale effects in operating expenses absorbed nearly all of it.

 

The 16.5-point decline in the GAAP operating margin line needs separate explanation or it badly misleads. The second quarter carried $150.7 million in restructuring and other charges, equal to 21.7% of revenue. Excluding that item, the GAAP operating loss was $55.0 million or 7.9% of revenue, against 13.1% a year earlier, meaning the GAAP operating margin improved 5.2 points year over year excluding restructuring. The scale of the programme was flagged in the first-quarter 8-K, where the company set out plans to cut roughly 1,100 roles at an estimated cost of $140.0 million to $150.0 million, mostly in the second quarter. The $150.7 million actually recorded sits essentially at the top of that range, indicating the one-off charge is substantially complete, which is why the third-quarter operating margin guidance can step to 17.58%.

 

 

Setting the two quarters side by side reveals a fact that cuts against the narrative of continuous gross margin erosion.

 

Non-GAAP gross margin recovered from 72.8% in the first quarter to 73.1% in the second, up 0.3 points sequentially, while GAAP gross margin recovered from 71.2% to 71.8%, up 0.6 points. The year-over-year compression is narrowing as well, from 4.3 points in the first quarter to 3.2 points in the second on a non-GAAP basis, an improvement of 1.1 points.

 

Year-over-year and sequential comparisons point in opposite directions, and both are facts, differing only in reference point. The annual comparison runs against 2025, before AI workloads entered the cost structure at scale, so it remains negative. The sequential comparison runs against a quarter in which they already had, so it is positive. Judging the turning point means watching the sequential direction and the change in the compression rate, and both point to compression easing. One analysis holds that non-GAAP gross margin holding near 73% would support the investment case, while the company's own account is that it is raising the number of tokens a fixed amount of compute can generate, an improvement path running through engineering efficiency rather than pricing.

 

What to check on the gross margin line this quarter is therefore not whether it returns to 76% but three things: whether the sequential recovery continues, whether year-over-year compression narrows further, and whether the company quantifies unit compute cost. The first two can be calculated straight from the tables, while the third will only appear on the call.

 

 

The second-quarter customer data in the Zacks preview provides the entry point for judging growth composition: customers contributing more than $100,000 annually numbered 4,698, up 27% year over year, with dollar-based net retention at 120%, against revenue growth of 36%.

 

The relationship between those three deserves unpacking. Revenue growth of 36% exceeding large customer growth of 27% means per-customer contribution rose. Dividing second-quarter revenue by the large customer count gives roughly $148,200 per quarter against roughly $138,500 a year earlier, an increase of about 7.0%. That is a rough measure, since the numerator includes revenue from all customers rather than large accounts only, so it tracks total revenue relative to the large customer count rather than strict per-customer spend. The 120% net retention rate is the cleaner indicator, showing that expansion within the existing base is a principal source of growth.

 

Current remaining performance obligations grew 35% year over year, marginally below revenue growth of 36%. That relationship carries diagnostic value in a subscription business: cRPO growth persistently below revenue growth generally means the contracted backlog awaiting recognition is thinning in relative terms, while the two running close together indicates stable growth visibility. The current one-point gap counts as close and is not a warning, but a widening to five points or more in the third quarter would warrant revisiting growth durability.

 

The company did not disclose cRPO in dollar terms in the second-quarter release, giving only the growth rate, so the absolute figure awaits the 10-Q. Citations need to note that, to avoid presenting a growth rate as an amount.

 

 

This is where the quarter is most liable to be over-read, and three things need separating strictly: traffic, product and revenue.

 

On traffic, more than 50% of second-quarter network traffic was non-human, which the company attributes to AI agents. That is a usage metric, establishing that AI agents have become the dominant source of requests on the network, and it generates no revenue by itself.

 

On product, the company has in fact built a mechanism for charging AI agents. The Monetization Gateway was announced on July 1, 2026, letting customers charge for any resource behind Cloudflare including web pages, datasets, APIs and MCP tool calls, implemented through the x402 protocol and the HTTP 402 status code: the server answers a request for a paid resource with 402 plus the price, the accepted asset and where to pay, and the client pays and retries with proof. It extends Pay Per Crawl, which charged crawlers for content only.

 

On revenue, the actual state is far more restrained. The product entered closed beta on September 30, open only to eligible US-based sellers and buyers with other geographies planned. The publicly named production participants number four: Cloudflare's own AI Gateway, Ceramic.ai, Stocktwits and API2PDF. Payments are verified and settled through Coinbase's x402 Facilitator, settling on Base in USDC. Neither announcement discloses what fee Cloudflare takes.

 

The conclusion is therefore clear: the company has never disclosed an AI revenue line, the Monetization Gateway was not even in closed beta for most of the third quarter, and four production participants do not constitute a financial line item. Any conversion of the 50%-plus non-human traffic share into an AI revenue contribution fills an undisclosed financial line with a usage metric. What is worth watching in this release is whether the company quantifies that business for the first time, not substituting an estimate for it.

 

Incidentally, the x402 payment rail itself is extending across networks, and Bitcoin Lightning's integration with x402 shows the protocol is not bound to a single settlement asset, which matters for assessing the long-run path of agent commerce while bearing on none of Cloudflare's results this quarter.

 

 

Full-year guidance calls for revenue of $2.864 billion to $2.870 billion, non-GAAP operating income of $443.0 million to $445.0 million and non-GAAP earnings of $1.25 to $1.26 per share. Feeding in the known figures reveals what the fourth quarter has to deliver.

 

First-quarter revenue was $639.8 million and the second quarter $696.1 million, so adding third-quarter guidance leaves implied fourth-quarter revenue of $791.2 million to $798.2 million, a midpoint roughly 7.9% above the third-quarter guidance midpoint sequentially, against the 5.81% implied for the third quarter. On earnings, first-half actuals total $0.54, so the full-year midpoint of $1.255 requires $0.715 in the second half, leaving roughly $0.375 implied for the fourth quarter after the $0.34 guided for the third.

 

Those implied figures mean guidance embeds an accelerating fourth quarter. A view on this release should follow from that: if the third quarter merely meets guidance without a full-year raise, the fourth-quarter requirement becomes harder. The company beat consensus in each of the last four quarters by an average of 9.30%, while the Zacks Earnings ESP currently stands at 0.00% with a Hold rating, and the model offers no conclusive call on a beat.

 

Valuation is the constraint on the other side. The stock closed at $266.54 on November 4, down 2.59% on the day, with after-hours trading at $263.47, and is up 42.4% year to date against a 3.1% decline in its industry index. Forward price-to-sales is 27.88 times, 7.2 times the industry's 3.85 and 2.9 times the 9.49 of Zscaler, the highest-multiple peer, with F5 at 5.23 and Check Point at 5.28. At that multiple, meeting guidance generally does not move the stock, and a raise or a new quantitative disclosure is needed. Live quotes and after-hours moves for NET can be compared against the 06:00 timestamp on MEXC's NET stock page, since the price reaction to earnings concentrates in the first thirty minutes after release and reading the closing price afterwards discards most of the information. Equity and crypto markets on MEXC typically see short-lived shifts in liquidity and spreads through dense earnings windows.

 

 

The central tension in this release is not whether AI drives growth but whether the gross margin cost of that growth has been priced. The second quarter answers it fairly plainly: non-GAAP gross margin fell 3.2 points year over year while non-GAAP operating margin fell 0.29, with 2.91 points absorbed in operating expenses. That means management's argument about low gross margin alongside low acquisition cost holds up in the data, at least at current scale. Focusing on the gross margin line alone produces a conclusion opposite to the operating reality.

 

What the market is most likely to misread is the 17.58% operating margin implied for the third quarter. Jumping 3.77 points from the second quarter's 13.81% looks like an efficiency breakthrough, while the substance is restructuring savings from 1,100 roles beginning to land with the one-off charge already cleared. The $150.7 million recorded in the second quarter equalled 21.7% of revenue, and excluding it the GAAP operating loss narrowed from 13.1% of revenue a year earlier to 7.9%. How much of the third quarter's margin improvement is structural operating leverage and how much is a one-step gain from headcount reduction has to be separated through the expense detail on the call. Reading the headcount savings as an AI business margin improvement is the easiest error available.

 

What genuinely warrants watching is the sequential direction of gross margin rather than its year-over-year level. Non-GAAP gross margin has already recovered from 72.8% in the first quarter to 73.1%, with year-over-year compression narrowing from 4.3 points to 3.2, and both point to compression easing. Should the third quarter extend the sequential recovery and narrow compression further, the concern that AI workloads structurally depress gross margin can be downgraded from a continuing slope to a one-time step. Should the sequential direction turn negative again, then unit compute cost improvement is not keeping pace with AI request growth, and that is the case genuinely requiring revaluation.

 

The cross-asset lesson concerns matching valuation to certainty. A forward price-to-sales of 27.88 times, 7.2 times the industry, demands not merely meeting guidance but sustained beats plus raises. In a week when December Fed hike odds sit high and the 10-year yield runs near 5%, high-multiple growth names carry amplified sensitivity to the discount rate, and a business like the Monetization Gateway, still in closed beta with no disclosed fee and four production participants, cannot supply new support for that multiple this quarter. Multiple repair is more likely to come from the margin line, which is already verifiable, than from an agent commerce narrative that has yet to enter the financial statements.

 

 

 

After the US market close on November 5, 2026, with the call at 5:00 p.m. Eastern, which is 2:00 p.m. Pacific, corresponding to 06:00 Singapore time on November 6. US daylight saving ended on November 1, putting Eastern time at UTC-5. The live webcast and a replay are on the company's investor relations site at cloudflare.net, with the replay available about two hours after the call ends.

 

 

Revenue of $736.0 million to $737.0 million, non-GAAP operating income of $129.0 million to $130.0 million and non-GAAP earnings of $0.34 per share on roughly 374 million weighted average shares. Revenue growth at the midpoint works out to 31.04% year over year, below 34% in the first quarter and 36% in the second. The operating income guidance implies 50.6% growth and an operating margin near 17.58%.

 

 

The two houses straddle the guidance range. MarketBeat's revenue estimate of $737.16 million sits $160,000 above the top of guidance with earnings at $0.3424, while Zacks' $733.3 million sits $2.7 million below the bottom with earnings at $0.35. The $3.8 million spread means whether the quarter reads as a beat depends on the reference used. The company beat the Zacks consensus in each of the last four quarters by an average of 9.30%.

 

 

Principally because of higher network costs as paid traffic grew relative to free traffic, together with the lower unit gross margin on AI workloads. Second-quarter GAAP gross margin was 71.8%, down 3.1 points year over year, and non-GAAP gross margin 73.1%, down 3.2 points. The company's account is that it is raising the number of tokens a fixed amount of compute can generate, an improvement path running through engineering efficiency rather than pricing.

 

 

The data does not support that conclusion. Second-quarter non-GAAP gross margin fell 3.2 points year over year while non-GAAP operating margin fell only 0.29, with the 2.91-point difference absorbed by sales, research and administrative expenses declining as a share of revenue. Management's argument is that AI requests carry lower gross margin but also lower acquisition cost, with upsells such as sandboxes and AI gateways offsetting at the operating level, and the second-quarter data is consistent with it.

 

 

Because the second quarter carried $150.7 million in restructuring and other charges, equal to 21.7% of revenue. Excluding that item the GAAP operating loss was $55.0 million or 7.9% of revenue, against 13.1% a year earlier, an improvement of 5.2 points excluding restructuring. The programme cuts roughly 1,100 roles, and the company had flagged $140.0 million to $150.0 million of charges in the first quarter with most falling in the second.

 

 

The company has never disclosed such a line. The 50%-plus non-human traffic share is a usage metric rather than a revenue one. The Monetization Gateway, which charges AI agents, was announced on July 1, 2026 and entered closed beta on September 30 for eligible US-based parties only, with four publicly named production participants and no disclosed fee take. Converting a traffic share into a revenue contribution has no disclosure basis.

 

 

Through the x402 protocol and the HTTP 402 status code, where the server answers a paid request with 402 plus the price, the accepted asset and where to pay, and the client pays and retries with proof. Payments are verified and settled through Coinbase's x402 Facilitator, settling on Base in USDC. Sellers set which requests require payment, the price and where proceeds go, with pricing rules able to match the URL, headers or query parameters.

 

 

Full-year revenue guidance of $2.864 billion to $2.870 billion, less first-quarter revenue of $639.8 million, second-quarter revenue of $696.1 million and third-quarter guidance, implies fourth-quarter revenue of $791.2 million to $798.2 million, a midpoint roughly 7.9% higher sequentially against the 5.81% implied for the third quarter. On earnings, first-half actuals of $0.54 against a full-year midpoint of $1.255 imply roughly $0.375 for the fourth quarter.

 

 

Forward price-to-sales is 27.88 times, 7.2 times the industry's 3.85 and 2.9 times the 9.49 of Zscaler, the highest-multiple peer, with F5 at 5.23 and Check Point at 5.28. The stock closed at $266.54 on November 4 and is up 42.4% year to date against a 3.1% decline in its industry index. At that multiple, meeting guidance generally does not move the stock, and a raise or a new quantitative disclosure is needed.

 

 

This article is for general information only and does not constitute investment advice, financial advice, legal advice, tax advice or any trading recommendation. Prices of crypto assets, equities and other related financial assets can move sharply, and the historical data, company disclosures and third-party expectations cited here cannot guarantee future results, while the company guidance, consensus estimates and analyst ratings referenced do not represent actual outcomes and will change with the earnings release and subsequent revisions. Readers should conduct their own research and reach independent judgments based on their own financial circumstances, investment objectives and risk tolerance, consulting a qualified professional where appropriate. The MEXC Crypto Pulse team accepts no liability for any direct or indirect losses arising from use of the information in this article.

 

 

MEXC Editorial Team provides the latest platform updates, product launches, trading feature guides, and industry insights. With a focus on developments in the global crypto market, it delivers transparent information, professional data, and forward-looking analysis. Both beginners and experienced traders can find firsthand updates, in-depth observations, and official announcements related to the MEXC ecosystem.

 

 

 

Want the fastest access to MEXC's latest updates? Join our official Telegram group now!

MEXC Platform: Download App | Zero Fee Trading | Buy USDT Now

Join MEXC Community: X (Twitter) | Telegram | Discord

Get Insights: MEXC Learn | Terms of Service | MEXC Blog | MEXC News

Account Verification: Understand KYC | How to Complete KYC

External Content Platforms: Substack | Medium | Paragraph | LinkedIn | X(News)