Best Quarter Ever Meets a 32% Drawdown: The Price Finally Pays for the Story

Results vs. Consensus

Cloudflare reported the September quarter after the close on November 5, 2026, on a day when the S&P 500 fell 4.4% and the stock fell 8.3% into the release. The print itself had none of that day's character. Revenue of $782.513M cleared the company's August guide of $736.0M to $737.0M by $46.0M at the midpoint and the Street's $739.7M by $42.8M. Non-GAAP operating income of $137.084M beat the $129.0M to $130.0M guide by $7.6M at the midpoint. Non-GAAP EPS of $0.38 cleared the $0.34 guide and consensus by four cents, the largest EPS beat in our coverage.

Q3 2026 Scorecard

Year-over-Year Comparison

Sequential Comparison

Revenue

Revenue of $782.513M grew 39.2% year over year and 12.4% sequentially, against 35.9% and 8.8% in the June quarter. Growth has now accelerated two quarters in a row, to 35.9% and then 39.2%, after a flat first quarter at 33.5% against 33.6% in Q4 2025. Management described the fastest growth "in nearly four years," and it came on a revenue base above $3B annualized. The $46.0M beat against the guide midpoint compares with $25.5M, $19.3M and $31.6M in the prior three quarters, so the beat was nearly one and a half times the prior high.

The disclosure that matters most for the model is that the upside was not concentrated. The CFO attributed it to variable revenue across a long tail of customers rather than a single contract, and the CEO said growth "inflected in October of 2025, then accelerated again in July of 2026, and then again at the beginning of this October." That last clause is a statement about the current quarter, and it sits behind a fourth-quarter guide that already implies 37.6% growth.

The geographic split moved further toward the United States. American revenue of $407.912M grew 44.3% and reached 52% of the total, while EMEA and APAC both grew 31.9%. International revenue fell to 47.9% of the total from 49.7% a year ago. The one surprise in the table is the smallest region: "other" grew 46.4%, the fastest of the four, which the call did not mention.

Margins

Non-GAAP gross margin of 72.4% fell 70bps sequentially and 287bps year over year. In August the CFO committed that gross margin "will stabilize around the level we have," which was 73.1%, and we wrote that any further decline should be read as a broken commitment rather than a mix explanation. We hold to that. The 10-Q names the cause: cost of revenue rose $79.0M year over year, of which $43.0M was third-party technology services, $19.4M co-location and bandwidth, and $10.1M depreciation. Management's framing on the call moved from "stabilize" to "within our target long-term range of 70%-77%."

Operating margin is where the quarter delivered. Non-GAAP operating expenses fell to 54.9% of revenue from 59.3% in Q2 and 60.0% a year ago. The composition is the point. General and administrative expense fell to 7.4% of revenue from 11.0%, the 10-Q attributing part of the year-over-year change to a 14% reduction in G&A headcount. Research and development fell to 13.0% from 14.9%. Sales and marketing rose to 34.5% from 33.4% as non-GAAP S&M grew 16.3% sequentially; the 10-Q shows an $18.1M year-over-year increase in revenue share paid to partners inside GAAP S&M. Management is spending on sellers and partners, and it is taking the savings from the back office.

EPS and Cash Flow

Non-GAAP EPS of $0.38 grew 40.7% on 376.461M diluted shares, up only 0.4% year over year. The 2026 convertible notes matured in August, and the company paid the $1,293.4M principal in cash, issued about 2.2M shares for the conversion premium and received about 1.4M shares back from the capped calls. Dilution from that instrument is now behind the stock.

The GAAP line is the one to watch. The reported GAAP net loss of $(34.065)M includes $11.018M of restructuring and $10.667M of legal reserve and settlements. Excluding both, the loss is $(12.4)M, against $(18.3)M on the same basis in Q2. Management guided to GAAP profitability in the fourth quarter. The obstacle is stock-based compensation and related payroll taxes, which rose to $174.103M from $140.593M in Q2 and $119.699M a year ago, or 22.2% of revenue. That line grew faster than revenue year over year for the first time this year, and nobody mentioned it.

Free cash flow of $82.080M was a 10.5% margin, up 9.5% year over year and 45.6% sequentially. Purchases of property and equipment were $121.809M, or 15.6% of revenue, against $49.961M (7.2%) in Q2. Adding back the $24.4M of severance paid in the quarter gives $106.5M, a 13.6% margin, against 22.3% on the same basis in Q2. The capital-expenditure step-up we flagged as unexamined for two quarters has started, and it has compressed underlying free-cash-flow margin by about nine points sequentially.

Segment Performance

Cloudflare reports one operating segment and does not disclose revenue by product or by Act. The two dimensions an outside analyst can track are geography, which the 10-Q discloses on a billing-address basis, and customer cohort, which management discloses on the call. Both are below.

Revenue by Region

United States

The American business grew 44.3%, accelerating from 40.7% in Q2, and added $53.5M sequentially, about 62% of the company's $86.5M sequential increase. Several of the named wins were American: a cabinet-level federal department signed a five-year $10.3M contract for the SASE portfolio to replace an end-of-life VPN, and a Fortune 100 cloud infrastructure provider signed a thirteen-month $15.5M Magic Transit contract after the network stopped attacks its legacy DDoS appliances missed.

"When the world's biggest cloud operators build next-generation AI infrastructure, they trust Cloudflare to protect it."
— Matthew Prince, co-founder & CEO

Assessment: The spread between the United States and EMEA widened to 12.4 points from 10.3 in Q2. That is the agentic thesis showing up where the AI companies are, and it is also a concentration the model must carry. The federal contract adds a budget-cycle exposure that the rest of the base lacks, and the new Congress makes that exposure slightly less predictable than it was a quarter ago.

Europe, Middle East and Africa

EMEA grew 31.9%, up from 30.4% in Q2, and its share fell to 26% from 27% a year ago. The one disclosed European win was a five-year $8.1M application-services and Workers contract with a media company replacing a seventeen-year legacy CDN deployment, chosen for bot management that identifies AI crawlers and opens a path to pay-per-crawl monetization.

Assessment: EMEA accelerated, but its spread to the United States widened to 12.4 points, so the August worry that the agentic tailwind is geographically narrow is not yet answered. The media win is the cleanest illustration this quarter of the content-control motion turning into revenue, because the buyer's stated reason was the AI-crawler problem rather than performance or price.

Asia Pacific and Other

APAC grew 31.9% and sequentially by 12.3%, matching the corporate sequential rate. The "other" region grew 46.4%, the fastest of the four, from a small base of $60.7M. Management did not comment on either region.

Assessment: The APAC sequential number matched the company's, but it follows +0.8% in Q2, and the year-over-year rate eased for a second straight quarter, to 31.9% from 32.3% in Q2 and 34.4% in Q1. The region is steady, not yet keeping pace with the acceleration. The "other" growth rate is too small a base to build on, and it is the kind of line that can carry a single large consumption customer. We would not extrapolate it.

Customer Cohorts and Retention

The cohort data is strong where it counts and softer where the headline points. Retention rose for a second consecutive quarter to 122%, and the CFO said that the larger the customer's commitment, the faster its revenue is growing.

"Of particular note, we added almost as many $5 million plus customers in the first three quarters of 2026 as we had $5 million plus customers in total at the end of just last year."
— Thomas Seifert, CFO

Assessment: Retention and the top-tier comment are the strongest evidence that expansion, not acquisition, is doing the work. The $100,000 cohort's growth rate is the caveat. It decelerated to 24.9% from 26.6%, because a record 308 net adds only barely exceeded the 297 added a year ago. The company describes this as a record, and it is one. It is not acceleration. The revenue acceleration is coming from customers spending more, which is consistent with the pool-of-funds explanation and less consistent with a story about a broadening logo base.

Key KPIs

Assessment: Two lines on this table point in a direction the headline does not. Total backlog grew 5.3% sequentially against 12.4% revenue growth, and its year-over-year rate fell from 38% to 34% while revenue accelerated. Current RPO grew faster than total RPO, so duration is shortening, the reverse of what we saw in Q2. Together with the CFO's comment on variable revenue, that says the quarter's upside came from consumption against existing commitments. That is a better kind of revenue to have and a harder kind to forecast. The second line is stock compensation, which rose 23.8% sequentially and is the single input that decides whether the fourth-quarter GAAP profitability guide holds.

Key Topics & Management Commentary

Overall Management Tone: The most exuberant call in our coverage, and the least defensive. In August management was confident and careful; this quarter it called the results "monster" and the security outlook "scary" in the same breath, and leaned on adjectives where it had previously leaned on numbers. It was less convincing where it said nothing: the gross-margin commitment, the stock-compensation increase, the October compute commitments and the Workers AI data incident were all absent from prepared remarks, and only one of them drew a question.

1. The Biggest Beat of the Four, and the Raise Was Twice the Beat

In August the full-year raise was 1.8 times the Q2 beat, and we called that the first time management had raised the part of the year it had not yet delivered. This quarter the ratio widened again. Full-year revenue guidance moved from $2,864M to $2,870M to $2,963M to $2,964M, a $96.5M raise at the midpoint against a $46.0M beat. The implied fourth quarter rose from about $794.7M to $845.5M.

"We again accelerated growth and became more profitable at the same time, delivering our fastest revenue growth in nearly four years and achieving our highest operating margin in the history of the company."
— Matthew Prince, co-founder & CEO

Assessment: This is the strongest guide revision of the year by every measure. Implied fourth-quarter growth went from 29.3% in August to 37.6%, so the guide now embeds a much smaller deceleration than it carried in May and August. The guide is still conservative by construction. If Cloudflare clears Q4 by its recent average, full-year growth lands near 38%.

2. Growth Inflected Three Times, and the Third Was in October

The most forward-looking sentence of the prepared remarks was a dating exercise. The CEO placed inflections in the growth rate in October 2025, July 2026 and "the beginning of this October," and paired it with a claim about share of the web.

"To give you a sense, based on third-party data, between July 1st and today, we added approximately 5% of the entire web to Cloudflare. That's wild. It's more than the total market of all our reverse proxy competitors combined, added to our network in a bit over a quarter."
— Matthew Prince, co-founder & CEO

He attributed the acceleration to two forces: developers choosing an isolate-based platform over container-based clouds for agentic workloads, and a security environment in which being behind Cloudflare has become the default answer to a wave of AI-discovered vulnerabilities.

Assessment: The October inflection is the most useful forward datum on the call, because it was volunteered for a quarter the guide already covers. The 5% claim rests on unnamed third-party data and likely counts sites rather than dollars, so we treat it as directional. Both claims point the same way as the guide: the fourth quarter is starting stronger than the third did.

3. The Margin Step Landed in Dollars, and Missed by a Tenth in Percent

Our August upgrade trigger was a Q3 operating margin at or above 17.6% alongside a Q4 revenue guide above $800M. Non-GAAP operating income was $137.084M against a $129.5M guide midpoint, a $7.6M beat. The margin was 17.52% against 17.6%, because revenue came in $46.0M above the guide. At the guided revenue of $736.5M, the same operating income would have been an 18.6% margin.

"Third quarter operating margin was 17.5%, an increase of 370 basis points sequentially and an increase of 220 basis points year-over-year."
— Thomas Seifert, CFO

The fourth quarter is guided to $156.0M to $157.0M of operating income on $845.0M to $846.0M of revenue, an 18.5% margin, against the 18.3% the August guide implied. Full-year operating income guidance rose $18.5M at the midpoint to $462.0M to $463.0M, against a $7.6M beat.

Assessment: The risk we described in August was that the savings would arrive a quarter late. They arrived on time and in larger dollars than guided. We grade the trigger as missed on the letter and met on the substance, and we do not pretend otherwise; the rating section explains why we upgrade anyway. The 380bps step that the company had never delivered in a single quarter came in at 371bps.

4. Where the Leverage Came From: the Back Office, Not the Sales Floor

The composition of the 371bps is the part of the quarter the call did not describe. General and administrative expense fell from $76.3M to $57.5M, a 24.6% sequential decline worth about 360bps of revenue. Research and development fell 2.2% in dollars and about 195bps as a share. Sales and marketing rose 16.3% in dollars and 110bps as a share, and gross margin gave back 70bps.

"In the past, sellers might have spent 60% of their time getting prepared for meetings, filling out Salesforce reports, doing all of that. That's not what they love doing. What they love doing is being in front of customers."
— Matthew Prince, co-founder & CEO

The CEO's description of the reorganization matches the expense lines: fewer people who "measure things," more builders, more sellers. The 10-Q shows G&A headcount down 14% year over year, sales and marketing headcount up 2%, and an $18.1M year-over-year increase in partner revenue share inside GAAP sales and marketing.

Assessment: This is a better composition than a uniform cut would have been, because the growth engine is being funded while overhead shrinks. It also means the easy savings are now in the base. Q2's G&A was the anomaly we flagged in August, and Q3's G&A sits below even Q1's $63.6M. The next 100bps of margin has to come from gross margin or from sales efficiency, and gross margin is moving the wrong way.

5. Gross Margin Broke Its Stabilization Commitment

In August the CFO gave the most gradeable sentence management had offered on this topic in a year: gross margin "will stabilize around the level we have." It did not. Non-GAAP gross margin fell 70bps to 72.4%, and the prepared remarks reframed the number against a long-term range rather than against the commitment.

"Third quarter gross margin was 72.4%, representing a decrease of 70 basis points sequentially and 290 basis points year-over-year, and within our target long-term range of 70%-77%."
— Thomas Seifert, CFO

The 10-Q gives the cause more plainly than the call did. Cost of revenue rose 54% year over year against revenue growth of 39%, and third-party technology services costs were the largest single driver at $43.0M of the $79.0M increase. Over nine months that line is up $104.0M. The CFO's answer on unit economics continued to argue that Workers trades gross margin for a lower cost to book, and that the trade shows up in operating margin.

Assessment: We said in August that we would treat a further decline as a broken commitment, and we do. A 70bps miss is not large, and the operating-margin argument is fair as far as it goes. The concern is the cause. A rising bill for third-party technology services, in a quarter when developer-platform growth was the main driver, is what it looks like when demand outruns owned capacity. That is consistent with the October compute commitments in topic 11, and it means gross margin is now a capacity question as much as a mix question. Bear 2 moves back to EMERGING.

6. Variable Revenue Is the Upside, and Backlog Is Growing Slower Than Revenue

The CFO said the beat came from customers joining pool-of-funds structures or consuming their commitments faster, renewing early and re-signing larger.

"We saw continued strength in our variable revenue. It was the primary contributor to the upside. The surprising or positive part of that was it was driven across a lot of customers either joining pool of funds or, as you said, burning through their pool of funds commitments faster, and therefore have to renew earlier, and as part of the renewal process, sign up to even bigger pool of funds concepts."
— Thomas Seifert, CFO

Total remaining performance obligations of $2,876.8M grew 5.3% sequentially and 34% year over year, against 7.4% and 38% in Q2. Current RPO, 65% of the total, grew 37% year over year.

Assessment: This is the August warning about "a higher standard deviation across the individual customers" playing out on the favourable side. Consumption against pool-of-funds commitments converts backlog to revenue faster, which is why RPO slowed while revenue sped up. It is a sign of real usage and it makes the next two quarters more dependent on renewals. If consumption is running ahead of commitments, early renewals should push RPO growth back up in Q4. If they do not, part of this quarter's revenue was pulled forward from contracts already signed.

7. Ten Million Developers, and 158,000 Paying Customers in a Quarter

Developers on the platform passed 10 million at the October customer conference, from more than 7.4 million at the Q2 call. Paying-customer additions nearly doubled again.

"We added more than 158,000 paying customers this quarter. That's 110% growth in paying customers year-over-year and nearly double the previous record of more than 80,000 we added in Q2."
— Matthew Prince, co-founder & CEO

The CEO was candid that much of this cohort is small, citing a personal assistant building a first app for about $5 a month, and said the commercial effect runs through large existing businesses moving more developers onto the platform, which accelerates pool-of-funds consumption. He also said almost every pool-of-funds deal now includes Workers as a major component. The largest Workers win disclosed was a one-year $11M pool-of-funds contract with a fast-growing AI company running tens of millions of AI-generated applications on Dynamic Workers.

Assessment: The second pillar of our thesis continues to compound faster than we modelled, and the revenue link is now described specifically: developer adoption feeds pool-of-funds consumption, which feeds retention. It remains unsized. The paying-customer count is a growth rate quoted against a base the company stopped publishing this year, so we read it as a funnel indicator rather than a revenue one.

8. Security Demand From an AI-Driven Vulnerability Wave

The second growth driver management named was security, and the framing was the most alarming we have heard from this company. The CEO said frontier AI labs and security firms finding flaws now disclose them to Cloudflare early through a trusted-reporter process, so protection can be deployed for all customers before public disclosure.

"Just last week and the week before, we had six of the worst internet security bugs of all time announced over the course of fewer than 14 days. We're not done. We have already gotten disclosures of many, many, many more extremely scary bugs, which researchers have found, have quietly disclosed to us, and we have protected our customers from."
— Matthew Prince, co-founder & CEO

He said the wave would be "bad for, I think, two years," that the question of how to protect the internet was discussed at the White House with AI leaders, and that the answer came back "Cloudflare." The CFO said security is "playing a heavier and heavier role" in net-retention expansion.

Assessment: The early-warning position is a real moat and an unusual one, because it comes from the share of the internet already behind the network rather than from a product feature. It is also the claim on the call least supported by numbers: security revenue is not broken out, and the White House anecdote is not a contract. We give it credit as the explanation for the federal and financial-services wins and as a reason the large-customer cohort keeps expanding, not as a quantified driver.

9. Act 4: First Revenue, No Size

The CEO announced the first revenue from what the company calls Act 4, products that help create a new business model for the internet in which agents pay for the content and services they consume.

"I'm proud to announce that we recorded in Q3 our first revenue for what we've called Act 4 products, those where we're helping invent the new business model of the internet, and I expect we'll talk a lot more about that in the quarters to come."
— Matthew Prince, co-founder & CEO

In Q&A he described two mechanisms: making crawling more efficient, citing a paid arrangement with OpenAI under which Cloudflare signals when pages change, and eventually micropayments between agents and sites. He said he did not know how long it would take to become a major revenue source but would "imagine that this becomes the largest part of Cloudflare's business" ten years out.

Assessment: First revenue is a milestone and an unquantified one. The crawl-efficiency product is the near-term piece and has a named paying customer; the micropayment layer is a decade-out aspiration. We continue to value Act 4 at close to zero in the base case and as a free option in the bull case. The ten-year claim is the CEO's view of the destination, not a forecast we can model.

10. GAAP Profitability in Q4, and the Stock-Comp Line That Decides It

In June the Investor Day target was GAAP profitability "at the latest by 2028." In August the CFO said the company was "pacing ahead." This quarter it became guidance.

"We also expect to achieve GAAP profitability in the fourth quarter, significantly earlier as compared to our outlook from this year's Investor Day to reach this milestone at the latest by 2028."
— Thomas Seifert, CFO

The arithmetic is tight. The Q3 GAAP net loss was $(34.1)M, or $(12.4)M excluding restructuring and the legal reserve. The fourth-quarter guide adds about $19.4M of sequential non-GAAP operating income, and the restructuring charges end with the plan. That closes the gap only if stock-based compensation stops rising. It rose $33.5M sequentially in Q3, to $174.103M.

Assessment: This is a genuine change in the character of the business and a commitment we can grade in February. It is also the clearest case on this call of a target that depends on an undiscussed variable. A repeat of Q3's stock-compensation increase would by itself erase the margin for GAAP profitability. We treat the guide as achievable and fragile.

11. Capex Arrived, and $461M of Compute Commitments Followed

For two quarters we called network capex the largest unexamined variable in the model: 9% and 7% of revenue against a 14% to 15% full-year guide. In Q3 it ran at 15%.

"Network CapEx represented 15% of revenue in the third quarter. As a reminder, there can be some variability in this metric quarter to quarter, and we continue to expect network CapEx to be 14%-15% of revenue for full year 2026."
— Thomas Seifert, CFO

Purchases of property and equipment were $121.8M in Q3 and $237.0M for nine months, 11.2% of nine-month revenue. On the raised revenue guide, 14% to 15% of the year is $414.9M to $444.5M, which requires $177.9M to $207.5M in Q4: 21% to 25% of fourth-quarter revenue and 1.5 to 1.7 times the Q3 spend. Separately, the 10-Q discloses that in October the company "entered into several large compute and co-location agreements for an estimated total undiscounted future payments of $461.1 million, with contract terms ranging between 3 and 10 years."

Assessment: The second of the two outcomes we described in August is now the live one: the spend is coming, and it is compressing free cash flow. The guide has been honoured with an offsetting raise, so our August downgrade condition did not fire. The October commitments are the more interesting disclosure. On the same call the CEO said "We are not a financial engineering company" and described a market digesting "really dumb deals." Committing $461.1M of multi-year compute and co-location capacity is not that, but it is a material capacity decision, equal to about 16% of this year's revenue, and it was not mentioned.

12. A New Revenue Chief Six Months After the Reorganization

Mark Anderson, who led the go-to-market transformation, retires at the end of the year and stays on as an advisor for at least a year. Christian Smith joins as President of Revenue. The CEO introduced him on the call alongside the strongest go-to-market numbers the company has reported: record sales productivity, up year over year for an eleventh consecutive quarter, the highest ACV growth ever, the highest quota attainment in more than five years.

"With his well-earned retirement coming at the end of this year, we wanted to find a successor with a proven track record of scaling global sales organizations to turbocharge our already robust go-to-market momentum."
— Matthew Prince, co-founder & CEO

The CEO's description of what may change under the new leader was specific: more forward-deployed engineers at the largest customers, a different ratio of solutions engineers to account executives, and more product-specialist overlays.

Assessment: The timing is the risk. The workforce reduction completed on September 30, our August note put any lag effect on bookings two to three quarters after completion, and the sales organization now gets a new leader inside that window. The handover is orderly, and the go-to-market numbers have never been better. That is exactly when a transition is easiest to underrate. Bear 3 stays at EMERGING until the first two quarters under the new leadership report.

13. The Restructuring Closed Inside Its Cap

The plan announced in May is substantially complete as of September 30. Total charges were $161.7M against the "up to $165M" estimate given in August, with $123.0M paid in cash against "up to $130M" and a remaining liability of $2.6M. The Q3 charge was $11.0M.

"And I'm proud of the fact that we had an incredibly generous severance package. We continued to vest people's shares, which was something that most companies haven't done."
— Matthew Prince, co-founder & CEO

Assessment: A cost estimate raised in August and then honoured is a better outcome than the May-to-August overrun suggested. The CEO's comment about continued vesting explains the $36.1M of stock compensation inside the restructuring charge and is a reminder that part of the severance cost is paid in shares rather than cash. Commitment three from our August list is met.

Guidance & Outlook

Management's framing was the same short sentence twice: the company guides "to what we see." The CFO tied the fourth-quarter guide directly to the breadth of the third-quarter upside, saying that because it came from a long tail of customers rather than a concentrated event, it informed how much of it went into the guide. The CEO's October inflection comment is the qualitative support for the number.

Implied quarter-over-quarter ramp: the new full-year midpoint of $2,963.5M less nine-month revenue of $2,118.3M leaves about $845.2M for Q4, consistent with the $845.5M quarterly guide midpoint. That is 8.0% sequential growth after 12.4% in Q3 and 8.8% in Q2, and 37.6% year over year. Second-half revenue of about $1,628.0M would be up 38.4%. Full-year growth at the midpoint is 36.7%, against 32.2% implied in August and roughly 29.6% in May.

Street at: the only fourth-quarter consensus we found, a single source, was $798.2M before the print, and the guide midpoint is $47.3M, or 5.9%, above it and above the highest estimate in that source. Full-year consensus of about $2,869M was $94.5M below the new guide. Pre-print 2027 consensus of $3,690M will have to move: rolling its own 28.6% growth rate onto the new 2026 midpoint gives about $3,811M.

Guidance style: unchanged in method and more generous in result. Cloudflare has now cleared its own revenue guide midpoint by $25.5M, $19.3M, $31.6M and $46.0M in four consecutive quarters. The margin guide is less padded than the revenue guide, as Q3 showed: operating income beat by 5.9% while the margin landed a tenth below. Our working assumption is that Q4 revenue clears the guide by $25M to $45M and the margin lands within 50bps of 18.5%.

Analyst Q&A Highlights

What Drove the Beat, and How Much of It the Guide Embeds

The first follow-up of the call went straight to the gap between a 39% quarter and a 38% guide, asking what was driving both. The answer described breadth rather than any single driver and repeated the company's guidance philosophy, without saying how much of the third-quarter upside was carried into the fourth.

Q: "You grew 39%, you are guiding to 38%, so very little decel. I guess from your perspective, what are the most important drivers, not only to the quarter, but the guide, as we sort of look at the exit rate this year?"
— Matt Hedberg, RBC Capital Markets

A: "Yeah. Well, we always guide to what we see. You heard in both Matthew's and my script that the demand is driven by a highly diversified momentum."
— Thomas Seifert, CFO

Assessment: The answer is a list of signals (developer count, traffic, net new customers, record cohort adds) rather than a decomposition. That is consistent with a business whose upside is now usage-driven and hard to forecast from the inside as well as the outside. The guide likely embeds the base trend and, on the CFO's next answer, only part of the variable surplus, which is why we still expect another beat.

Pool-of-Funds Burn and How It Shapes the Forecast

A question tied two trends together: more revenue shifting to consumption, and customers burning through pool-of-funds commitments faster than planned. The answer named variable revenue as the primary source of upside and said explicitly that its breadth informed the fourth-quarter guide.

Q: "Talk to us about how that informs your forecasting going forward and any observations you can share with us on the pool of funds piece of the business in particular."
— Gabriela Borges, Goldman Sachs

A: "Because it has been driven by a long tail of customers, this, of course, influences how we build guidance for the fourth quarter, that it was not a concentrated event, but a long tail of events across many customers, and it is part of the reason why we guided to what we guided."
— Thomas Seifert, CFO

Assessment: This is the most important modelling disclosure of the call. It confirms that the upside is consumption against existing commitments, which explains why backlog slowed while revenue accelerated, and it says the company was willing to carry part of a usage surplus into the guide because it was diversified. The risk is symmetric: a usage-driven beat can become a usage-driven miss in a pause, and the guide is now less padded against that than it was in May.

When Act 4 Becomes Real Money

The question asked for a timeline on monetizing the new business model of the internet and the levers that could pull it in or push it out. The answer described two mechanisms, crawl efficiency already paid for and micropayments still to come, and declined to give a timeline.

Q: "Maybe give us a little bit of a sense for timeline on when you think that Act 4 can start to be monetized in real dollar amounts, and some of the key levers that could perhaps make that timeline take longer or perhaps pull in relative to whatever your base case is."
— Gabriela Borges, Goldman Sachs

A: "That is coming. I don't know exactly how long it will be before it turns into major sources of revenue, but if you look out in the more distant horizon, if you look 10 years out, I would imagine that this becomes the largest part of Cloudflare's business."
— Matthew Prince, co-founder & CEO

Assessment: An honest non-answer on timing and an ambitious answer on destination. The same response described the crawl-hint arrangement as one where the company is "recognizing significant amounts of revenue," which sits oddly against the prepared remarks' "first revenue." Until a number is published, the two statements cannot be reconciled from outside, and we keep Act 4 out of the base case.

What a Slower Data Center Buildout Would Mean

Asked on the day the market sold off on a possible slowdown in U.S. data center construction, the question asked whether such a slowdown would make an isolate-based architecture more valuable. The answer positioned Cloudflare explicitly against the financing structures behind the AI infrastructure buildout.

Q: "A lot can happen in terms of data center construction in the future, but hypothetically, if it were to slow, it feels like it could make your agentic cloud more valuable. I am curious how you think about that scenario and how it could impact Cloudflare's business, if at all."
— Saket Kalia, Barclays

A: "We are not a financial engineering company. We are focused on how we can actually deliver real operational excellence and get real advantages for every CapEx dollar that we spend in order to deliver more benefit. We have watched a bunch of really dumb deals get done. Obviously, the market is digesting those right now. Those just are not the sorts of deals that we would ever do."
— Matthew Prince, co-founder & CEO

Assessment: The strategic answer is consistent with everything this management has said about capital discipline, and in a market punishing AI infrastructure it was the right one to give. It does not address the company's own capacity position. Third-party technology services costs rose $43.0M year over year and the company signed $461.1M of compute and co-location commitments a month later. A scarcer-capacity world may favour the most efficient user of hardware, and it also raises the price of the hardware that user still has to rent.

The Gating Levers on Workers Unit Economics

The question recalled the Investor Day point that the developer platform carries a lower gross margin and asked what would bring its unit economics to or above the core business. The answer restated the trade-off and claimed the operating margin as the evidence, without a number.

Q: "Given the acceleration that you're seeing in this part of the business, could you remind us of what the core levers or gating levers you have to get Workers unit economics at or above the core business?"
— Sanjit Singh, Morgan Stanley

A: "Our Workers product falls into this category where we have slightly lower gross margins, but significantly lower cost to book, for example. This is directly reflected in the operating margin we are achieving, and this quarter was a good testament of that."
— Thomas Seifert, CFO

Assessment: The argument is coherent, and this quarter's composition partly undercuts it. The operating-margin gain came mostly from general and administrative expense, not from a lower cost to book, since sales and marketing rose as a share of revenue. The CFO also described "continued gross margin expansion" as an engineering lever in a quarter when gross margin fell. Fourth quarter running, and still no published unit-economics metric.

Whether the Vulnerability Wave Is Reaching Budgets

The question asked whether security budgets are inflecting in response to the wave of new vulnerabilities and whether giving protection away is the right go-to-market. The answer was the most vivid of the call and stayed qualitative.

Q: "Are you seeing cyber budgets inflect? What are CISOs doing right now or getting in terms of dollars to fix this never-ending avalanche?"
— Joseph Gallo, Jefferies

A: "I think right now what CISOs are doing is panic. It feels like the sky is falling."
— Matthew Prince, co-founder & CEO

Assessment: The answer described buyer behaviour (major banks signing contracts rather than relying on the free tier) without any budget or pipeline figure. Last quarter's version of this question, put to the CFO at the end of the call, went unanswered when time ran out; this quarter's got an anecdote. Two quarters in, the security tailwind is asserted and not yet measured.

Six Months After the Reorganization

The question asked for concrete evidence of how the May restructuring changed the way the company operates. The answer described the builders, sellers and measurers framework and made a headcount statement that will be checkable at the next print.

Q: "Just if you can give us some concrete examples of what the aftermath of some of those difficult restructuring moves has been in terms of your ability and your successes in operating the business at this level."
— Fatima Boolani, Citi

A: "Again, as I said at the time when we laid people off, I think our headcount's going to be larger by the end of the year than it was even before the layoffs. We're hiring like crazy. It's who we're hiring and the roles that we're hiring for that are changing."
— Matthew Prince, co-founder & CEO

Assessment: The statement is ambiguous on timing. In May the CEO said headcount would exceed 2026 levels again by 2027. If "the end of the year" means December 31, the company needs several hundred net hires in one quarter from a flat 4,700, which the fourth-quarter operating-margin guide does not obviously accommodate. If it means the end of 2027, it repeats the May commitment. We read it as the latter and will check the year-end headcount.

What the New Revenue Leader Changes

The follow-up asked where the new President of Revenue sees the greatest room to improve execution and whether the playbook changes. The answer was more specific than most on the call.

Q: "You highlighted doubling down on go-to-market execution. Where is the greatest opportunity to improve execution? Do you envision a continuation of the current playbook, or what tactical changes across enterprise sales or channel should we expect?"
— Joseph Gallo, Jefferies

A: "How many forward deployed engineers do we have that are out there, which we already have at our largest customers, helping them get the most out of Cloudflare? That certainly is going to increase. What's the ratio of solutions engineers to AEs?"
— Matthew Prince, co-founder & CEO

Assessment: More forward-deployed engineers and a richer technical overlay are cost-adding changes, which fits the rising sales and marketing line. They are also the right changes for a business whose largest customers are growing fastest. The model should carry sales and marketing at or slightly above 34% of revenue through 2027 rather than assuming further leverage there.

What They're NOT Saying

1. The Workers AI data exposure. On October 26 the company disclosed that a hosted model on Workers AI returned 155 responses containing other customers' data between August 27 and September 9, traced to a context-window configuration mismatch and an inference-engine bug. The call never mentioned it, and no analyst asked. The 10-Q reports no legal proceeding it considers likely to be material. For a company whose growth case rests on customers trusting it with agent traffic, the omission is the issue, not the scale.
2. The October compute commitments. $461.1M of undiscounted future payments under "several large compute and co-location agreements" with terms of three to ten years appear in the 10-Q commitments note only. Nothing in the prepared remarks or Q&A refers to them, on a call where the CEO criticized the AI infrastructure deals the market is digesting.
3. The gross-margin commitment. August's "stabilize around the level we have" was replaced by "within our target long-term range of 70%-77%." The change of reference point was not acknowledged and no analyst asked about gross margin directly.
4. Stock-based compensation. $174.103M, up 45.5% year over year and 23.8% sequentially, rising to 22.2% of revenue. It is the variable that decides the fourth-quarter GAAP profitability guide, and it was not mentioned.
5. A full-year free-cash-flow statement. In May the claim was "unchanged"; in August it was "unchanged" excluding the higher severance. This quarter there was no full-year statement at all, in the first quarter of the capex step-up. The company has still never published a 2026 free-cash-flow figure.
6. The fourth-quarter capex implied by the guide. Reiterating 14% to 15% on the raised revenue base requires roughly $178M to $208M of fourth-quarter spend, 21% to 25% of revenue. The CFO reminded the call that the metric varies quarter to quarter and moved on.
7. The slowdown in backlog growth. RPO growth fell from 38% to 34% year over year while revenue accelerated. The CFO read out both numbers without connecting them.
8. The large-customer growth rate. The $100,000 cohort grew 24.9%, down from 26.6%. Management described the 308 additions as a record, which they are, and cited 25% growth without saying the rate had slowed.
9. The legal reserve. A $10.667M legal reserve and settlement charge, excluded from non-GAAP results as a matter "outside the ordinary course of our business," is not identified in the 10-Q or the release.
10. The Deno Land acquisition. Closed October 8 for about $26.5M in cash and disclosed in the 10-Q subsequent-events note. Like VoidZero last quarter, a developer-platform acquisition did not come up on a call largely about the developer platform.
11. Partner-sourced revenue and the non-human traffic share. Both were headline metrics in August (31% and above 50%). Neither was updated.
12. The size of anything by product. Act 4 has first revenue, Workers is in "almost every" pool-of-funds deal, and security plays a "heavier and heavier role." None is sized.

Market Reaction

• Pre-print setup: NET closed at $244.51 on November 5, down 8.3% on the day as the S&P 500 fell 4.4% and the software ETF fell 4.4% in an AI-led selloff after the midterm elections. The stock entered the print up 24.0% year to date and 7.0% over twelve months, but down 31.1% over thirty days and 32.3% below its October 9 closing high of $360.93. The 52-week closing range was $160.19 to $360.93. The S&P 500 was up 7.4% year to date.
• After-hours move: shares rose roughly 17% to 19% in the November 5 extended session, to about $290.
• Next-day session: NET opened at $286.97 on November 6, a 17.4% gap, and that open was the session high. It traded down to $265.00 and closed at $268.75, up 9.9% or $24.24. Volume of 9.4M shares was 3.0 times the 30-day average of 3.1M.
• Index and peer context: the S&P 500 rose 0.3% and the software ETF slipped 0.1% on the same session. Zscaler fell 3.1% and CrowdStrike was flat, so the move was idiosyncratic by roughly ten points.
• Since our last note: the stock is 8.3% below the $293.14 August 21 close, against a 4.0% decline in the S&P 500 and a 4.3% decline in the software ETF over the same period.

The shape of the session repeated August's. Then the stock gapped 11.8% and closed up 5.6%; this time it gapped 17.4% and closed up 9.9%, opening at its high and giving back 6.3% from there. In August we read the fade as a fully valued stock being sold into good news. This time the setup was different, and the fade means something else. The stock had just fallen a third in four weeks for reasons unrelated to the company: rising yields in mid-October, a software selloff on fears that AI disrupts application vendors at the end of October, and an AI-infrastructure selloff after the election on November 5. The holders selling into the gap were most likely de-risking a high-multiple position in a falling tape rather than doubting the quarter.

The macro context matters for the thesis in a specific way. The November 5 selloff was about a possible slowdown in U.S. data center construction. Cloudflare fell with the AI-infrastructure group that day, and the question on the call about exactly that scenario got the answer that Cloudflare benefits from efficiency rather than from capacity. The market has not yet decided which side of that trade Cloudflare is on. The 10-Q's October compute commitments say the company itself is a buyer of capacity, if a disciplined one.

The investment point is the price. In August the stock traded at 36.1 times guided 2026 revenue after a 48% re-rating, and we said the multiple had already paid for the fix. It now trades at 32.3 times a guide that is $96.5M higher, and at 25.0 times our 2027 estimate against 28.4 times in August. The business improved by more than we modelled in August, and the price went down. That combination is what moves the rating.

Street Perspective

Debate: Is the Acceleration Durable, or a Consumption Spike?

Bull view: The bull camp reads two consecutive quarters of acceleration, retention at 122%, a fourth-quarter guide implying 37.6% growth, and an October inflection volunteered by the CEO as evidence that agentic workloads and the vulnerability wave have reset the growth rate. On this reading pool-of-funds burn is the mechanism working as designed, and early renewals at larger sizes will refill backlog.

Bear view: The bear camp points at the backlog line. RPO growth slowed to 34% from 38% while revenue accelerated, the upside was explicitly variable revenue, and the $100,000 cohort's growth rate decelerated. Consumption revenue that accelerates in a boom decelerates in a pause, and the guide now carries part of the surplus.

Our take: The bulls have the better near-term evidence and the bears have the right leading indicator. The October comment and the guide make a strong fourth quarter likely. The test is fourth-quarter RPO: if early renewals are real, backlog growth re-accelerates; if it does not, part of 2026 was borrowed from 2027 commitments. We model 29% growth for 2027 rather than extrapolating 38%.

Debate: Is Gross Margin a Mix Effect or a Capacity Cost?

Bull view: The optimists argue that gross margin is the wrong metric for a platform where the fastest-growing product trades gross margin for a much lower cost of sale, and that the 371bps operating-margin gain proves the trade works. A 70bps sequential decline inside a stated 70% to 77% long-term range is noise.

Bear view: The skeptics note that management committed to stabilization three months ago and missed, that third-party technology services were the largest driver of a 54% rise in cost of revenue, and that the company then signed $461.1M of compute and co-location commitments. Renting capacity to meet demand is a gross-margin cost that compounds with growth.

Our take: The bears have the better argument on this specific quarter, and the bulls have the better argument on the long run. The operating-margin trade is real, but this quarter's leverage came from G&A rather than from the cost to book. We model non-GAAP gross margin at about 72% into 2027, not the 73% we modelled in August, and treat a print below 71% as a reason to revisit the rating.

Debate: After a 32% Drawdown, Is NET an AI-Infrastructure Casualty or Beneficiary?

Bull view: The bull argument is that the stock was sold with the AI-infrastructure complex on a data-center-slowdown thesis that, if anything, favours the most capital-efficient platform. The multiple compressed while estimates rose, so the drawdown has created an entry point in a business that just guided to GAAP profitability.

Bear view: The bear argument is that NET is still a high-multiple software name in a tape that is de-rating high-multiple software, at roughly 200 times this year's guided EPS. The gap-and-fade on a large beat says sellers are still in control, and the company is itself a buyer of compute capacity in a market where that capacity may get scarcer and more expensive.

Our take: The bull side has moved from an argument about the business to an argument about the price, and the price argument is now on its side. The bear's tape argument is real and can persist for months; it is a reason to size the position moderately, not a reason to stay at Hold when our own framework shows nearly 18% upside. Our fair-value work follows.

Model Update Needed

Valuation impact. At the November 6 close of $268.75 on 358.4M shares outstanding at the 10-Q cover date, NET carries a $96.3B market capitalization. Net of $5,128.7M of cash and securities against $4,441.9M of carrying-value convertible notes, enterprise value is $95.6B. That is 32.3 times the full-year guided revenue midpoint of $2,963.5M and 199 times the full-year guided non-GAAP EPS of $1.35. On the same basis in August it was 36.1 times and 234 times. On our 2027 revenue estimate of $3,823M the stock trades at 25.0 times, against 28.4 times on our August 2027 estimate at the August price. At the print-day close of $244.51 it traded at 22.7 times 2027, within a turn of our August bear-case multiple.

Fair value framework. We keep the August framework unchanged, enterprise value to next-year revenue at 36, 30 and 22 times for bull, base and bear, and change only the estimate and the balance sheet. Holding the multiples fixed means the rating change comes from our numbers and the price, not from a new view of what the stock deserves.

Weighting the base case at 50% and the two tails at 25% each gives a probability-weighted twelve-month value of about $317, or 17.8% above the current price. In August the same weighting gave about $304, 4% above the then price of $293.14. Our fair value rose about 4% and the price fell 8.3%, which is the whole of the rating change. For reference, the current-year version of the framework, 33 times 2026 revenue, now gives $275, about 2% above the price, against 8% below in August. The base case requires the multiple to recover from 25.0 times to 30 times 2027 revenue, slightly above the 28.4 times it traded on that year in August. That is the assumption a reader should challenge first.

Thesis Scorecard Post-Earnings

The pillars below are the standing thesis carried since initiation in May, with Bear 4 added in August, graded against this quarter's print and call. The status tags are the ones in the thesis of record.

Overall: the thesis is stronger on the operating side and slightly weaker on quality. All four bull pillars are confirmed again, on harder numbers than in August. One bear point, gross margin, came back after one quarter of remission, and the capital-discipline bear point gained new evidence on both sides. The valuation bear point, the one that set our Hold, weakened because the price fell while the guide rose.

Action: upgrade to Outperform. We hold ourselves to the trigger we published in August, and on its letter it did not fire: the fourth-quarter guide of $845.5M cleared the $800M bar by $45.5M, but the Q3 margin printed 17.52% against 17.6%, a miss of eight basis points caused entirely by revenue landing $46.0M above guide. Our August downgrade triggers did not fire either: the margin was well above 16%, and the capex step-up compressed free cash flow but came with a raise of 2.1 times the beat. We are upgrading on the re-derived numbers. With our multiples unchanged, the probability-weighted value is about $317 against $268.75, and that is a stock we expect to beat the market. Conviction is moderate, not high, because gross margin, stock compensation and the new revenue leadership are all moving against us at the margin. We would move back to Hold on any one of three outcomes, each outside our own estimates: a fourth-quarter non-GAAP operating margin below 17.5%, meaning no step from Q3 against an 18.5% guide; an initial 2027 revenue guide implying growth below 25%, against our 29%; or non-GAAP gross margin below 71.0%, against our 72%. The triggers are independent, and any one of them is enough.