Every Q1 Criticism Answered, and the Multiple Re-Rated 48% to Pay for It

Results vs. Consensus

Cloudflare reported the June quarter after the close on August 6, 2026. Where the May print was a good quarter delivered alongside a shock, this one was a good quarter delivered alongside a raise, and the difference shows up in every line of the scorecard. Revenue of $696.061M cleared the Street's $666.4M by $29.7M and the company's own May guide midpoint of $664.5M by $31.6M, the third consecutive quarter of clearing guidance by more than $19M and the largest clearance of the three. Non-GAAP operating income of $96.111M beat the $90.0M to $91.0M guide by $5.6M at the midpoint. Non-GAAP EPS of $0.29 cleared both the guide and consensus by $0.02.

Q2 2026 Scorecard

Year-over-Year Comparison

Sequential Comparison

Revenue

Revenue of $696.061M grew 35.9% year over year and 8.8% sequentially, against 33.5% and 4.1% in the March quarter. That is the fastest year-over-year growth Cloudflare has printed since the first quarter of 2023, delivered at more than three times the revenue base, and it is the fact that decides how the rest of this report reads. The composition supports the number. Growth was broad across cohorts: management disclosed a record 986 net additions of customers above $100,000 in annualized revenue over the trailing twelve months, and said every cohort from $100,000 through $5 million set a record for year-over-year net additions. Revenue from customers above $100,000 reached 73% of the total, up from 71% a year ago.

The disclosure that resolves last quarter's open question is dollar-based net retention. In March it printed 118%, down two points sequentially, and management attributed the decline to an unusually strong new-customer contribution that does not enter the retention calculation. That explanation was arithmetically coherent and impossible to verify from outside, and we wrote at the time that a second consecutive sequential decline would matter considerably more than the first. Instead retention rose to 120%, its highest level in this cycle and six points above the year-ago quarter, while new-customer bookings simultaneously grew at their fastest rate in more than five years. Expansion and acquisition both accelerated in the same quarter, which is the combination the mix explanation predicted and the expansion-weakness explanation did not.

The geographic split is the one place where the quarter is less uniform than March. The United States contributed $354.447M, or 51% of revenue, and grew 40.7%. Europe, the Middle East and Africa contributed $187.039M and grew 30.4%. Asia Pacific contributed $99.444M and grew 32.3%, and other regions contributed $55.131M and grew 31.8%. International markets fell to 49% of revenue from 51% a year ago. In the March quarter the four regions grew within a three-point band of each other; this quarter the spread is ten points. American demand for agentic infrastructure is running ahead of everywhere else, which is plausible on the merits and is also a concentration the model should carry rather than ignore.

Margins

Non-GAAP gross margin of 73.1% is the first sequential improvement in eight quarters. That is a smaller number than it sounds like in isolation and a larger one in context. In isolation, 30bps is noise. In context, the entire bear case on gross margin rested on the claim that the decline was structural and open-ended, and management has now delivered one quarter that says it is not. The CFO attributed the stabilization to the same mechanism that caused the decline, running in the other direction.

"Paid versus free traffic on our network continued to grow year-over-year, again, driving additional allocation of network costs from sales and marketing into cost of revenue. However, as we discussed at Investor Day, this trend is showing signs of beginning to stabilize."
— Thomas Seifert, CFO

Year over year the picture is still negative: 73.1% against 76.3%, down 320bps, and GAAP gross margin down 312bps to 71.8%. The mix shift toward the developer platform is real and the developer platform carries a lower gross margin than the application-services business. What changed is the trajectory rather than the level, and management now guides the level to hold.

Operating margin is where the reorganization shows up. Non-GAAP operating expenses fell to 59.3% of revenue from 62.1%, and non-GAAP operating margin rose 240bps sequentially to 13.8%. The composition of that leverage is uneven and worth naming. Sales and marketing grew 27.6% against revenue growth of 35.9%, falling to 33.4% of revenue from 35.5%. Research and development grew 24.5%, falling to 14.9% from 16.3%. General and administrative grew 44.9%, rising to 11.0% of revenue from 10.3%, and it is the only expense line growing faster than revenue. Non-GAAP G&A excludes the acquisition-related costs and the legal reserve booked in the quarter, so this is the clean line, and it expanded in a quarter when the company removed a fifth of its workforce.

EPS and Cash Flow

Non-GAAP EPS of $0.29 grew 38.1% against non-GAAP operating income growth of 32.9%. The diluted share count used for the non-GAAP calculation was 373.683M against 365.264M a year ago, so share creation is running at roughly 2.3% annually and is a real, if modest, drag that the operating growth is outrunning.

The GAAP line deserves its own sentence because it is closer to mattering than at any point in this company's history. The reported GAAP net loss was $(169.981)M. Add back the $150.693M restructuring charge and the loss is $19.288M on $696.061M of revenue. Management framed the same arithmetic slightly more favorably on the call and used it to claim the 2028 GAAP-profitability goal set at the June Investor Day is now conservative.

Free cash flow of $56.384M at an 8% margin was up 69.4% year over year and down 32.9% sequentially. The sequential decline is entirely severance: the restructuring rollforward in the 10-Q shows $98.567M of cash paid in the six months, effectively all of it in Q2. Adding that back gives $154.951M of free cash flow, a 22.3% margin, which is the underlying cash-generation rate of the business this quarter and the most encouraging number in the release. It is also, necessarily, an adjusted figure the company did not publish.

Segment Performance

Cloudflare reports a single operating segment and does not disclose revenue by product line, so the two dimensions available to an outside analyst are geography, which the 10-Q discloses on a billing-address basis, and customer cohort, which management discloses on the call. Both are given below. The product dimension, which management organizes as a four-Act framework and which drives most of the commentary on every call, remains entirely unsized, and that gap is discussed later in this report.

Revenue by Region

United States

The American business grew 40.7% and now contributes more than half of revenue for the first time in the periods disclosed in this filing. That is a nine-point spread against the next-fastest region and a ten-point spread against the slowest. Every named customer win in the prepared remarks that carried a dollar value was American, European or APAC, including a five-year $31.8M application-services and Zero Trust contract with a digital-native media company, a five-year $7.7M Magic Transit and network-firewall expansion at a large United States federal agency, and an eighteen-month $15.9M contract with a Fortune 1000 technology company. The federal win is the most interesting of the three because it describes a displacement rather than an expansion: a legacy provider's outage locked more than 100,000 users out of a mission-critical system for days, and Cloudflare cut the agency's global rule-change time from a one-week service-level agreement to thirty seconds.

Assessment: American outperformance of this magnitude is a demand signal and a concentration risk in the same number. The demand signal is that the agentic re-platforming is happening fastest where the AI companies are, which is exactly what the thesis predicts. The concentration risk is that a business that grew uniformly across four regions in March now depends disproportionately on one, and United States federal exposure specifically carries budget-cycle risk that the rest of the base does not.

Europe, Middle East and Africa

EMEA grew 30.4%, the slowest of the four regions, and lost a point of revenue share. Management gave no regional commentary beyond the growth rates and did not address the deceleration relative to the United States. One of the disclosed wins was European: a Global 2000 technology company signing a three-year $11M application-services and Zero Trust contract to eliminate a stack of five incumbent legacy point solutions.

Assessment: 30% growth is not a problem in absolute terms and the consolidation-driven win described above is exactly the Act 2 motion the company wants. The watch item is relative: if the American spread persists at ten points for another two quarters, the agentic tailwind is a geographically narrow phenomenon rather than a global one, and the total addressable market framing that management uses gets harder to underwrite at full size.

Asia Pacific and Other

APAC grew 32.3% and other regions grew 31.8%, both a point or two below the corporate average and both losing or holding share. The one disclosed APAC win is instructive on the land-and-expand motion: a rapidly growing technology company signed a one-year $4M pool-of-funds contract for the Workers developer platform, building on an $8.7M application-services contract signed in the prior quarter, and now routes every request through a Cloudflare Worker using KV and Durable Objects as its routing and tenant-configuration layer.

Assessment: That single account is the clearest published illustration of the flywheel management describes: security and performance first, developer platform second, within twelve months. It is one anecdote and the company selects which anecdotes to publish, but the sequence and the timeline are specific enough to be checkable in aggregate through the cohort data, and the cohort data agrees.

Customer Cohorts and Retention

Management's framing of the cohort data was the most emphatic part of the prepared remarks, and unusually for this company it was quantified rather than characterized.

"In fact, every one of our large customer cohorts from $100,000 to $500,000 to $1 million to $5 million in annualized revenue added a record number of net new customers year-over-year in the second quarter."
— Thomas Seifert, CFO

Assessment: The cohort table is the strongest evidence in the release and it is the part hardest to explain away. A record in the $100,000 tier alone would be a volume story about the low end. A record in every tier through $5 million, with the cohort growth rate accelerating from 25.2% to 26.6% and retention rising six points year over year, is a statement that the enterprise motion survived the removal of a fifth of the workforce. That was the single largest identified risk coming out of the March quarter, and one quarter of evidence says it did not materialize. One quarter is not the whole answer: the reorganization was not substantially complete until September 30, and bookings damage from a thinned solutions-architecture and customer-success layer would show up two to three quarters after the savings, not one.

Key KPIs

Assessment: Two KPIs on this table do work the income statement cannot. The first is remaining performance obligations. Total backlog grew 38% against reported revenue growth of 35.9%, which is the cleanest available forward indicator for a business whose revenue recognition is becoming less ratable. Current RPO grew 35%, fractionally slower than both, and the gap is the point: the long-dated portion of the backlog is compounding faster than the near-dated portion, which means customers are signing longer rather than shorter. The second is revenue per employee, which rose roughly 27% sequentially on a headcount base that fell by about 800. That is the reorganization's arithmetic made visible, and it is the reason the operating-margin step-up has stopped being purely hypothetical.

Key Topics & Management Commentary

Overall Management Tone: Uniformly confident, and for the first time in several quarters confident about the same things on both sides of the P&L. The March call had two registers that never joined, a bullish demand story and a defensive reorganization story, with nothing connecting them; this call had one, and the numbers did the connecting rather than the rhetoric. Where management was less convincing was on the items it chose not to raise: the restructuring cost overrun was disclosed in a single sentence between two good numbers, and the capital-expenditure guide was reiterated without acknowledging that two quarters have now run at half the annual rate. Analyst pushback was almost entirely absent, which is itself a datum.

1. The Raise Is Finally Bigger Than the Beat

This is the number that separates this quarter from the last one, and unlike in March it was not discussed on the call because it did not need to be. In May, Cloudflare guided full-year revenue to $2,805M to $2,813M, a midpoint of $2,809M, after beating its own Q1 guide by $19.3M and raising the year by $19.0M. Rest-of-year revenue was therefore unchanged, and we called that the single most important fact in the release: a company describing agentic AI as the biggest tailwind in its history had declined to put any of it into the balance of the year.

The full-year guide is now $2,864M to $2,870M, a midpoint of $2,867M. The raise is $58.0M. The Q2 beat against the company's own guide was $31.6M. Rest-of-year revenue is therefore up $26.4M, which is the first time in this coverage that management has raised the part of the year it has not yet delivered.

The operating-income line moved the same way and by more. The full-year non-GAAP operating income guide went from $418M to $421M to $443M to $445M, a $24.5M raise at the midpoint against a $5.6M beat, so $18.9M of the raise is prospective rather than delivered.

Assessment: This is the most important change in the quarter and it resolves the criticism at the center of our May note. Implied second-half revenue growth went from roughly 28% to 30.1%, so the guide no longer embeds a deceleration that contradicts the narrative. It is worth being precise about what has been conceded: management moved the rest of the year up by 0.9% of full-year revenue, which is a real change of posture rather than a large change of number, and the second half still decelerates from 35.9% to 30.1%. Cloudflare has cleared its own revenue guide midpoint by $25.5M, $19.3M and now $31.6M in three consecutive quarters. Treating the guide as a floor is at this point the base case rather than the bull case.

2. The Operating-Margin Bridge Is Half Built, and Q3 Is Where It Has to Land

In March the entire margin story was prospective: 11.4% delivered, 13.6% guided for Q2, and a 17.0% implied second half contingent on a reorganization that would not be substantially complete until the end of Q3. We described that as an unearned raise. Q2 delivered 13.8%, twenty basis points above the guide and 240bps above the March quarter, which is the first installment paid.

Q3, H2 and FY 2026 revenue and operating income are guidance midpoints. The Q4 and H2 lines are residuals after subtracting reported Q1 and Q2 and, for Q4, guided Q3, from the full-year midpoint.

The step that remains is larger than the one just taken. Q2 improved 240bps sequentially; Q3 has to improve another 380bps in a single quarter, and Q4 another 70bps on top of that. The reorganization is the mechanism and, per the 10-Q, the company "expects to recognize the remaining charges in the three months ended September 30, 2026 as it substantially completes the Plan by September 30, 2026." Headcount ended Q2 at 4,700, so most of the payroll reduction is already in the base, and Q3 gets a full quarter of it where Q2 got a partial one.

Assessment: The bridge is now credible in a way it was not in May, because the first span held and because the headcount is actually out. It is still the entire investment case for the next two quarters. A 380bps sequential margin expansion is not something this company has ever done, and the risk is no longer that the savings are imaginary but that they arrive one quarter later than the guide says. A Q3 print at 16% rather than 17.6% would be a good operational outcome and a bad stock outcome at 36 times revenue.

3. The Reorganization Cost More Than Advertised

The single least-discussed disclosure of the quarter was delivered in one sentence between the EPS line and the free-cash-flow line. Full-year restructuring charges rose from the $140M to $150M guided in May, of which $105M to $110M was cash, to up to $165M of which up to $130M is cash. The 10-Q breaks the new estimate into $125M of cash severance and benefits, $35M of stock-based compensation and $5M of other costs.

"Excluded from these non-GAAP results were severance and other restructuring charges of $151 million for the second quarter, of which $99 million was paid in the second quarter. For full year 2026, we now expect severance and other restructuring charges of up to $165 million with up to $130 million expected to be cash related. While higher than initially anticipated, we prioritized speed of resolution, particularly internationally to ensure a continued focus on execution."
— Thomas Seifert, CFO

The Q2 charge alone was $150.693M, at the top of the range guided for the entire year fifteen weeks earlier, with more to come in Q3. Cash paid in the quarter was $98.567M and the remaining accrued liability at June 30 was $24.273M.

Assessment: A 10% to 15% cost overrun on a reorganization of this size is unremarkable in isolation, and the stated reason, paying to close out international consultation processes quickly rather than letting them drag, is both plausible and the right trade. Two things about it are worth holding onto. First, it happened inside a single quarter against a range set the quarter before, which says the May estimate was built fast. Second, the non-GAAP treatment now excludes restructuring entirely: in our May note we flagged that the non-GAAP definition as filed did not exclude severance, and it now does. That change was not announced or discussed; it appears in the reconciliation. It is a defensible accounting choice and it also means $150.7M of real cost sits outside every headline number management guided the Street to.

4. Free Cash Flow Acquires a Qualifier

In May the claim was unhedged: full-year free-cash-flow expectations were unchanged despite $105M to $110M of cash severance. We called that the most quietly aggressive statement on the call, because absorbing roughly $107M of unplanned outflow without moving a target requires either offsetting in-year savings of the same size or a prior expectation that carried that much cushion, and neither was stated. This quarter the same sentence came with a condition attached.

"Excluding the impact from the higher severance and other restructuring costs, our free cash flow expectations for 2026 remain unchanged."
— Thomas Seifert, CFO

The words "excluding the impact from the higher severance" were not there in May. The reported figures are consistent with the incremental cost being absorbed nowhere: free cash flow of $56.384M was down 32.9% sequentially, and operating cash flow fell to 16.9% of revenue from 24.7% in Q1.

Assessment: This is a small walk-back stated in the smallest possible number of words, and the reason it matters is that no full-year free-cash-flow dollar figure has ever been published for 2026. A target that is unchanged except for the parts that changed cannot be tested against anything. The underlying cash generation is genuinely strong: adding back the $98.567M of severance paid gives a 22.3% free-cash-flow margin for the quarter. The disclosure practice is the issue rather than the cash.

5. Gross Margin Stabilizes, and the Unit-Economics Substitution Continues

Gross margin was the cleanest bear point coming out of March: two consecutive sequential declines, 430bps below the prior-year quarter, explicitly guided lower, with management asking the Street to switch to an undefined "unit economic margin" deferred to an Investor Day that had no date. Two of those four conditions have now changed. The Investor Day was held on June 9 at the New York Stock Exchange, and gross margin rose 30bps sequentially, its first improvement in eight quarters. The other two have not: no level or back-series for unit economics has been published, and the answer to the direct question about second-half gross margin redirected to the new metric again.

"Well, as we said at Investor Day, the important metric for us to focus on is total unit economics because the behavior of gross margin across the various products is just very different. You can see with all the metrics that we published today that we are tracking ahead of everything we track. So we're quite confident that the total unit economics for the -- across all products is going to increase over the course of this year. Gross margin, I think, will stabilize around the level we have."
— Thomas Seifert, CFO

Assessment: "Gross margin will stabilize around the level we have" is the most useful sentence management has produced on this topic in a year, and it is a commitment that can be graded in October. The unit-economics framing remains a request rather than a disclosure: a company can ask investors to weigh a different metric only after publishing it with history, and after two quarters and an Investor Day, there is still no number. The right posture is to take the stabilization commitment at face value, model gross margin flat at 73%, and treat any further decline as a broken commitment rather than a mix explanation.

6. Non-Human Traffic Crosses 50%, and the Agentic Commerce Stack Gets Real

The headline disclosure of the prepared remarks was that more than half the traffic crossing Cloudflare's network in Q2 was not generated by a human, the first time that has been true. The CEO's framing of the trajectory was deliberately extreme and deliberately caveated.

"And to give you a sense of how this trend is playing out and with the big caveat that I have called it wrong at every point along the way, the -- if the current trends continue, we think in 5 years, nonhuman traffic will be as much as 1,000x as much as human traffic. In other words, humans will be a rounding error on the Internet, not because human traffic goes down, but that's just how fast we're seeing nonhuman traffic grow."
— Matthew Prince, co-founder & CEO

The forecasting history he cites is itself the evidence: in November 2025 the internal model put the crossover in the second half of 2027, in March 2026 it moved to the first half of 2027, and it happened in May 2026. Against that traffic backdrop, the company used the intra-quarter and early-Q3 product events, Content Independence Day and Agents Week, to announce the commercial layer: Monetization Gateway, which lets a customer sell any resource behind Cloudflare including a page, an API, a dataset or an MCP tool; wallets, which let a buyer pay autonomously through an agent; and cloudflare.pay, an identity and trust layer for agent-to-merchant transactions. A research pilot with OpenAI on content-creator compensation was announced alongside them.

Assessment: The traffic crossover is the cleanest confirmation available of the first pillar of our thesis, because it is a measurement rather than a projection and because Cloudflare is one of very few companies positioned to measure it. The commercial layer is a different matter and should be valued at close to zero today. It is pre-revenue, it is unpriced, it depends on industry-wide standard adoption that management itself frames as a critical-mass problem, and the CEO's own description of the engineering requirement is that the system must be three orders of magnitude larger than the world's largest payment network. Treat the traffic as thesis-confirming and the payment rails as a free option.

7. The Developer Inflection Doubles Again

Developers on the platform passed 7.4 million, with roughly two million added in the June quarter alone against one million in the March quarter and 1.5 million in all of 2025. The CEO's reaction to the figure is the most quotable line of the call and also a fair summary of how far outside trend it is.

"I was -- I made the team triple check the developer numbers because we added more developers in a quarter than we did in all of last year. And we thought last year was pretty good."
— Matthew Prince, co-founder & CEO

The stated drivers are that Workers is structurally well suited to agentic workloads, being lightweight, billed only for work performed, and fast to spin up and down; that the population of people writing code is expanding beyond professional developers through vibe-coding platforms; and that those platforms deploy to Cloudflare by default. Management named Lovable, Replit and Base44 at Wix as examples of where generated code is landing. Separately, more than 80,000 paying customers were added in the quarter, which the CEO compared to the company's entire paying-customer base at its 2019 IPO.

Assessment: This is the second thesis pillar and it is compounding faster than we modelled. The important qualifier is that developer count is an adoption metric, not a revenue metric, and Cloudflare still does not disclose developer-platform revenue. What did change this quarter is that management stopped describing Workers as an adoption play. Asked directly to size the book, the CEO declined to give a number but said the platform "has become a meaningful contributor to revenue" after a period of "playing for adoption" rather than revenue. That is a qualitative upgrade with no quantity attached, which is the pattern across most of this company's product disclosure.

8. VoidZero: A $164M Acquisition That Nobody Mentioned

On June 1, Cloudflare acquired VoidZero, Inc., a developer-toolchain company, for total purchase consideration of $164.2M comprising $98.7M in Class A stock and $65.5M of cash net of cash acquired. The purchase price allocated $143.0M to goodwill and $26.5M to intangibles, and goodwill on the balance sheet rose from $233.5M at March 31 to $376.2M at June 30 as a result. Separately, and larger than the cash component of the deal, the company entered into stock-based compensation arrangements with the acquired team worth $106.4M, of which $98.8M remains to be recognized over a weighted-average 3.9 years.

The transaction appears only in Note 13 of the 10-Q. It was not in the press release, was not mentioned in the prepared remarks, and no analyst asked about it.

Assessment: The economics are defensible. Roughly $271M of total consideration and retention for a toolchain team, in a quarter when the company added two million developers and is explicitly building a developer-platform moat, is a rational use of stock at a 36 times revenue multiple. What is not defensible is the disclosure asymmetry. The 10-Q states the acquisition "did not have a material impact on the Company's consolidated financial statements," which is true of the income statement in the stub period and not true of the balance sheet, where it moved goodwill by 61%. A company that builds most of its prepared remarks around the developer platform and spends no time at all on the developer-platform acquisition it closed five weeks before the call is choosing which parts of the story to tell.

9. Network Capex: The Guide That Has Not Moved While the Spend Has Not Started

Network capital expenditure ran at 7% of revenue in Q2 against 9% in Q1, and management reiterated a full-year guide of 14% to 15%.

"Network CapEx represented 7% of revenue in the second quarter. As a reminder, there can be some variability in this metric quarter-to-quarter, and we expect network CapEx to be 14% to 15% of revenue for full year 2026."
— Thomas Seifert, CFO

Purchases of property and equipment in the first half totalled $115.192M, or 8.6% of first-half revenue. Hitting 14% of the full-year revenue midpoint requires $401.4M for the year and therefore $286.2M in the second half, which is 18.7% of implied second-half revenue and 2.5 times the first-half spend. At the top of the guide, 15%, the second half needs $314.9M, or 20.6% of second-half revenue and 2.7 times the first half.

Assessment: One of two things is true and neither has been acknowledged. Either the guide is stale and full-year capital expenditure lands closer to 10% of revenue, in which case free cash flow is materially better than anyone is modelling and management has left that upside undisclosed for two consecutive quarters. Or the spend is genuinely coming, in which case second-half free cash flow gets compressed by roughly $170M relative to a straight-line assumption, in exactly the quarters where the margin step-up is supposed to be demonstrating itself. This was the same open question in March, the gap has widened rather than closed, and no analyst asked about it on either call. It is the largest unexamined variable in the model.

10. GAAP Profitability Moves Inside the Frame

At the June Investor Day, management set a goal of GAAP profitability by 2028 at the latest. Asked how the second-half guide tracks against it, the CFO reframed the target as conservative.

"I think the keyword was latest by end of 2028. We are pacing ahead of that very clearly without the impact of the restructuring charges, the GAAP net loss would have been around $18 million. So that is almost striking distance."
— Thomas Seifert, CFO

The filed arithmetic is close: the reported GAAP net loss of $169.981M plus the $150.693M restructuring charge leaves $19.288M. Stock-based compensation and related payroll taxes of $140.593M is the remaining obstacle, and it fell to 20.2% of revenue from 25.8% a year ago.

Assessment: This is a genuine change in the character of the business and it is underweighted in the current debate, which is entirely about non-GAAP margin. A company whose GAAP loss is 2.8% of revenue before restructuring, with stock compensation falling five and a half points as a share of revenue in a year, is approaching the point where the non-GAAP adjustments stop doing the heavy lifting. It also cuts the other way on valuation: the market is already paying 234 times full-year guided non-GAAP EPS, and the GAAP number will be lower than the non-GAAP number for years.

11. The Business Model Is Getting Harder to Forecast, and Management Said So

Asked about pool-of-funds contracts and the revenue-timing headwinds flagged at the Investor Day, the CFO gave the most useful modelling comment of the call.

"But as we said at Investor Day, the business model is also going to evolve. It's moving away from a purely ratable SaaS model towards a much more diversified mix of pool of funds, consumption-based structures and also what we call T-shirt sizes. As the business accelerates, that also leads to more and more customers burning through their T-shirt sizes faster, pool of funds are getting consumed faster and getting renewed. So we are really stacking these things on top of each other. And that leads to, you would call it, a higher standard deviation across the individual customers."
— Thomas Seifert, CFO

Three of the eight customer wins described in the prepared remarks were pool-of-funds structures, including a one-year $7.5M deal with a generative-AI company and a one-year $6M deal for agent container infrastructure.

Assessment: This is management pre-announcing higher quarterly variance, and it is the honest read on what consumption pricing does to a subscription model. It cuts both ways for the stock. On the upside, consumption revenue converts an agentic-traffic boom into revenue faster than ratable contracts do, which is part of why growth accelerated. On the downside, it removes the predictability premium that ratable software multiples are partly paid for, at exactly the moment the multiple is at its highest. The CFO's stated response, continued prudence in setting the guide, is the correct one and is also why the guide should keep being treated as a floor.

12. The Refusal to Join the AI Infrastructure Arms Race

The final question of the call asked why Cloudflare has not signed the multi-billion-dollar GPU infrastructure contracts its competitors are announcing. The answer was the clearest articulation of capital-allocation philosophy management has given.

"Sure, Patrick. So I think not all revenue is created equal. And if you're selling what is just commodity compute, if you're basically letting an AI company use your balance sheet,and your credit rating in order to buy servers that are the same as everybody else's servers. Then like that's just not an attractive business for us."
— Matthew Prince, co-founder & CEO

The supporting argument was that Cloudflare sells work performed rather than capacity rented, that hyperscaler GPU utilization is structurally low because the customer rather than the provider is responsible for scheduling, and that Cloudflare's isolate-based sandboxing is lighter weight than containers and therefore supports far higher utilization per dollar of hardware.

Assessment: This is the right answer and it is also the answer that makes the capital-expenditure question above more pressing rather than less. A company that declines to rent capacity, and instead monetizes utilization, still has to buy the hardware that utilization runs on. Management is describing a strategy of spending less capital per unit of revenue than the hyperscalers while simultaneously guiding capital expenditure to double as a share of revenue in the second half. Those two statements are not contradictory, but only one of them has been quantified.

Guidance & Outlook

Implied quarter-over-quarter ramp: the full-year revenue midpoint of $2,867.0M less reported first-half revenue of $1,335.816M leaves $1,531.2M for the second half, or 30.1% year-over-year growth against $1,176.5M in the second half of 2025. Deducting the $736.5M Q3 guide midpoint leaves $794.7M implied for Q4, which is 29.3% growth. The guide therefore embeds deceleration from 35.9% in Q2 to 31.0% in Q3 to 29.3% in Q4. That is a gentler slope than the one implied in May, when the second half was guided to roughly 28%, and management moved it up rather than letting the Q2 beat absorb it.

Street at: the Q3 revenue guide of $736.0M to $737.0M sits above the pre-print consensus trajectory, and the full-year raise of $58.0M is 1.8 times the $31.6M Q2 beat against the company's own guide. On operating income the ratio is even wider: a $24.5M full-year raise against a $5.6M beat. Both lines are second-half weighted, and Q3 carries the heavier burden of the two.

Guidance style: unchanged on revenue, and materially more forward-leaning on the balance of the year. Cloudflare has cleared its own revenue guide midpoint by $25.5M, $19.3M and $31.6M in the last three quarters, so the practice of setting a beatable number continues. What changed is the willingness to raise the unearned portion, which management declined to do in May and did in August. The margin guide remains the aggressive half: 17.6% for Q3 against 13.8% delivered is a bigger single-quarter step than the company has ever printed, and the guide gives itself no room to arrive late.

Analyst Q&A Highlights

How Non-Human Traffic Actually Monetizes

The opening question went to the mechanism rather than the milestone: if most traffic is now machine-generated, does that show up as new logos or as expansion, and at the large end of the base or the small end. The answer never addressed the segmentation and instead reframed the question around a new business model, distinguishing malicious non-human traffic, which Cloudflare blocks at no incremental charge as a matter of policy, from wanted agent traffic, which it wants to serve as efficiently as possible and eventually to charge for at the transaction level.

Q: "I'd love to dig into the changing profile of traffic and how that's monetized. You've talked about how the majority of traffic now is nonhuman, and we all see the inflection in traffic. Maybe the question is, do you see that driving more new logos or are existing customers buying more? And then from a different related lens, is it driving more business with large customers or smaller ones?"
— Saket Kalia, Barclays

A: "And I think that the business model of the Internet for the last 27 years has been largely defined by advertising and really defined by Google. I think the business model of the next 27 years of the Internet is going to be very different, and there's no company in a better position to define what it looks like than Cloudflare."
— Matthew Prince, co-founder & CEO

Assessment: The question asked where the money is coming from today and the answer described where it might come from in five years. That is not evasion so much as a genuine gap in the company's disclosure: the segmentation requested has never been published and could not have been answered from the podium. It does mean the single most important commercial question about the agentic thesis, whether agent traffic is a land motion or an expansion motion, remains unanswered after two calls.

Whether Gross Margin Has Actually Bottomed

The follow-up was the most consequential exchange of the call for anyone modelling the business, because gross margin had fallen for two consecutive quarters into this print and management had guided it lower. The response conceded the level and committed to it, which is more than the prior two calls produced, while continuing to argue that the metric itself is the wrong one.

Q: "it was great to see the stabilization in gross margins this quarter. And we talked a bunch about that at Analyst Day. But maybe you could just talk to us a little bit about some of the puts and takes there. And particularly, how you see gross margins sort of ebbing and flowing through the second half?"
— Saket Kalia, Barclays

A: "Gross margin, I think, will stabilize around the level we have. But as I said, more importantly, it is to focus on the unit economics. And here, we will continue to see expansion over the course of the second half of the year."
— Thomas Seifert, CFO

Assessment: A commitment to stabilization "around the level we have" is a gradeable statement and the first one available on this topic. It converts a structurally-declining-margin bear case into a flat-margin base case, which is worth several turns of multiple on its own. The second half of the answer is the recurring problem: a metric invoked as more important than the one being asked about, for the third consecutive quarter, still without a number.

Why Two Million Developers Arrived in One Quarter

The most-repeated statistic of the call drew the obvious question about durability: with abundant alternatives, what makes developers commit to this platform rather than sample it. The answer identified three drivers, two of which are about the platform and one of which is about the market, and the third is the one that matters most for the trajectory.

Q: "I think, Matthew, you said 2 million developers added this quarter. That was obviously impressive. I guess with so many options for developers and questions about open weight and open source models these days, what are the most important elements about Workers that gives developers the confidence to platform on it, really, as you know, the Internet is fundamentally changing to support machine-to-machine traffic?"
— Matthew Hedberg, RBC Capital Markets

A: "The first is that Cloudflare Workers is turning out to just be the perfect platform for building agents and agentic workloads. It's extremely lightweight, it's -- you only get charged for when it's actually doing work. You can spin things up and spin them down very, very quickly."
— Matthew Prince, co-founder & CEO

Assessment: The technical argument is credible and unfalsifiable from outside. The market argument, that the population of people writing code is expanding beyond professional developers and that generated code lands on Cloudflare by default, is the one that would explain a doubling of the run rate in a single quarter, and it is also the one most exposed to the platform partners changing their default deployment target. The disclosure gap persists: two million developers is an adoption metric, and developer-platform revenue has never been sized.

Whether Pool-of-Funds Renewals Have Cleared Their Revenue Headwind

A question rooted in the Investor Day materials asked whether the revenue-timing drag from early pool-of-funds renewals is now behind the business. The answer said yes on trajectory and no on predictability, and volunteered a warning about quarterly variance that had not been asked for.

Q: "with respect to the pool of funds, I think you had a chart there about some of the larger customers really early renewing on a pool of funds basis and some of the revenue timing and revenue headwinds associated with that. As we got into Q2 and as we look into the back half, do you think -- are we at a point where those customers have sort of leaned into usage and we're sort of past those initial revenue headwinds when it comes to those early pool of fund renewals?"
— Sanjit Singh, Morgan Stanley

A: "So the momentum over multiple quarters is clearly pointing up into the right. You don't -- you not only see this in revenue, you also see this in RPO and CRPO. But moving -- how we move from quarter-to-quarter becomes a little bit more difficult to forecast. That's why we continue to be prudent in terms of how we set the guidance."
— Thomas Seifert, CFO

Assessment: This is the most useful answer on the call for anyone building a quarterly model, and it explains the guidance philosophy better than any of the direct questions about conservatism did. Management is telling the Street that the transition to consumption pricing widens the distribution of quarterly outcomes and that the guide is being set with that in mind. The corollary is that beats should be expected to continue and that a single soft quarter should not be over-read.

Whether the Structural Growth Rate Has Permanently Reset Higher

After three years in a high-twenties to low-thirties band and a quarter that broke out of it, a natural question is whether the ceiling has moved. The answer stacked four independent arguments and then the founder-CEO added the shortest and most revealing one.

Q: "you all have been in this very consistent growth rate for the past 3 years now in the high 20s, low 30s. And if I think about the trends that are happening in your business, is there a scenario where the structural growth rate of Cloudflare actually ends up being higher over the next 3 years as all of these business try to come together?"
— Gabriela Borges, Goldman Sachs

A: "The one for sure is the innovation flywheel and that we address and disrupt a TAM that becomes bigger and bigger from $30 billion at our IPO to north of $300 billion at this point in time. I think the flywheels that come with Act 3 and now especially with Act 4 are accelerating. Every time we move to a new Act, the maximum deal size goes up by factors, maximum $1 million contracts at Act 1, double-digit million dollars at Act 2, hundreds -- triple digits in Act 3."
— Thomas Seifert, CFO

Assessment: The deal-size ladder is the most concrete framing management has offered for why the growth rate could reset, and the disclosed contract values in this quarter's prepared remarks are consistent with it, including a $31.8M five-year application-services contract that would have been at the top of the range two years ago. The addressable-market figure should be treated as directional marketing rather than analysis. The founder-CEO's addendum, that he remembers when growth rates were higher and would not object to getting back there, is the closest thing to a growth aspiration this management team has voiced, and it was delivered as a joke.

The Path to GAAP Profitability

The question tested the Investor Day commitment against the second-half guide. The answer moved the goalpost closer rather than defending it, and disclosed a number that appears nowhere in the press release.

Q: "At Investor Day, you set a goal of GAAP profitability by 2028 at the latest. Just based on the guidance that you gave for the second half, how are we tracking against that goal? And what are some of the dials that you have in terms of relative or incremental opportunities?"
— Jonathan Ho, William Blair

A: "I think the keyword was latest by end of 2028. We are pacing ahead of that very clearly without the impact of the restructuring charges, the GAAP net loss would have been around $18 million. So that is almost striking distance."
— Thomas Seifert, CFO

Assessment: An answer that pulls a target forward in the same breath as disclosing an unpublished figure is a confident one, and the arithmetic behind it survives checking. The dials cited, top-line acceleration, operating leverage and internal automation, are the same three that produced this quarter's margin step, so the claim is that the current trend simply continues. The one variable left out is stock-based compensation, which at $140.6M in the quarter is still seven times the residual loss and is the actual gate on GAAP profitability.

Sizing the Workers Book, and Whether Growth Is Reverting to Self-Serve

The most direct attempt of the call to extract a product-level revenue figure was declined, twice and explicitly, before the answer pivoted to the enterprise-versus-self-serve half of the question.

Q: "I was hoping you could put a little bit of a quantitative framing on how big the book of business, so to speak, has become for you on Workers and Workers AI combined, just so we can have a sense and feel for the magnitude of impact this is having on your top line in terms of proportion."
— Fatima Boolani, Citi

A: "So I will start out not fully answering your question and then Thomas can continue to not fully answer your question. I think we haven't broken out by product area, what our revenue or revenue growth rate is. But I can say that the Workers platform continues to be something that really drives an enormous amount of both new customer adoption and real revenue growth. I think for a while, we were saying that we were playing for adoption, we weren't playing for revenue. I think that we have -- that has tipped and we are now very much -- again, not trying to maximize every dollar by any means, but that Workers has become a meaningful contributor to revenue."
— Matthew Prince, co-founder & CEO

Assessment: The refusal was good-humoured and total, and the substantive content is a stated transition from adoption to monetization with no quantity attached. "Meaningful contributor to revenue" is an upgrade from prior language and cannot be modelled. Until the company sizes the developer platform, the fastest-growing and lowest-gross-margin part of the business is a residual that outside analysts infer from the gross-margin line, which is precisely why the gross-margin debate has been so hard to settle.

Opting Out of the AI Infrastructure Arms Race

The last question of the call asked why Cloudflare has not signed the multi-billion-dollar GPU contracts its peers are announcing, and paired it with a second question for the CFO about whether rising cyber risk was showing up in the numbers yet. Only the first half was answered; the call ended before the CFO spoke.

Q: "One of the decisions that you've made is not to participate in the AI CapEx kind of arms race. We're seeing hyperscalers and neoclouds sign these kind of massive GPU AI infrastructure contracts, including tonight, one of your competitors disclosing an aggregate nearly $3 billion AI infrastructure contract value. I guess, Matthew, what was your thinking in kind of not joining this AI infrastructure arms race?"
— Patrick Colville, Scotiabank

A: "The hyperscalers, the traditional first-generation clouds are in the business of buying a server and then trying to sell it back, lease it back and get 5 turns of revenue off of it. We're in a very different business, where we're selling actually work getting done. And then instead of it being up to the customer to get as much out of the server as possible, we need to do the work to get as much out of the underlying equipment as possible and build the scheduling and the efficiency and everything else that's out there."
— Matthew Prince, co-founder & CEO

Assessment: The strategic answer is coherent and consistent with a decade of this company's capital discipline. The structural point about the exchange is that a two-part question ended the call with one part unanswered, and the unanswered part was the one addressed to the CFO about whether a named demand driver is in the numbers or still in the pipeline. Nobody followed up, and management did not volunteer it in the closing remarks.

What They're NOT Saying

1. The VoidZero acquisition, in its entirety. A $164.2M purchase that closed June 1, plus $106.4M of retention compensation, appears only in Note 13 of the 10-Q. It was absent from the press release, absent from prepared remarks that were largely about the developer platform, and absent from Q&A. Goodwill rose 61% in the quarter as a result.
2. Any resolution of the capital-expenditure gap. Network capital expenditure has run at 9% and then 7% of revenue against a full-year guide of 14% to 15% that has been reiterated twice. The second half now requires roughly 19% to 21% of revenue, or two and a half times the first-half spend. Neither management nor any analyst raised it on either call.
3. A full-year free-cash-flow figure. Still none, for the third consecutive quarter, and the "unchanged" language now carries an exclusion for the higher severance. There is no published number against which either the claim or the exclusion can be tested.
4. Any run-rate savings figure for the reorganization. The cost is now disclosed to the million and has been revised upward. The benefit is disclosed nowhere. The only way to infer it is to back it out of the operating-margin guide, which is what the market is now doing at 36 times revenue.
5. A definition, level or back-series for "unit economic margin." Third consecutive quarter of being told this is the metric that matters, first quarter after the Investor Day where it was supposedly framed, and still no number. Gross margin is now committed to stabilizing, which makes the substitution less urgent and its continued absence more conspicuous.
6. The paying-customer base. The 10-Q states the company "ceased including this metric because it has become significantly less important to us over time," yet the CEO cited 80,000 additions and 74% growth from the podium. A growth rate quoted against a base the company has formally stopped publishing is not a disclosure, it is a headline.
7. Revenue by Act, or by product. The four-Act framework structures the strategy, carries an explicit gross-margin ranking and an explicit deal-size ladder, and was used to answer the pipeline question. No Act has ever been sized. The direct question this quarter was declined on the record.
8. Rule of 40. In May the CFO committed to "north of 46%" for 2026 and "north of 50%" for 2027. Neither number was mentioned this quarter. On the reported figures the metric is running at 49.7% using revenue growth plus non-GAAP operating margin, comfortably ahead of the commitment, which makes the silence odd rather than concerning.
9. Net quota-carrying sales capacity. In May, management said account-executive capacity was exempted from the reduction and that net capacity would still accelerate in 2026. There was no update, and no analyst asked, in the first quarter after the cuts were actually executed.
10. Regretted attrition among the roughly 4,700 who remain. No question asked, nothing volunteered, for the second consecutive quarter. Headcount was disclosed as a single number with no composition.
11. Whether rising cyber risk is in the numbers. The last question of the call put this directly to the CFO. The call ended without an answer and the closing remarks did not return to it.
12. The change to the non-GAAP definition. Restructuring and other charges are now excluded from non-GAAP operating income and non-GAAP net income. They were not excluded under the definition filed with the Q1 10-Q. The change was neither announced nor discussed; it is visible only by comparing the two reconciliations.

Market Reaction

• Pre-print setup: NET closed at $284.43 on August 6, up 44.3% year to date, up 32.0% over the trailing twelve months and up 5.8% over the trailing thirty days, against a 52-week closing range of $160.19 to $301.33. The S&P 500 was up 12.6% year to date over the same period. The stock entered the print roughly 6% below its 52-week closing high and having already recovered the entire May de-rating.
• After-hours move: shares rose approximately 15% to 17% in the August 6 extended session, with the move beginning on the revenue line and extending on the full-year raise.
• Next-day session: NET opened at $318.00 on August 7, an 11.8% upside gap, traded a $295.89 to $324.73 range, and closed at $300.27, up 5.6% or $15.84 on the day. Volume of 8.8M shares was 2.7 times the 30-day average of 3.2M.
• Index context: the S&P 500 rose 0.6% on the same session. The move was overwhelmingly idiosyncratic.
• Since the print: the stock climbed to a closing high of $330.83 on August 13, the day a $2.50B convertible-note offering settled, then gave the gain back over the following six sessions to close at $293.14 on August 21, roughly 3% above the pre-print level.

The shape of the reaction is more interesting than its direction. The stock gapped 11.8% higher and finished the session at less than half that gain, having traded as high as 14.2% up intraday. Selling into a gap on a broad beat-and-raise is what a fully-valued stock does: the buyers who wanted exposure to the quarter already owned it into the print, having bid the shares up 44% year to date, and the marginal seller was taking the multiple rather than doubting the numbers.

The two-week path since reinforces that reading. The August 13 high close of $330.83 coincided with the pricing of $2.50B of zero-coupon converts struck at a 60% premium, an event that mechanically brings convertible-arbitrage flow into the shares, and the entire move faded over the following six sessions. As of the August 21 close of $293.14, the stock is up about 3% versus where it traded the afternoon before the release, against a quarter that raised full-year revenue by $58M, operating income by $24.5M and EPS by six cents. The market has taken the beat and declined to pay more than the multiple it already had.

That is the entire investment problem in one sentence. In May, an excellent operating quarter was met with a 23.6% de-rating because the story changed. In August, an excellent operating quarter that fixed the story was met with a 5.6% session and a full round trip inside two weeks, because the multiple had already recovered before the print. The stock entered this release at 35.0 times full-year guided revenue on the new guide, against 24.4 times at the May close, so most of the reward for solving the March problems had been paid in advance.

Street Perspective

Debate: Has the Structural Growth Rate Reset Higher, or Is This a Cycle?

Bull view: The bull camp reads a 240bps acceleration to the fastest growth rate in three years, retention rising to 120%, record cohort additions at every tier through $5 million and backlog growing 38% as evidence that the agentic re-platforming has moved Cloudflare out of its three-year high-twenties band permanently. On this reading, the company is early in a demand cycle that the guide still under-describes, and the correct comparison set is no longer infrastructure software but the platforms that captured the last architectural shift.

Bear view: The bear camp answers that one quarter of acceleration, from 33.5% to 35.9%, is a thin basis for re-rating a business, that the guide itself decelerates to 29.3% by the fourth quarter, and that consumption pricing which accelerates revenue in a boom also decelerates it in a pause. Management's own comment about a "higher standard deviation across the individual customers" is cited as the company conceding the point.

Our take: The bulls have the stronger evidence and the bears have the better base rate. Acceleration in a business this size is genuinely unusual and the breadth behind it, four regions, every cohort, both new and expansion, makes a one-off explanation hard to construct. But the company itself is guiding to deceleration, and the burden of proof for a permanent reset is several quarters, not one. The practical answer is that the growth debate is no longer where the risk is; the multiple debate is.

Debate: Does the Second-Half Margin Step-Up Land in Q3 or Slip to Q4?

Bull view: The optimists point out that headcount is already down to 4,700, that the reorganization is substantially complete by September 30, and that Q3 therefore gets a full quarter of a cost base that Q2 only got a partial one of. Q2 beat its own margin guide by 20bps and expanded 240bps sequentially without the full benefit, so the 380bps required in Q3 is arithmetic rather than ambition.

Bear view: The skeptics note that the company has never delivered a 380bps sequential operating-margin expansion, that general and administrative expense grew 44.9% year over year in a quarter when a fifth of the workforce left, and that the restructuring already cost 10% to 15% more than budgeted, which is evidence that this plan is running hotter than management modelled. A Q3 margin of 16% would be a fine result and a 160bps miss.

Our take: The bulls are right that the savings are real and the bears are right about the timing risk, and at this multiple only the timing matters. The tell for us is the general and administrative line: it is the only expense category growing faster than revenue, it is the category most likely to absorb reorganization friction, and it is inside the non-GAAP number the guide is set against. We would put the probability of Q3 landing at or above 17.6% at somewhat better than even, and note that the stock is priced as though it were near certain.

Debate: Is 36 Times Forward Revenue Defensible for This Business?

Bull view: The bull argument is that the multiple should be read against a business with 36% growth, a stabilizing gross margin, an operating margin heading to 18% within two quarters, GAAP profitability inside two years and a call option on agentic payment rails that is in the price at zero. Against forward-twelve-month revenue rather than this year's, the multiple is closer to 28 times, which is not obviously expensive for the growth and margin combination.

Bear view: The bear argument is arithmetic. The stock has re-rated 48% in fifteen weeks, from 24.4 times full-year guided revenue at the May close to 36.1 times now, and it did so before the margin step-up was demonstrated and before the reorganization was complete. At 234 times full-year guided non-GAAP EPS, the equity is a duration instrument, and the sell-side target distribution running from roughly $310 to $400 has the lowest revised targets implying mid-single-digit upside.

Our take: The bull case is a fair description of the business and an incomplete description of the security. Everything the bulls cite is true and almost all of it is already in the price: the re-rating happened between May and August, in advance of the evidence, which is precisely why the print itself produced a 5.6% session and a full round trip. Our fair-value work below puts the probability-weighted twelve-month outcome within a few percent of the current price, with roughly 26% of upside in the bull case against 22% of downside in the bear. That distribution is a Hold, and it is a Hold that gets more interesting in either direction after the Q3 print rather than before it.

Model Update Needed

Valuation impact. At the August 21 close of $293.14 on 356.1M shares outstanding as of the 10-Q cover date, NET carries a $104.4B market capitalization. Net of $4,162.8M of cash and securities against $3,270.3M of carrying-value convertible notes, enterprise value is $103.5B. That is 36.1 times the full-year guided revenue midpoint of $2,867M and 234 times the full-year guided non-GAAP EPS midpoint of $1.255. Against the same metrics at the May 8 close, 24.4 times and 164 times, the equity has re-rated 48% on the revenue multiple in fifteen weeks. It entered this print at 35.0 times, so the print itself contributed roughly a point of multiple.

The August financing is close to neutral on enterprise value and worth carrying explicitly. The $2.50B of 2031 zero-coupon converts added roughly $2,462.3M of net proceeds against $2,500M of new principal, less $259.5M paid for capped calls struck to reduce dilution. The 2026 converts matured on August 15 with roughly $1,125M of principal outstanding at a conversion price of $191.34, which at any price above that level settles into equity rather than cash and converts roughly $1.1B of debt into approximately 5.9M shares. Net of both, the company holds more cash, carries a longer-dated liability structure at a conversion price 70% above the current share price, and has modestly more shares.

Fair value framework. We continue to anchor on enterprise value to revenue, the only forward metric management commits to in dollars, and roll the anchor to our FY27 estimate of $3,641M, which assumes 27% growth against the FY26 guide midpoint and sits below the current exit run rate implied by the Q4 guide. On that basis the stock trades at 28.4 times.

Weighting the base case at 50% and the two tails at 25% each gives a probability-weighted twelve-month value of roughly $304, or about 4% above the current price. For reference, the same framework applied to the current fiscal year rather than the next one puts fair value at $268 on a 33 times multiple, roughly 8% below the current price. Neither framing produces a return that beats the market with confidence, and neither produces a decline large enough to justify a negative rating against a business accelerating at 36%.

Thesis Scorecard Post-Earnings

The pillars below are the standing thesis established at initiation in May, graded against what this quarter's print and call revealed. The status tags are the ones carried in the thesis of record.

Overall: the thesis is materially stronger than it was in May. Every one of the four bull pillars is now confirmed on disclosed data rather than three of four, the fourth having moved from prospective to partially delivered. One of the three bear points has been challenged by the company's own numbers and downgraded, and a second has been challenged without being resolved. The one bear point that strengthened is the valuation point, and it strengthened by more than the operating case improved: a 48% re-rating against a 240bps growth acceleration and a $58M guide raise.

Action: hold. The operating business has answered every question we raised in May, and it answered them faster than we expected. The market answered them first. Our stated upgrade trigger in May was a third-quarter report showing operating margin above 16% with $1M-plus deal growth intact, and Q3 has not happened; buying ahead of it now means paying 36 times forward revenue for a 380bps margin step that this company has never before delivered in a single quarter. We would move to Outperform on a Q3 print at or above 17.6% operating margin accompanied by a fourth-quarter revenue guide above $800M, and to Underperform if Q3 margin lands below 16% or if the capital-expenditure guide is honoured in a way that visibly compresses second-half free cash flow without an offsetting raise.