DK Street Journal

Atlassian's artificial-intelligence assistant does not have a price. Rovo Search, Chat and Agents are bundled into every paid Jira, Confluence and Jira Service Management subscription, and each licensed seat earns a monthly pool of credits set by plan tier — 25 on Standard, 70 on Premium and 150 on Enterprise, with a Rovo chat or agent request costing ten. As of May 2026 the company was not billing anything for usage above that allowance, and has promised at least 90 days' notice and an explicit customer opt-in before it does.

That makes Atlassian — the Sydney-founded vendor of Jira project tracking, Confluence knowledge pages and Jira Service Management desks, sold by the seat to more than 350,000 customers — the clearest test of the question hanging over all per-seat software: what happens to a seat meter when the software starts doing the work the seat was hired for. Its own risk register names the exposure. Its answer is not a meter but a ladder.

The upgrade is the product

Chief executive Mike Cannon-Brookes named the mechanism on the August 6 call: "the Teamwork Collection with approximately 10x as many Rovo credits included is a great motivator, inducement for upgrade in terms of hitting those limits." Customers on that bundle carry four to five times more paid seats per instance than stand-alone Jira or Confluence accounts. One product, Rovo Dev Standard, is priced outright at $20 per user per month; no revenue or customer figure has been attached to it.

What Atlassian publishes instead is adoption: Rovo in use at over 80% of the Fortune 500, Rovo-assisted actions up 50% sequentially, adopters growing annual recurring revenue more than twice as fast as non-adopters. It publishes no absolute paid-seat count and no net expansion rate. monday.com, which sells per-seat work management to overlapping buyers, discloses that its AI revenue doubled sequentially to roughly 1% of recurring revenue and 17% of net new, and that net dollar retention was 109%. An industry report from May 2026 recorded Atlassian's first-ever decline in enterprise seat counts; management's June-quarter language of broad-based seat expansion contradicts it, and no outside reader can adjudicate, because the number does not exist in the disclosure.

A very good year, guided in half

The business momentum is real. Quarterly revenue growth accelerated through fiscal 2026 to 27.6% in the June quarter, cloud grew 31% and remaining performance obligations rose 44% to $4.82bn. Operating income swung to $211.7m from a $28.5m loss. Most striking, gross margin expanded 381 basis points to 86.9% — where ServiceNow's fell 680 basis points to 70.7% and Elastic's slipped to 74.5%. Inference has not reached Atlassian's cost of revenue.

Next year it does. Fiscal 2027 is guided to about 13% revenue growth, recurring-revenue growth of 18% against 23%, and gross margin down to 84.5% explicitly on Rovo usage and hosting. Data Center revenue is guided to fall 17%, the reversal of a pull-forward Atlassian created itself when it announced in September 2025 that most Data Center products end in March 2029.

What the price now assumes

The shares rose 85% between July 29 and August 28. The August 7 earnings reaction supplied 35.8% and the August 27 sector session — Salesforce up 22.6%, ServiceNow 9.2%, the software complex outrunning the Nasdaq — another 10.3%. The remaining 21 sessions compounded to 23.5%, against a run of target increases beginning with Bank of America's upgrade to Buy at $175 and RBC's $215 initiation on August 22.

Price per dollar of trailing gross profit went from about 4.1 times at the end of February to 8.97 times — a 118% expansion while trailing gross profit grew 29%. monday.com trades near a third of that on faster growth; Microsoft, whose GitHub and Copilot stack competes for the same developer budget, trades at 17 times on a 68% gross margin. Forward earnings of 34.6 times rest on consensus 2027 earnings per share of $5.498, fractionally below the $5.505 just delivered.

The business earns the re-rating's first leg: accelerating cloud, a GAAP profit swing, and expanding margin in a quarter when peers' margins broke. What it does not earn is the second leg, priced for a year the company itself describes as half as fast and no more profitable, with the AI bill arriving in cost of revenue while the AI revenue arrives nowhere.

The September quarter is the first reported under that guide. It will show cloud growth, recurring revenue and margin. It will not show how many seats Atlassian sold.

The sell side has stopped debating how fast The Trade Desk can grow and started modeling how long it will shrink. Consensus now carries revenue falling 4.6% this year and a further 4.5% in 2027 at the Ventura, California company, which runs an independent demand-side platform — the software advertising agencies use to buy connected television, display, audio and mobile inventory on behalf of brands.

That reset followed the August 6 second-quarter call, and it dragged the price with it. HSBC's Mohammed Khallouf cut the shares to Reduce on 17 August with a $10 target, down from $20, sending them to their lowest level since January 2019; Guggenheim, Evercore ISI, DA Davidson, BMO and Scotiabank also downgraded, and the average target fell 43% to $13.89 — roughly where the stock actually trades. The stake now is the floor rather than the miss: a company that still generated $136m of free cash flow in the June quarter, at a 13.5% trailing free-cash-flow yield, is being priced as though the meter it charges against is permanently smaller.

The meter is running slower

Growth did not break in one quarter. It ran 17.7%, then 14.3%, then 11.8%, then 3.0% on June-quarter revenue of $715.1m. Third-quarter revenue is guided to at least $650m, about 12% below last year, with adjusted earnings before interest, tax, depreciation and amortization of around $160m against $241m in the second quarter. Operating income fell 13% year on year and net income 28.6% — profit contracting faster than the top line slows.

Underneath, the volume signals are uneven rather than uniformly bad. Connected TV grew double digits and is more than half of revenue; joint business plans with large advertisers grew 38%; accounts outside the top 500 grew more than 50%. The shortfall is concentrated in a handful of the biggest accounts, in consumer-goods and automotive categories that management put at about a quarter of platform spend, with some brands moving from premium decisioning to cheaper guaranteed buying.

The pricing story is not where it is usually assumed to be. Management's position is that its take rate has been flat for a decade, and that Kokai, its unified identifier and new artificial-intelligence tools went in without rate increases. The visible take-rate compression is at Criteo, the retail-media and performance advertising company, whose media spend grew 9% to $1.1bn last quarter while its own contribution net of traffic-acquisition costs fell 12% in constant currency. What squeezes The Trade Desk instead is the price of the media itself: Amazon made ads the default on Prime Video, adding roughly 50 billion impressions and taking its own connected-TV prices from about $35 per thousand views to $28 within a year, with buyers reporting broad declines of 10% to 30%. A percentage-of-spend meter runs slower on cheaper impressions — and Amazon's demand-side platform, which charges nothing on its own inventory, went from under 10% to just under 20% of global programmatic share in about fifteen months.

The macro alibi

"These numbers are not a reflection of our company or the long-term opportunity in front of us," chief executive Jeff Green told investors on the August 6 call, pointing to tariffs, inflation and a weaker consumer alongside the company's own execution. The word "macro" was used 14 times on that call. Meta grew revenue 28.0% in the same quarter. Magnite, the sell-side platform that monetizes the connected-TV inventory The Trade Desk buys, grew 11.2% with gross profit up 20.7% and operating income up 42.2%; its shares are up 76% since late February while The Trade Desk's have roughly halved. The impressions are being transacted. The independent buy side is not capturing them.

What the price embeds

The market values the company at about 2.7 times trailing gross profit, against roughly 21 times at the August 2025 peak — between Magnite at 7.1 times and Criteo, in outright decline, at 0.9. On cash the same ordering holds: 7.1x enterprise value to trailing cash earnings versus Magnite's 22.3x. Buybacks are real, cutting diluted shares 5.2% year on year, with $78m repurchased last quarter and $269m of authorization left against $1.5bn of cash. Green himself bought roughly 6m shares in early March at weighted average prices between $23.49 and $25.08 — about $150m, which he called the biggest personal purchase of his life.

On earnings the picture reverses. Consensus models $0.402 of earnings per share this year against $0.91 last year, which puts the shares at 33.8 times forward earnings and 16.2 times trailing — dearer on next year than on last. Nor has the chart resolved anything: 15 sessions since 10 August between $12.95 and $14.14, with closing lows of $12.95 and then $13.03, is a flat shelf rather than a sequence of higher lows, inside a downtrend unbroken since 13 November.

So the year-long de-rating is earned, and the case that it has overshot has to be made on cash flow alone. The cash multiple prices a business in permanent decline; the forward earnings multiple prices one whose profits halve and then stabilize. Both cannot be right, and the guide gives no help — management has said forward visibility is now a constraint on what it can forecast at all.

The company has told investors to expect less revenue in the September quarter than it booked a year ago. Whether Green's March purchase was early or simply wrong turns on how many quarters like that follow it.

Klaviyo added revenue faster than it added gross profit in the June quarter, and told investors the gap will widen through the rest of the year. Revenue reached $370.6m, up 26.4%; gross profit grew 21.3%, with reported gross margin falling to 72.6% from 75.7% a year earlier.

The reason is the meter itself. Klaviyo, the Boston platform that stores consumer profiles for e-commerce brands and sends their email and text campaigns, bills on the number of active profiles in an account plus the sends attached to them, with plans typically permitting ten times the send volume of the profile count, per its own billing documentation. Braze's annual report states that its pricing "is based in large part on the number of consumers that customers reach and the volume of messages customers send," in its fiscal 2026 Form 10-K. That design makes both businesses immune to the seat-count deflation repricing the rest of enterprise software — a shrinking marketing department does not shrink a retailer's contact database. It does not make them immune to cost of goods, because the fastest-growing unit of demand is a text message, and text messages are bought from carriers.

The meter is expanding

On the customer side the premise holds. Klaviyo disclosed more than 205,000 customers and 4,477 above $50,000 of annual recurring revenue, up 36% — roughly ten points faster than revenue, and about 40% of the total. Net revenue retention was 109%, absorbing a three-point headwind from lapping last year's enforcement of active-profile billing, a drag management said runs through the first quarter of 2027. Braze, a New York vendor whose software ingests app and web behavior and orchestrates cross-channel campaigns for consumer brands, reported dollar-based net retention rising to 110% and a fourth consecutive quarter of organic acceleration, with customers above $500,000 of recurring revenue up 33%. Its BrazeAI Decisioning Studio, sold separately, contributed $5.7m in the April quarter against revenue of $211.0m.

The pass-through is expanding faster

Klaviyo's non-GAAP gross margin of 73.4% was down about three points, which the company put down to text growth and carrier fees. "Text has strong unit economics, driven by its lower cost of acquisition and higher rates of expansion," chief financial officer Amanda Whalen told investors on the August 5 call. The near-term guidance says something harsher: gross margin steps down again in the third quarter, with a greater-than-normal seasonal decline in the fourth, as text grows faster than the company. A third-quarter pricing change shifts carrier fees onto customers and is expected to be neutral to this year's revenue and margin. At Braze, non-GAAP gross margin fell to 67.4% from 69.3%, which chief financial officer Isabelle Winkles attributed chiefly to premium messaging volumes and staffing tied to the AI product.

The seat-priced comparators do not escape, for different reasons. HubSpot grew June-quarter revenue 19.8% to $911.7m but added 7,000 net customers against an expected 9,000 to 10,000 and cut its second-half target to 5,000 to 6,000 a quarter, with retention at 102%. Sprout Social, which sells social-media management on seats, grew 10.8% and rallied on a roughly 20% headcount cut; its customers above $30,000 of recurring revenue grew 20% and now supply more than 61% of subscription revenue — the same upmarket mix shift, at a company with no per-message meter at all.

What the shares have and have not earned

Over the month to August 28 Braze rose 28.7% and Klaviyo 3.3%. Remove the August 27 and 28 sessions — Salesforce's quarterly beat and its deepened Anthropic partnership lifted the whole software complex, Salesforce itself up 22.6% in a day — and Klaviyo is down 8.0% on the month while Braze still holds a 13.9% gain. Klaviyo, whose shares fell 13.6% the session after its print, trades at 5.88x trailing and 5.34x forward gross profit, against roughly 4.5x in late May and a modest premium to HubSpot's 4.23x; Salesforce, growing 10.8%, is at 4.96x. Braze trades at 7.43x trailing gross profit against about 5.0x in late May — the dearest per gross-profit dollar of the five names here, on the lowest gross margin, having disclosed nothing since May 27.

The verdict splits. Braze's business earns the direction of its move: revenue accelerating for four quarters, retention rising, an AI product with a revenue line of its own. What it has not earned is the size of the re-rating, which arrived without a single new number. Klaviyo's meter is genuinely widening while its growth decelerates for a fourth straight quarter and its margin compresses, and its shares have gone nowhere on their own account. What both companies share is that usage pricing converts consumer message volume into revenue at a worsening exchange rate — gross-profit dollars grew about five points slower than revenue at each.

Braze reports after the close on Tuesday, September 8. It is the first disclosure since the shares re-rated, and the line to read is not the growth rate but what survives of it after the carriers are paid.

NetApp spent fiscal 2026 buying memory into a market that had roughly doubled the price of it, and still shipped a record year of all-flash arrays without giving up gross margin. Then, on May 28, the company guided the year ahead down.

The guide — non-GAAP gross margin of 68.5-69.5% for fiscal 2027, against 71.3% delivered — is worth roughly $149m of gross profit on NetApp's own $7.325-7.575bn revenue guide, and the mechanism reaches past one vendor: every maker of flash arrays buys the same dynamic random-access memory and NAND flash that artificial-intelligence servers are bidding away. NetApp's own risk-factor language now names "our ability to manage our gross profit margins, including managing component costs." The quarter management called the trough ended July 31 and is reported after the close on September 2.

The input

Conventional memory contract prices rose an estimated 93-98% in the first quarter of 2026 and a further 58-63% in the second, with enterprise solid-state drive contract prices up 53-58% and then 48-53%, as AI server demand absorbed supply and cloud providers locked the rest up under long-term agreements. The step-up is moderating rather than reversing: TrendForce expects server memory contract prices to rise a further 13-18% in the third quarter, with NAND up 10-15%.

Year one, absorbed

NetApp, which sells ONTAP data-management software, the all-flash arrays it runs on, and first-party storage services inside Microsoft Azure, Amazon Web Services and Google Cloud, took that shock without headline damage. Revenue growth accelerated through the year — 1.2%, 2.8%, 4.4%, then 12.5% in the April quarter — and full-year non-GAAP gross margin edged up to 71.3% from 71.1%.

The strain is visible one level down. Hybrid Cloud product gross margin fell to 56.3% in fiscal 2026 from 58.0%, about 170 basis points, on $3.19bn of product revenue. And the part of the business that would insulate NetApp from component prices is small. Public Cloud segment gross margin expanded to 85.7% in the April quarter from 68.4% two years earlier, but the segment is 9.3% of revenue and 11.4% of gross profit, and its revenue grew 3.5% for the year to $688m after the divestiture of Spot by NetApp. Keystone, the on-premises storage-as-a-service offering, booked $300m of total contract value, up 34% — bookings, roughly 4% of revenue. NetApp discloses no annual recurring revenue figure of any kind.

Growth came from boxes. All-flash revenue hit a record $1,216m in the April quarter, up 18%. That is the line memory prices.

What earnings were made of

Operating leverage carried the year: GAAP operating income rose 26.8% on revenue up 5.4%, while gross profit rose 6.2%. Roughly a third of the 12.1% growth in non-GAAP earnings per share came from a 3.8% reduction in diluted shares, to 201m, funded by $950m of buybacks alongside $413m of dividends. Billings grew 6.4% in the April quarter against revenue up 12.5% — the standard signature of demand pulled forward — though chief executive George Kurian told analysts that accelerated customer decisions had only a minimal effect on the quarter.

What the shares did

The 54% advance since May 21 was not a grind: a 12.4% session on the expanded Google Distributed Cloud air-gapped agreement and a 22.4% session on the April-quarter print together compound to 37.6%, leaving 66 other sessions worth 12.1%. Price per dollar of trailing gross profit went from about 4.96x at the May flip to 7.49x, roughly 51% of expansion against 5.7% growth in gross profit dollars. It still leaves NetApp the cheapest of the three box vendors — Everpure, the former Pure Storage, trades at 13.05x trailing gross profit on 35% quarterly revenue growth, Dell at 12.06x, and Dell's own April-quarter gross margin fell to 17.75% from 21.12%.

The verdict

The software meter is real and improving, but at a shade over a tenth of gross profit it cannot explain a 54% re-rating; the hardware line can, and the hardware line is the one being taxed. Chief financial officer Wissam Jabre described the July quarter as "more or less the trough" for product gross margin, with recovery dependent on price increases taking hold. That makes the coming print a test of pricing power, not of demand.

Consensus for fiscal 2027 sits at the very top of the range NetApp itself guided, which leaves the guided margin compression no cushion. On September 2 the trough stops being a forecast.

Accenture agreed this week to buy two more consulting businesses: McCoy & Partners, a Dutch specialist in SAP transformation work, on August 25, and COMWARE, a Tokyo technology-services provider bringing more than 180 professionals, on August 27. Both go into Accenture Edge, the mid-market unit. The Dublin-headquartered firm, which bills large enterprises for cloud, cybersecurity, data and application-modernization programs largely by consultant time, spent the best week of its year buying billable heads.

That is worth noting because the re-rating those purchases landed inside — Accenture's shares are up 38% from their July 15 low, Cognizant's 48% — rests on a proposition neither company will now quantify. Accenture told investors that the first quarter of fiscal 2026 would be the last in which it broke out advanced-AI bookings, on the grounds that the technology is embedded in nearly everything it sells. The final figure was $2.2bn, double the prior year, atop roughly $11.5bn booked cumulatively against $4.8bn of cumulative advanced-AI revenue. Cognizant has never disclosed a generative-AI bookings number at all. The central question — is AI adding contract value faster than it removes billable hours — is no longer answerable from either firm's reporting.

The meter, not the order book

The available proxy is Cognizant's, and it splits. The Teaneck, New Jersey outsourcer, which serves financial services, healthcare, products and communications clients from an offshore-heavy delivery base, reported trailing-twelve-month bookings of $29.1bn, up 5%, a book-to-bill near 1.3 — while bookings signed in the June quarter itself fell 6%. The quarter carried seven deals with total contract value of $100m or more. Longer contracts, not more of them.

What is being sold inside those contracts is also changing shape. Fixed-price and transaction-based work has grown for three consecutive years, and some clients now ask for AI-infused rate cards that embed pre-training and inference costs. "Human effort, platforms, software, frontier services, all bundled for an output which is not effort-based, but outcome-based," chief executive Ravi Kumar S told investors on the July 29 call. He also supplied the demand check: one in four Global 2000 companies have paused AI deployments, and roughly two-thirds have not yet realized measurable productivity gains.

Infosys, the Bangalore-based offshore provider, names the mechanism outright. Management confirmed real deflation on large-deal renewals, with clients demanding productivity give-backs at renewal and increasingly mid-contract, and guided fiscal 2027 constant-currency revenue to 1.5–3% — roughly half a point organic once a 1.7% acquisition contribution is stripped out. Its AI-first revenue reached 8.2% of the total, up from 5.5% two quarters earlier. Real, scaling, and not enough: total revenue grew 2.9%. Infosys trades at 15.0x forward earnings against 14.3x trailing, the market pricing earnings down.

What Accenture has actually shown

Accenture's May-quarter revenue of $18.72bn grew 5.6%, decelerating from 8.3% in February. New bookings were $19.3bn, down 2%. Gross margin was flat near 32.8%. Headcount rose 1%, to 798,739. Fourth-quarter revenue is guided to $17.75bn–$18.4bn. The one place AI leverage shows is unit economics: Accenture's staff generated about $91,500 of revenue each over the past year, up 5.7% on near-flat headcount, against Cognizant's $60,700, up 1.8% on a workforce 12,900 larger than a year ago. Accenture bought that leverage with 11,000 job cuts under a restructuring charged at $865m, chief executive Julie Sweet saying the firm is "exiting on a compression timeline" people for whom reskilling is not a viable path.

The shares have moved on none of this. Accenture has published nothing since June 18; the trigger was the late-July rotation out of AI infrastructure names, already documented here. The sell-side arrived after: Citi lifted its Accenture target to $190 from $135 on August 24 and TD Cowen to $173 from $151 on August 26 — the first already spent at Friday's $189.61 close, the second below it. Accenture's forward multiple has expanded from roughly 10x in late July to 13.7x, against 15.0x trailing and a 10.8% free-cash-flow yield; Cognizant sits at 11.1x forward against a 14–15x reference and 13.7x trailing. Both are cheap against their own history. Neither is cheap because anything improved.

Cognizant earns part of its move: a sixth consecutive quarter of adjusted operating-margin expansion to 16.0%, full-year adjusted earnings guidance raised to $5.70–$5.82, financial-services revenue up 12%, headcount down 900 sequentially, and its trend measures crossed into positive territory on August 26. Accenture earns almost none of it, because it has said nothing. That the offshore ADRs sat out — Infosys up 10%, Wipro flat — argues this is a re-rating of two US-listed names rather than an industry demand recovery.

Accenture's fiscal year closed August 31. The fourth-quarter print lands in late September, date not yet confirmed; last year's equivalent came on September 25. It will be the firm's first word in three months, and the first full-year bookings number with no AI line in it to check.

The biggest single day of Palo Alto Networks' advance this year was delivered by another company's earnings. On August 27 the shares rose 12.8% — the session after CrowdStrike reported July-quarter revenue of $1.47bn against $1.44bn expected and adjusted earnings of 31 cents against 29 cents, with Nvidia's results the same week adding to the impulse. The whole listed security complex moved together: Okta up 33.7%, CrowdStrike up 21.3%, Zscaler up 11.0%, Fortinet up 6.6%. The next session Palo Alto fell 3.2% to $370.68 while Check Point rose 3.2% and Tenable rose 2.1% — it gave back roughly a quarter of the move on its own.

That matters because Palo Alto has not published a financial statement since June 2, covering the quarter that ended in April, and the shares have roughly doubled since May 1 on no company disclosure at all. The company reports fiscal fourth-quarter and full-year results after the US close on Tuesday, September 1, with a webcast at 1:30pm Pacific. Consensus looks for about $0.98 a share on roughly $3.35bn of revenue, against a fiscal 2026 guide of $11.415bn to $11.425bn in revenue and $20.9bn to $21.0bn of remaining performance obligation.

What the last disclosure actually said

Palo Alto sells hardware and software firewalls plus a widening stack of security subscriptions to large enterprises and governments. April-quarter revenue of $3.002bn grew 31.1%, but $388m of it came from CyberArk and Chronosphere, the identity and observability businesses acquired in February, leaving organic growth near 14%. Gross margin fell to 67.6% from 74.2% two quarters earlier as purchase-accounting amortization landed in cost of revenue, producing a GAAP operating loss of $183m.

The contracted metrics point the other way. Next-generation-security annual recurring revenue reached $8.1bn, up 60%, while remaining performance obligation grew 36% to $18.4bn; net of the acquisitions, the recurring base still grew faster than the backlog — roughly 28% against 22%. A recurring line outrunning contracted obligations is what displacing incumbent products looks like; backlog outrunning the recurring line is what stretching contract length looks like. Palo Alto discloses neither average contract duration nor the size of its customer-financing book, so the second reading cannot be ruled out from public filings. Cash held up: adjusted free cash flow was $910m against $578m, and the trailing-twelve-month margin 38.5% — though the 37.5% fiscal-year guide implies a step down in the July quarter.

"Q3 was a standout quarter for Palo Alto Networks, with accelerating organic bookings growth as customers turn to us to secure their AI deployments at scale," chairman and chief executive Nikesh Arora said with those results.

The other two grew faster

CrowdStrike, which sells the cloud-native Falcon platform by subscription, posted a fifth consecutive quarter of acceleration to 25.8% growth, record net new annual recurring revenue of $333m, up 51%, and raised its fiscal 2027 net-new-ARR growth outlook to 34%. "We are in an arms race. AI is driving more cyber attacks. AI is driving more cyber spending," chief executive George Kurtz told investors on the August 26 call. It is also selling into CyberArk's socket: identity ARR of $585m, up 34%, with privileged account security up thirty-five times year on year.

Fortinet, whose FortiGate firewalls run on in-house silicon, grew June-quarter revenue 25.6% to $2.048bn with billings up 33% and product revenue up 52% to $773m at a record 38% non-GAAP operating margin. Its management told investors customers now view it as a platform provider comparable to Palo Alto and CrowdStrike. The firewall base is not quietly decaying; it is re-accelerating, and not only for Palo Alto.

The budget impulse behind all of this is external and datable. Anthropic's Mythos model release in June reset enterprise perception of AI-enabled attack capability — CNBC wrote on June 5 that it "rejuvenated the cybersecurity sector" — and Arora said "Mark my words, Mythos has increased the terminal value of the entire cybersecurity industry," citing 1,200 inbound companies and more than 800 customer meetings in six weeks.

What the price now assumes

At $302bn of market value, Palo Alto costs about $40 for every dollar of gross profit it earned over the past year and about $30 against the coming year's — roughly double the $20 it cost in early May. Fortinet costs about $20 trailing and $19 forward; CrowdStrike about $57 and $52. Its 311x trailing price-to-earnings ratio is amortization noise, not a valuation signal, since the quarter was a GAAP loss.

The sector's re-rating is earned by the sector: two of the three delivered accelerating numbers and one of them raised its forward guide. What Palo Alto has not yet earned is the portion of its own multiple that assumes it is the share gainer rather than a passenger. Consensus prices about 21% compounding into fiscal 2027; the last organic print was 14%. The displacement evidence is real and it is one quarter old, resting on a decomposition the company itself does not publish.

For four months the shares have moved on a rival's earnings, a chipmaker's, and one AI model release. Tuesday afternoon, Palo Alto has to supply its own.

Manhattan Associates has spent years moving customers of its warehouse, order-management and point-of-sale software from perpetual licenses onto cloud subscriptions, and the conversion is far enough along that the company's fastest-growing number never touches the income statement. Contracted revenue not yet recognized — what the company calls remaining performance obligation — reached $2.47bn at the end of June, up 23% from a year earlier. Reported revenue grew 9.3%, to $297.8m.

That gap is arithmetic, not spin. A license sale lands largely at signature; the same customer on a subscription pays over a multi-year term, so the meter compounds while the profit-and-loss statement lags. What turns the mechanism into a story is that the market accepted the argument all at once. Manhattan's second-quarter report on 28 July beat consensus on both lines — adjusted earnings of $1.39 a share against $1.32 expected, revenue against $287.7m — and the shares gained roughly a third across the two sessions spanning it, with press coverage attributing about 27 points of that to 29 July alone. At $225.83 the stock sits at a 52-week high on about 41 times consensus 2026 earnings of $5.50, against roughly 25 times at its 27 February close of $135.43. A full year of de-rating was undone in a month, for consensus earnings growth of 10.3% this year and 11.5% next.

What the meter says

The bookings side earns its keep. Cloud revenue reached $127m, up 26%, in a third consecutive record-bookings quarter, and management raised full-year guidance to revenue of $1.160–1.166bn, cloud revenue of $505.5m and backlog of $2.62–2.68bn — 18–20% growth, which the company expects to land toward the high end. Conversions of existing on-premise customers made up more than 40% of new cloud bookings and new customers more than 25%, so the pipeline is not purely a migration of the installed base. Excluding the license runoff, revenue grew 13% rather than 9%.

The socket is contested. Manhattan sells warehouse management against Blue Yonder, SAP, Oracle and Körber, and ABI Research's ranking puts the top five vendors together at only 25–30% of that market — a fragmented field in which a stated win rate above 70% is a claim about deals seen, not share held.

What the accounts say

The same quarter that produced record backlog produced falling profit. Generally accepted accounting principles operating income dropped 10.2% to $66.2m; gross margin compressed to 53.2% from 56.8%. A restructuring charge of $8m, worth $0.11 a share, explains part of it, and management said the savings are being reinvested with no margin benefit this year. On an adjusted basis operating margin was 34.9%, and operating cash flow rose 22% to $91m. The balance sheet carries $186m of cash and no debt, with $275m of a $500m buyback authorization already spent this year. A currency tailwind of roughly one point in the first half is expected to reverse to a one-point headwind in the second.

The re-rating was not Manhattan's alone

Eight specialized enterprise software names moved together after 29 July, and the ones with no transaction meter moved hardest: Intapp rose 39.7%, AppFolio 25.4%, Blackbaud 16.6% on revenue growth of 3.0%. Bentley Systems, which sells infrastructure-engineering software and grew 12.8%, rose 2.1%. Software broadly had fallen about 27% from October 2025 to mid-July on the belief that AI would erode software economics, then bounced roughly 15% from 22 July as earnings beat and Nvidia's Jensen Huang publicly dismissed the disruption thesis. Blackbaud, at 9.3 times forward earnings, rallied while promising very little: "organic total revenue growth of 4% to 6% annually," chief executive Michael Gianoni told investors on 29 July, alongside a target of double-digit annual earnings-per-share growth through 2030. Descartes Systems, whose customs-filing network bills per transaction, rose 2.8% and reports on 10 September.

The verdict

Manhattan earns the bookings half of its move: the backlog is contracted, the cloud line is compounding at 26%, and guidance went up rather than down through a soft retail spending cycle. What the business does not yet earn is the multiple. Roughly 41 times forward earnings is a growth-software price paid for around 10% consensus earnings growth, against 26.9 times for Bentley and 9.3 times for Blackbaud, and it was reached in a month in which the whole group re-rated for reasons that had nothing to do with warehouses. Backlog is a promise about future revenue; it is not evidence that reported margins stop compressing while services and reinvestment carry the mix.

Manhattan has not yet set a date for its third-quarter report. When it comes, the backlog figure will arrive in the same release as the earnings it is meant to justify — and for the first time in a year, it is the one holding up the price.

Snowflake has done something most of its neighbors have not: put an actual dollar figure on its artificial-intelligence business. The company that sells the Data Cloud — a consumption-priced platform where enterprises pool their data and pay for the compute they burn querying it — reached a $100m AI revenue run-rate in its fiscal third quarter of 2026, a quarter earlier than it had guided.

Set that against an annualized product-revenue base of roughly $5.5bn and it is about 2% of the company. Which means the thing that has actually accelerated — product revenue growth from 30% in the January quarter to 34% in the April quarter — is core warehouse consumption stirred up by AI workloads, rather than AI products being billed. That distinction is the whole question on Wednesday September 2, when Snowflake reports its July quarter after the close.

The consumption is real

"Snowflake delivered a milestone quarter, with product revenue of $1.33 billion, up 34% year-over-year, marking the strongest sequential dollar acceleration in our history," chief executive Sridhar Ramaswamy told investors on the May 27 first-quarter call. "AI continues to be a powerful tailwind for Snowflake."

The supporting metrics agree. Net revenue retention — what existing customers spend this year against last — has ground up from 124% to 126% across five quarters. Customers with more than $1m of trailing product revenue went from 606 to 779 over the same stretch, growing 29%, faster than revenue. Remaining performance obligations — contracted work not yet recognized — stood at $9.21bn, up 38%.

Two cautions inside that. RPO fell sequentially from $9.77bn at the January year-end, and its growth rate decelerated from 42% while revenue growth rose. And Snowflake publishes neither a current-RPO figure nor an average contract duration, so a backlog swelled by longer multi-year paper cannot be distinguished from one swelled by demand. Elastic, reporting a day earlier, splits its committed cloud from its self-serve cloud; Snowflake does not offer the equivalent.

Beyond the $100m, AI is quantified in accounts: Cortex Code in more than 7,100 customer accounts in the April quarter, up from 4,400.

Where the margin comes from

GAAP gross margin peaked at 67.8% in the October quarter and has slipped for two quarters to 66.6% — consistent with graphics-processor-served inference costing more to serve than a table scan. The 75% non-GAAP product gross-margin guide is held partly by bandwidth savings under a new Amazon Web Services contract, a procurement term rather than the AI products' own economics. That contract runs the other way on cash: Snowflake committed $6bn of compute and AI spend to AWS over five years, roughly $1.2bn a year against a company whose entire annual cost of revenue is about $1.5bn.

The improvement in the operating loss, narrowed 27% to $326m, is expense discipline: operating costs grew 9.8% while revenue grew 33.5%. Buybacks are not the mechanism either — $873.5m repurchased in fiscal 2026 against $1.48bn of stock compensation, with diluted shares still rising to 345.4m.

What the shares have already paid for

Snowflake closed at $330.97 on August 28, up 22% in thirty days and 86% from the two sessions before May's print, and its 50-day average has sat above its 200-day since late June. The market now pays 33.74x trailing gross profit, against 15.66x in early May and 27.73x in late July. Datadog, the monitoring platform, trades at 27.42x while growing revenue 35.6%; MongoDB, the document-database company, at 18.50x growing 25%. Datadog also supplied the cautionary case, falling 20.4% in the August 6 session on a large AI customer's usage cut despite that growth rate. No Snowflake-specific news explains the thirty-day move; the likelier reading is the broad software rotation plus positioning into September 2.

So: the business earns the direction of the move and not its size. Consumption, retention and large-account counts all improved, and the AI disclosure is better than the adoption-count fog its peers offer. But the company guided July product revenue to $1.415–1.420bn, four points of growth below what it just reported; gross margin is drifting the wrong way; and Databricks, private and self-reporting, claims $7bn of annualized revenue growing about 80%, with its directly competing warehouse product at $1.5bn. Snowflake is now the most expensive dollar of gross profit among its listed peers while growing slower than one of them.

Snowflake has promised Amazon $1.2bn a year of compute it must find customers to consume. September 2 is the first read on whether the demand arrives on that schedule.

The quarter split in two

Elastic's fiscal first quarter, reported after the close on 27 August, broke the company's cloud revenue into two lines that behaved nothing alike. Elastic Cloud — the hosted version of the Elasticsearch search-and-analytics engine that enterprises use for logging, security monitoring and, increasingly, as the retrieval layer underneath AI applications — grew 20% to $235.2m. Underneath that blended figure, annual committed-contract cloud revenue grew 27% while monthly self-serve cloud was flat, which management attributed to expected softness among smaller businesses.

The distinction carries the argument. Elastic had re-rated steeply into this print — from 3.85x trailing gross profit in early May to 6.75x on 21 August — on a guide that implied only about 13% growth, and the July quarter is the first new disclosure since. It says the money arriving is contracted rather than opportunistic. Remaining performance obligations expected to convert within a year reached $1.15bn, up 21%, and total obligations $1.9bn, up 27% — both well ahead of the 15.1% revenue growth Elastic reported, at $478.1m. That beat the $470.1m analysts expected and the company's own May guide.

The customers are there; the dollars are not disclosed

Elastic ended July with more than 1,800 customers at $100,000 or more of annual contract value, after adding over 80 net new ones, its largest quarterly addition on record. Those customers now supply 90% of sales-led subscription revenue, up from 87% a year ago.

On artificial intelligence, the only quantity offered is a headcount of customers. "37% of our 100,000-plus ACV customers are now using Elastic for AI, up from approximately 21% a year ago," chief executive Ashutosh Kulkarni told analysts on 27 August. "That is more than 670 high-value customers now using Elastic for AI use cases with 70 net additions quarter-over-quarter in Q1." No revenue figure, no annual recurring revenue number and no attach rate was attached to vector search or agentic retrieval; the monetization stays buried inside consumption.

Something is being consumed, though, and it shows up as cost. Gross margin fell to 74.5% from 76.7%, so gross profit grew 11.8% while revenue grew 15.1% — the arithmetic of serverless capacity and inference cost of service. Net expansion slipped to 111% from 112%. The operating loss narrowed to $3.65m from $9.44m, non-GAAP operating margin reached 16.2% despite $13m of restructuring charges, and $40m of buybacks took the diluted share count down about 1.2% year on year.

The raised full-year guide of $1.998–2.010bn implies second-half growth near 15.4% — marginally faster than the quarter just delivered, so no deceleration is embedded. Chief financial officer Navam Welihinda told investors on the call that "the sustained progress we see in CRPO, sales productivity pipeline and operating leverage reinforce our confidence in our revenue growth and margin expansion plans."

Reallocated budget, not new budget

Whether this is fresh spending or somebody else's is testable. Cisco, which owns the Splunk log-analytics and security-information franchise Elastic displaces, grew its observability product line 6% in its own July quarter while total revenue rose 17.6%; it trades near 21.9x forward earnings. MongoDB, the closest rival for retrieval workloads, grew 25% in its April quarter and trades at 18.50x trailing gross profit against Elastic's 7.60x — a gap that no longer looks like a bargain bin.

Elastic closed 17.0% higher on 28 August, the largest gain of the session in enterprise software, against MongoDB's 3.1% and Snowflake's flat close: an idiosyncratic response to disclosure. Shares are up 72% in three months. Brokers moved after the fact — Stifel to $107, Oppenheimer to $119, RBC to $120 — leaving the close already above the lowest of them.

What the business earns

The committed backlog and the record high-value customer adds earn most of this move: demand for retrieval is being signed, not merely sampled. What the business has not earned is the doubling of its gross-profit multiple since May against a growth rate that consensus still models at roughly 15% next year too. The trailing price-to-earnings figure of 27.5x is no help here — it is an artifact of a $435.9m tax valuation-allowance release booked in April.

So the question has narrowed rather than closed. Elastic has proved the contracts exist; it has not yet proved that AI work is worth more per customer than the search and logging it replaces, and the falling gross margin is the first evidence that it may cost more to serve. The next four quarters, on management's own timetable for net expansion recovering, will settle which.

Amazon's cloud division supplied roughly three-quarters of the additional operating profit the company earned in the June quarter. AWS operating income rose to $16.6bn from $10.2bn a year earlier — about $6.4bn of the $8.29bn increase in consolidated operating income, against $1.6bn from North America and $0.2bn from International.

That decomposition matters because the case for Amazon's shares this year has rested on an untested premise: that hyperscale AI spending would eventually appear as segment profit rather than as depreciation. In this quarter it did. The unresolved half is that the spending is still accelerating faster than the profit — Amazon raised 2026 cash capital expenditure guidance to about $220bn from about $200bn, blaming memory-chip cost inflation, and trailing free cash flow has turned negative by roughly $7.6bn even as operating cash flow grew 33%.

The cloud line, by itself

AWS revenue of $42.2bn grew 36.7%, the fifth straight quarter of acceleration and the fastest in eighteen quarters, taking the annualized run rate to $169bn. The sequential addition of more than $4.6bn was roughly 80% larger than any previous quarterly increase in AWS history. Segment margin reached 39%; management put the underlying expansion at 520 basis points excluding a one-time derivative gain, which is the figure that describes the business.

Amazon also disclosed what the hypothesis assumed it hid: an AWS artificial-intelligence revenue run rate above $25bn growing at triple-digit rates, and a chips business spanning Trainium and Graviton at a similar scale. Chief executive Andy Jassy told investors on the July 30 call that customers want the silicon on its own: "We do have an increasing number of customers who are interested in us providing the Trainium chips to them, separate from our cloud, and we're actively having those conversations and exploring, and I expect there's a real chance we'll do that in the future," he said. Management said AWS will not have enough capacity to meet 2026 demand, expects the same in 2027, and has already committed significant 2028 capacity. Anthropic and Amazon have agreed to expand to as much as 5 gigawatts of new compute across Trainium generations.

Where the franchise is losing ground

The profit is real; the relative position is not improving. Microsoft's Azure grew 43% on the identical June quarter, passing $100bn of annual revenue, and Google Cloud grew 82% to $24.8bn with segment margin widening from 20.7% to 35.6% — approaching AWS's level from a far smaller base. AWS's backlog of $496bn is now smaller than Alphabet's $514bn cloud backlog, despite AWS booking roughly 1.7 times Google Cloud's quarterly revenue. Alphabet disclosed that just over half of its backlog converts within 24 months; no comparable duration figure for AWS is discoverable, so whether Amazon's number reflects longer contracts cannot be tested.

One suspicion about the earnings quality inverts on inspection. Amazon did not stretch its depreciation schedule to flatter profit: it shortened the estimated useful life of a subset of servers and networking equipment from six years to five, effective January 1, 2025, citing the pace of AI development, at an expected cost of about $0.7bn to 2025 operating income plus roughly $0.6bn from early retirements. The margin expansion came through that, not around it. Net income of $62.6bn is a different matter — it includes $53.4bn of non-operating income, largely a mark-to-market gain on the Anthropic stake, and describes nothing about operations.

What the shares have paid for it

The stock is up 26.9% over 180 days, but the last month's gain was earned in two sessions: a 15.32% move on July 31, the largest in the name in at least a year, following the July 30 release. Shares peaked at $284.02 on August 3 and have given back 6.2% since, a retreat market coverage attributes to capital-spending anxiety rather than any AWS deterioration, alongside a New Jersey attorney general suit filed August 4 and a Jeff Bezos filing to sell more than $4bn of stock.

Against that, the price per dollar of Amazon's annual gross profit — what a share costs relative to trailing gross profit — has fallen to 7.28x from 7.75x in early May, even as market value rose to about $2.87tn. Trailing enterprise value to earnings before interest, taxes, depreciation and amortization of 11.9x sits below Alphabet's 13.1x and Microsoft's 19.0x. The headline 21.2x trailing earnings is an artifact of the Anthropic mark; consensus FY2027 earnings of $10.61 a share, which assumes it does not repeat, implies about 25x.

So the re-rating is paid for. The business earned the move: accelerating revenue, expanding margin, and the majority of incremental profit coming from the segment the story is about. What nothing in the quarter settles is share — AWS is compounding profit while growing slower than both rivals and carrying a smaller order book than one of them.

The next test is arithmetic, not narrative. Amazon guided September-quarter operating income to $22.5bn–$26.5bn, below the June quarter, with a Prime Day timing shift blamed for most of the apparent slowdown. Capacity Jassy says he cannot build fast enough eventually becomes depreciation whether or not the demand shows up.