DK Street Journal

Veeva Systems sells cloud software to drug and device makers only — clinical-trial, regulatory-submission and quality systems on one side, sales-rep customer-relationship management and prescription data on the other. Revenue in the quarter ended 31 July grew 17.6%, the fourth consecutive quarter of acceleration. The composition is the surprise. Of the $108m added to subscription revenue year on year, $68m came from research-and-development and quality products and $40m from the commercial side — 63% of the increase from the clinical and regulatory franchise.

That matters because the story attached to Veeva all year has been commercial: the forced move of pharma customers off the legacy product Veeva hosted on Salesforce's platform onto its own Vault CRM, with the old system scheduled to end on 31 December 2029. The migration is going Veeva's way. It is not yet what is driving the money.

The acceleration arrives with a services bill

Subscription revenues were $766.8m, up 16%. Professional services — implementation and managed work — were $161.2m, up 24%, and rose to 17.4% of total revenue from 16.5% a year earlier. That mix cost more than subscription scale added: gross margin was 75.0%, down about a quarter of a percentage point year on year. Migrating a top-20 biopharma is consulting labor, and the labor is showing up in the growth rate.

Operating income rose 40%, but against a soft base — the year-ago quarter, around the settlement of Veeva's long antitrust fight with IQVIA, carried a 24.8% operating margin. Sequentially, margin eased from 30.9% in the April quarter to 29.6%. Management raised full-year guidance to $3.682-3.687bn, above the roughly $3.65bn consensus, yet the arithmetic of that raise implies second-half revenue growth near 13.7%, almost four points below the quarter just reported.

The CRM wins are conversions

Veeva won Biogen and Regeneron in the quarter, taking it to 12 of the top 20 biopharmas committed to Vault CRM, with two more decisions expected by year-end; management called it the best CRM quarter it has had. Regeneron's global commitment was announced on 25 August, two weeks after Eli Lilly's. These are Veeva's own customers choosing Vault over the alternative rather than new accounts. "We see Veeva maintaining over 70% share in the CRM space, which means we'll continue to execute," Paul Shawah, executive vice president of strategy, told investors on the 26 August call — a defense of position, not a claim of gains.

Salesforce disputes even that. Its Life Sciences Cloud claims more than 140 customers, roughly double a year earlier, naming Novartis, AstraZeneca and Pfizer, and chief revenue officer Miguel Milano has said publicly that Salesforce is gaining share from Veeva.

What is not disclosed

Veeva publishes no calculated billings and no headline remaining-performance-obligation figure, so there is no contracted backlog against which to test the subscription acceleration. Salesforce, reporting days earlier, disclosed current remaining performance obligations of $33.5bn, up 14% in constant currency, and annualized Agentforce revenue above $1.5bn. Veeva's agentic platform, Falcon, has five early adopters and no disclosed price. "We are the rate limiter right now," chief executive Peter Gassner said on the call. "We have to get that product ready, start working with the early adopters, but interest in Falcon is very high." The only priced new product on the call was Aspen, a horizontal CRM at $50 per user per month plus usage overages.

What the market paid

Shares rose 14.3% on 27 August, but about nine points of that was already in after-hours dealing the night before, and the remainder landed inside a broad session in which software stocks rallied hard: Atlassian and HubSpot, with no news of their own, rose 10.3% and 8.2%. Salesforce, on its own record quarter and an Anthropic partnership, rose 22.6%. Over three months Veeva is up 77%, and now trades at 13.2x trailing sales against roughly 9.2x in early May, and 31x forward earnings against consensus that has revenue growth easing from 15% this fiscal year to 12% next.

The beat and the guidance raise are real, and the regulated system-of-record franchise is intact. What the disclosure does not carry is a reason for a multiple that has expanded more than 40% since spring: the acceleration is services-led, the margin comparison is against a settlement-depressed quarter, the company's own guide embeds deceleration, and the AI products investors are paying for have no revenue line, no attach rate and no price.

Two of the top 20 biopharmas will pick a CRM before the year is out. Each decision is either one more conversion Veeva already had, or the first evidence that Milano is right.

Bandwidth cut its forecast for 2026 political campaign messaging to $13m from $15m on its July 29 call, a small revision inside an otherwise raised guide. It is also the clearest available evidence that the November US midterms are not what re-rated the stock: political texting works out to about 1.4% of guided full-year revenue, and contributed roughly $1m in the June quarter.

What inflated the top line was the carriers. Bandwidth, a Raleigh, North Carolina company that owns and operates its own carrier network and sells voice, messaging and emergency-services programming interfaces to enterprises and to the software vendors that resell them, guided full-year revenue to $900m–$910m, about 20% growth. It guided its own cloud communications measure — total revenue less pass-through messaging surcharges — to $622m–$626m, or 11%. Work from the midpoints against 2025's $753.8m and roughly $89m of the $151m of guided growth, near 59%, is surcharge money collected from customers and remitted to mobile carriers without markup. That split is arithmetic from disclosed midpoints, not a reported line item.

Three fee increases, one revenue line

The surcharges are not Bandwidth's doing. US mobile carriers raised application-to-person per-message fees three times this year — T-Mobile effective January 19, AT&T on April 1 and Verizon on May 1. Providers bill them through and hand them over. The effect on reported growth is dramatic and on profit is nil: Bandwidth's revenue went from -1.1% in the fourth quarter of 2025 to +19.8% and +22.2% in the two quarters since.

Gross profit did not follow. It rose 9.6% in the June quarter to $78.5m while revenue rose 22.2% to $219.9m, and GAAP gross margin fell 4.1 points to 35.7%. The company's own 10-Q quantifies it: first-half cost of revenue rose $61m, of which $41m was higher pass-through messaging surcharges. Bandwidth remains loss-making at the GAAP operating line, a fifth straight quarter.

The half that is real

"Strong customer demand across Voice and Messaging, combined with the structural advantages of our owned-and-operated global network, drove another quarter of margin expansion and profit growth," chief financial officer Daryl Raiford said in the July 29 release. On the non-GAAP measures he is right. Adjusted EBITDA rose 27% to a record 18.3% margin, non-GAAP gross margin improved about a point to 59.4%, and full-year non-GAAP earnings are guided up 48%. Customer name retention runs above 99% and revenue per customer hit a record $256,000. Bandwidth was picked in March to carry voice and messaging for Salesforce's Agentforce Contact Center, and all five of the quarter's million-dollar wins included its Maestro orchestration or AI services — though no dollar figure or customer count for AI voice traffic has ever been disclosed.

The expansion metric is the tell. Dollar-based net retention was 107%. Twilio, the closest listed comparable and a company that quantified the identical carrier-fee drag on its own August 6 call, reported 116% net expansion and grew gross profit 20.4% on the same 22% headline revenue growth.

What the price has already paid for

Bandwidth has risen about 245% over twelve months, to $51.85 on August 27 from $15.04 a year earlier, and is still 34% below its July 9 peak of $78.44 — the two sessions around earnings took 19% and then 29% out of it. Against gross profit, the only workable anchor while GAAP earnings are negative, the shares trade at 5.41x trailing and 4.95x forward, down from 6.52x in May but far above the roughly 1.5x implied a year ago. Twilio's 13.54x looks expensive beside it until the growth rates are laid alongside. On earnings, 29.6x forward is being paid for a year in which consensus has revenue growing 4.3%. Piper Sandler's James Fish lifted his target to $52 on August 19 while staying Neutral, citing risk-reward "especially looking at likely reported growth rates beyond this year."

The verdict splits cleanly. Bandwidth's earnings power has genuinely improved — the incremental margin on cloud communications growth is real and the balance sheet was refinanced at a zero coupon. What nothing in the business explains is a re-rating built on a revenue line that is 59% borrowed from mobile carriers and lapses on its own schedule. The AI voice story may eventually fill the hole, but it has produced anecdotes and no disclosed dollars.

Bandwidth has not yet set a date for third-quarter results; the last two years put it in late October. That report is the first read on whether midterm messaging showed up at all, and the last one before the carrier fee increases start comparing against themselves in January.

Okta, which sells the identity software enterprises use to log their staff into applications — single sign-on, directory, multi-factor authentication, plus the acquired Auth0 developer portfolio — reported its July quarter after the close on 26 August. It was the print the whole August rally in the shares had been pointing at, and it arrived with a genuine bookings beat and not a single quantified figure for machine or agent identity. Current remaining performance obligations, the contracted revenue Okta expects to recognize within a year, grew 14% to $2.585bn against total revenue growth of 10.6%.

That gap matters because of what the price now embeds. Okta closed 27 August at $172.91, roughly 175% above its 10 April low of $62.93, and it fetches 11.96 times trailing gross profit against 5.97 times on 3 May, while trailing gross-profit dollars grew about 6% over that stretch. At 45 times forward earnings, the shares are priced against consensus revenue growth of 10.0% this fiscal year, 9.5% next and 10.3% after that — no re-acceleration modeled anywhere.

Where the bookings actually came from

Management quantified the new-product portfolio, not the agent one: identity governance, privileged access and identity threat protection together made up 30% of second-quarter bookings and lifted average deal value by roughly 40%. Customers above $1m of annual contract value grew 22% year on year to more than 600, while those paying more than $100,000 a year grew only 6% to 5,255. This is a large-enterprise expansion story. Okta also won Department of Defense Impact Level 5 authorization, the highest unclassified defense cloud clearance, and the public sector — under a tenth of revenue — supplied the quarter's biggest deal.

On agents, the disclosure was verbal. "Still immaterial. Still very small. We're very early innings… for FY '27, we don't think it's going to be material," Tighe said on the call, adding that fiscal 2028 and beyond hold "a real possibility for this to be material." Chief executive Todd McKinnon was warmer — "In the fullness of time, it could be the biggest category of cyber" — and described one customer whose Claude agent instances multiplied roughly thirtyfold within weeks of an evaluation starting. The plumbing is being laid: Agent single sign-on now ships inside the core product with a consumption cap built in but not enforced, the Cross-App Access protocol went generally available with Anthropic, and the Permiso acquisition, closed the same day, adds about 400 post-authentication risk detections to Okta's roughly 90. None of it carries a disclosed revenue line.

Margins improved; the growth rate didn't

Subscription revenue rose 12% to $793m and net retention held at 107%, a point better than a year ago but unchanged sequentially. Non-GAAP subscription gross margin edged up two-tenths of a point to 83.9%, so governance and Auth0 mix is not diluting profitability. Free cash flow was $227m, a 28% margin. But the non-GAAP operating margin of 28.2% improved just half a point — the seven-point jump in the reported operating margin is largely stock compensation and amortization rolling off. And third-quarter bookings guidance of $2.59–2.60bn implies 11–12% growth, below the 14% just printed.

The shares rose about 29% on 27 August, inside the broad security session the day already carried, and the price targets raised afterward by Morgan Stanley, BMO, Jefferies and others top out near $180 — some 4% above the close.

SailPoint moved on Okta's numbers

SailPoint, the identity-governance specialist whose franchise overlaps Okta's governance push directly, gained about 12% the same session having disclosed nothing. Its last operating figures are from 9 June: annual recurring revenue of $1.163bn, up 26% as of 30 April, on a quarter that still ran a $79.8m operating loss. It reports fiscal second-quarter results before the open on 9 September. Because the two run very different gross margins — 79.6% at Okta, 64.7% at SailPoint — gross profit is the comparable denominator, and on it SailPoint is the more expensive at 15.58 times, growing roughly twice as fast.

The verdict

The quarter Okta delivered is real and it is not the quarter the price is discounting. Faster bookings, wider margins, raised guidance and a debt-free balance sheet with $2.3bn of cash justify a business re-rating off April's distressed level. What nothing in the disclosure supports is the leap from roughly five times gross profit to twelve on an unchanged 10% grower whose finance chief just dated agent monetization to next fiscal year at the earliest and whose own bookings guidance steps down. Okta's structural argument — neutrality against Microsoft's bundled Entra ID, with McKinnon saying "I think Microsoft is copying us" — is a five-year argument being paid for in one quarter's session.

One customer's Claude agents multiplied thirtyfold in a few weeks, McKinnon said. Not one of them has yet turned up in a number Okta is willing to print.

Cloudflare, which runs a global network selling content delivery, bot and denial-of-service mitigation, Zero Trust access and a serverless developer platform, told investors it will spend 14-15% of revenue on that network in 2026. Through the June quarter it had spent 7%. Essentially the entire step-up falls into the two quarters ahead, and it lands in the one line of the income statement that has already turned against the company.

Gross profit grew 30.2% in the June quarter while revenue grew 35.9% — cost of revenue is outrunning sales. That gap is the argument at Cloudflare's price. The shares have re-rated steadily since early May and took another leg on 27 August inside a market-wide software repricing led by Salesforce and CrowdStrike, none of it Cloudflare news. What buyers are paying for is the proposition that Cloudflare can carry the coming wave of machine traffic at software margins. The second half is when that proposition meets the invoice.

The quarter that answered May

On 8 May the shares fell 23.6% in a single session, the largest one-day fall in the company's history, after a first-quarter beat was paired with June-quarter guidance of only $664-665m and a cut of more than 1,100 jobs, about 20% of staff, at charges now put at $165m. Revenue came in at $696.1m, roughly $31m above that guide, and the full year was raised to $2.864-2.870bn.

Underneath, the contract metrics improved across the board. Remaining performance obligations reached $2.732bn, up 38% — faster than revenue. Dollar-based net retention hit 120%, six points better than a year earlier. Customers paying more than $100,000 a year reached 4,698, up 27%, with a record 986 net additions over twelve months; large customers are now 73% of revenue. "We delivered a stellar second quarter, highlighted by revenue accelerating to $696.1 million, up 36% year-over-year, and record growth in total paying customers, large customers, and developers on our platform," co-founder and chief executive Matthew Prince told investors on 6 August.

Where the margin went

Reported gross margin was 71.76%, against 74.88% a year earlier — but up half a percentage point from March, and on the company's non-GAAP measure 73.1%, which management called the first sequential improvement in eight quarters, attributing the annual decline to a mix shift between paid and free traffic that it said was stabilising. Free cash flow was $56.4m, 8% of revenue, against a $302m gap between the GAAP operating loss and the non-GAAP operating income in the same quarter.

Management has been explicit that it will not buy its way into inference: it says it is refusing the commodity server business in favor of a network run at roughly ten times better utilisation. Akamai, the content delivery pioneer led by co-founder Tom Leighton, took the other road and has published the bill. Its June-quarter revenue grew 5.4% to $1.100bn while gross profit fell slightly and margin dropped to 55.8% from 59.1%; capital expenditure was $347m, 32% of revenue, with the full year guided near 40%, buybacks suspended and $3.5bn of zero-coupon convertibles raised against more than $2.8bn of multi-year artificial-intelligence compute commitments. Akamai told investors those contracts carry 60-70% cash gross margins and 20-30% operating margins after depreciation, with costs landing before revenue.

Fastly, the smallest of the three at a $3.9bn market value, is the counter-case: revenue up 23.3% to $183.3m, its fastest in four years, a record 63.3% gross margin, and capex guided to 10-12% of revenue. Its growth is narrow — the top ten customers supply 37% of revenue and produced 87% of sequential growth, and net retention was 117%.

What the price already assumes

Cloudflare trades at 60.0x trailing gross profit, against 50.5x when its uptrend began on 5 May and 37.4x in late February. Over six months the shares rose 79% while trailing gross profit grew 12.9%; roughly three-quarters of the move was the multiple. On management's own full-year guide of $1.25-1.26 in non-GAAP earnings per share, the forward price-to-earnings ratio is near 245x. Akamai fetches 6.44x trailing gross profit and 16.1x forward earnings; Fastly 9.24x. Over twelve months all three rose together — Fastly most of all — but since 5 May they separated: Cloudflare up 26.1%, Akamai down 5.7%, Fastly down 24.4%.

The operating business earns the first half of that. Backlog growing faster than revenue, retention climbing and record large-customer adds are not price increases dressed up; they are expansion. What nothing in the disclosure yet earns is the rest. No revenue figure or customer count exists for Workers, AI Gateway, pay-per-crawl — under which customers now return more than a billion payment-required responses to crawlers a day — or the Monetization Gateway announced on 1 July, whose agent wallets can claim a handle but cannot yet be funded. Nearly 2 million developers joined the Workers platform in the June quarter, more than in all of 2025, taking the total above 7.4 million. Adoption is documented. Monetisation is asserted.

So the margin question is live in the direction that matters, and it resolves on schedule. Third-quarter results are due 29 October, the first print in which the deferred network spending has to appear — or the guide has to come down.

Nutanix closed its fiscal year on 31 July with 32,360 customers running its enterprise cloud platform — the hypervisor, software-defined storage and Kubernetes stack that lets a company run virtual machines on hardware it owns — and more than 3,000 of them were added during the year. Chief executive Rajiv Ramaswami told The Register in a briefing on the fourth-quarter results that most chose Nutanix as an alternative to VMware, and that the migrations have another five years to run.

That claim is the entire investment case, and it is unusually hard to audit, because Nutanix stopped disclosing ACV Billings with its first quarter of fiscal 2025. Billings is where a software company can flatter itself by signing longer contracts; without it, the test of whether Broadcom's post-acquisition repricing of VMware licences is converting into new customers paying new money has to be run on three substitutes — recurring revenue, contracted backlog and average contract duration.

The duration test

Remaining performance obligations — signed business not yet recognized as revenue — reached about $3.44bn at 31 July against roughly $2.69bn a year earlier, growth near 28%. Full-year revenue grew 12.4%, to $2.854bn. Backlog expanding at more than twice the rate of revenue is exactly the pattern a duration game produces, so duration is the thing to check: it came in at 3.3 years, a tenth of a year longer than in fiscal 2025, and down from 3.4 years in the third quarter. Three points of the twenty-eight, roughly. The rest is volume.

Annual recurring revenue, which is duration-neutral by construction because it divides each contract's value by its term, ended the year at $2.549bn, up 16% — 3.6 points ahead of revenue growth, and one point slower than the prior year.

The wins arrive at a steady clip

The new-logo cadence through fiscal 2026 was about 640, then 710, then over 700, then 650. Ramaswami has said Nutanix adds roughly 500 to 1,000 customers a quarter with nearly all coming from VMware. Against a stated target of some 165,000 accounts in Broadcom's installed base, a flat 650-a-quarter clip is a long war rather than a rout — and it is why the headline growth rates keep easing.

The box costs more than the software

The constraint is no longer competitive. AI hardware demand has pushed memory prices up by as much as 70% this year, raising server prices and stretching delivery times; Nutanix licenses software, but the customer must buy the machine. Chief financial officer Rukmini Sivaraman told investors the fiscal 2027 outlook assumes those conditions persist all year, and that a moderately higher share of orders will carry future licence start dates as customers wait for servers — bookings that sit in backlog instead of converting, per the fourth-quarter call. The company's answer is to sever migration from a hardware refresh: expanded support for storage arrays from Dell, Pure Storage, NetApp and Lenovo lets a customer adopt the stack on equipment already owned, and management credited external-storage deals with several seven-figure wins in the quarter. NetApp named the Nutanix platform in its own May results deck as a planned integration.

The July quarter itself was the best of the year: revenue of $757.1m, up 15.9% against 10.0% in the third quarter, GAAP operating income of $70.0m, up 124%, and free cash flow of $277.6m against $207.8m. Gross margin slipped about a point, to 86%. Three weeks before that print, on 4 August, Nutanix cut 5% of its workforce, roughly 400 jobs.

What the price already pays for

Fiscal 2027 is guided to $3.18–3.23bn, about 12% growth, with free cash flow of $850–950m against $841m delivered. The shares rose 6.1% on 27 August on 8.0m shares, roughly triple normal volume, and now sit at 7.62x trailing gross profit against 4.68x in early May — a re-rating of some 63% on a business guiding to slower growth.

The trailing price-to-earnings of 12.5x should be ignored: it rests on a one-time $1.2bn release of a valuation allowance on US deferred tax assets. Excluding it, fiscal 2026 earnings were $1.04 a share, near 67x. Forward earnings are 31.7x. Only cash looks undemanding — the guided free cash flow implies close to a 4.8% yield on a market value of $18.9bn.

Among the neighbours, DigitalOcean, which rents compute to developers and small businesses, is growing at 28.6% and accelerating, but its gross margin has fallen every quarter from 61.4% to 55.0% and it trades at 11.43x forward sales — twice Nutanix's growth at nearly double the multiple. Rackspace, which manages other people's clouds rather than licensing a platform, printed flat revenue of $666.3m at a 19.4% gross margin, carries roughly $2.8bn of debt against a stockholders' deficit, and trades at 0.31x sales. Neither is a hypervisor vendor; both mark the ends of the range Nutanix sits between.

So the displacement is real and the money is new — backlog grew on customer count, not on contract length, and the July quarter's operating leverage was genuine. What the business does not yet explain is the second half of the move. A 63% re-rating since May prices a company compounding faster this year than last, and Nutanix guided the opposite. The swing factor is no longer whether VMware customers want to leave; it is when their servers arrive.

Ramaswami says the migrations have five more years in them. The binding constraint on that arithmetic has quietly changed hands — from Broadcom's price list to the memory market's.

Workday reported its fiscal second quarter after the close on 27 August, and the figure that moved the shares was not the revenue line. Total subscription backlog — everything customers have contracted to pay, including years out — came in at $27.403bn, up 8.0%, against roughly $28.6bn analysts had modelled. The stock fell about 7% in after-hours trading, to near $180 from a $193.57 close, erasing some $3.3bn of market value.

That number matters because of what Workday, Paycom and ADP actually sell. All three meter the same thing: a person on a payroll, billed monthly. Workday sells cloud human-resources, payroll and financial software to large enterprises; Paycom sells a single-database payroll and hiring suite to small and mid-sized American firms; ADP runs payroll and human-resources outsourcing for both, including a business that co-employs staff for smaller companies. They are the one rung of enterprise software where the argument that artificial intelligence is eliminating jobs can be checked against a disclosed operating number rather than a narrative.

The meter reads flat

That number is ADP's pays per control — employees on client payrolls, measured same-store. It grew 1% in the fourth quarter of fiscal 2026 and 1% for the full year, with fiscal 2027 guided to 0–1%. Employer Services revenue over the same quarter rose 7% to $3.70bn with segment margin up to 34.4%. Price, module attach and mix did the work; headcount contributed almost nothing. The wider labour data agree that this is stall rather than collapse — the hiring rate was 3.4% in June with 7.4 million openings, a market that neither hires nor fires. Research built on ADP's own payroll microdata found a 13% relative employment decline for 22-to-25-year-olds in the most exposed occupations since late 2022, older workers steady: fewer new seats, not vanished ones.

A flat meter does not kill a subscription business, but it does move all the growth onto price and product. Workday's July quarter shows both halves. Twelve-month backlog of $9.034bn grew 14.2%, still a shade ahead of the 13.9% subscription revenue growth, so near-term contracted seats are holding. The long-dated portion, which is where future seat expansion gets priced, went from 17.6% growth a year ago to 10.9% in April to 8.0% now — a sequential gain of $109m. Full-year subscription guidance moved to $9.940bn–$9.950bn from $9.925bn–$9.950bn, with margin guidance raised to 31.0%.

Workday's counter is to stop selling by the employee. Artificial-intelligence revenue reached nearly $600m annualized, up more than 200%, and "AI driving more than 25% of our new ACV and more than 5,500 customers now using at least one of our organic agents," co-founder and chief executive Aneel Bhusri said on the call. Those agents are billed through Flex Credits, consumed per metered task off a rate card — work performed, not employees employed.

Paycom's answer: fewer of its own

Paycom does not disclose billable employees at all. It reported 20,321 clients at the end of 2025, up 5%, implying revenue per client of roughly $101,000 against $97,000 a year earlier — about half its growth came from price and attach. June-quarter revenue was $531.2m, up 9.8%, with recurring revenue up 11.0%, and operating margin reached 31.7% from 23.2%. "Our strong second-quarter results came in ahead of expectations, reflecting the strength of our automation strategy and disciplined execution," founder-CEO Chad Richison said on 5 August. Roughly a fifth of the shares have been retired, diluted count falling to 45.9m from 56.3m.

What the prices earned

Workday is down 15.7% over twelve months and up 55.5% over three, its largest session a 12.5% jump on 14 August, when a reported Silver Lake take-private approach was read as a floor under software valuations. Paycom's 47% thirty-day gain begins with a 23.8% session the day after its beat-and-raise. Across the nine back-office software names, thirty-day gains average about 20% against a twelve-month loss near 15%; besides Paycom, the big movers were Veeva and ServiceNow, on company events with nothing to do with headcount.

So the split is clean. Paycom's move is bought and paid for by disclosure; Workday's was bought on a deal report and a rotation, and the disclosure, when it landed, went the other way. Yet the two vendors whose meter is literally the payroll count are the cheap ones: measured against gross profit, Workday's market value is 6.26 times and Paycom's 6.23 times, while ADP commands 10.76 times — a 73% premium for 5–6% guided revenue growth. Workday's own reading has fallen from roughly 9.2 times a year ago, and it trades at 17.97x forward earnings against 39.3x trailing; Paycom sits at 19.6x forward against 25.1x trailing. ADP's forward earnings multiple, 23.2x, is barely below its trailing 26x — and its fiscal 2026 gross profit rose just 1.2% on 6.7% more revenue.

The useful conclusion is not that seats are disappearing. It is that they have stopped arriving, and every dollar of growth now has to be argued for on price, modules or metered agents. Workday put that in writing this quarter: the company built on charging per employee is charging for its newest product per task instead.

Corning has signed three multiyear fiber contracts with the largest builders of artificial-intelligence data centers in the space of five months, and the factories meant to fill them are still under construction. The Hickory, North Carolina cable expansion that Meta anchored broke ground in March and is expected to be fully operational only at the end of 2027 — roughly eighteen months after the volume was booked.

That gap is the whole question at Corning, a 67,200-employee maker of optical fiber, cable and connectors that also supplies glass substrates for displays, Gorilla cover glass for phones, ceramic emissions substrates for cars and polysilicon for solar. For two years the company's re-rating has rested on promises about content per rack. The June quarter is the first in which the fiber line is unambiguously the company: optical communications sales of $2.07bn were 44% of core sales, and their $502m of year-on-year growth accounted for roughly three-quarters of the increase in the whole business.

The split inside optical

The growth is not the carrier business. Sales to telecom carriers were roughly flat in the quarter on order timing, though up 17% across the first half. Enterprise networks — the data-center line — grew 65%, with generative-AI optical sales nearly doubling. Segment net income rose 77% to a record $438m, a 21% margin after tax.

Everything else stood still. Glass Innovations, the display and Gorilla business, grew 1% to $1.46bn. Automotive grew 2% to $471m. Solar sales rose 90% to $438m and still posted a $7m net loss after a maintenance shutdown. The arithmetic settles the question of who is paying for the year: consolidated gross profit rose $236m to $1,628m, while the optical segment's net income alone rose $191m. Display pricing and the yen hedge did not do this work; fiber did.

Margins moved with it. Chief financial officer Ed Schlesinger said core gross margin expanded 1.2 points to 39.6% and core operating margin 1.9 points to 20.9%, with return on invested capital up to 14.9%. Chief executive Wendell Weeks told investors on the July 28 call that demand still exceeds Corning's ability to produce its highest-density optical products, and that any additional output could be sold immediately.

Contracted, not merely announced

The customer commitments are dated and specific. On January 27 Corning and Meta announced a multiyear agreement worth up to $6bn, with Meta as anchor customer for the Hickory expansion. On May 5 NVIDIA and Corning announced a long-term partnership under which Corning raises US optical-connectivity capacity tenfold and US fiber capacity by more than half across three new plants in North Carolina and Texas; NVIDIA took rights to invest up to $3.2bn in Corning. On June 8 Amazon added a multibillion-dollar supply agreement and 1,000 North Carolina jobs. Management said on the July 28 call that long-term agreements will become the "lion's share" of the optical business, with expansions underpinned by contracts that share investment risk with customers.

What is not disclosed is a revenue dollar. Generative-AI optical products are reported as growth rates and design status. Corning laid out a $10bn photonics revenue map by 2030 — co-packaged optics that replace copper inside the rack — and gave no 2027 guidance for it, naming the timing of scale-up adoption as the largest single adjustment in its risk-weighted plan. The engineering case is concrete enough: above 130,000 accelerators a cluster needs three optical layers instead of two, lifting fiber content per accelerator to 1.3–1.5 times current levels by 2028, and a fully optical scale-up would take about 160 fibers per accelerator against roughly 16 today. Capital spending is guided to about $2bn this year against $1.42bn of free cash flow in the quarter.

What the shares have done

Corning closed at $151.99 on August 27, 40.6% below its June 29 peak of $255.69 and 20.4% below three months ago, having lost roughly a fifth in a single session after the July results. The cause was the other half of the company: September-quarter core sales guided to $4.9–5.0bn, at or below consensus, alongside a warning that handset units would fall by a mid-teens percentage this year as memory-chip prices raise phone costs. Universal Display, which sells emitter materials into the same phone cycle, described the identical mechanism and guided to the low end of its year; its June revenue fell 11.4% to $152.2m and it trades near 8.5x trailing gross profit, priced as ex-growth. The optical names then took a second, macro hit in mid-August as the 30-year Treasury yield reached 5.33%, the highest since 2007, with Lumentum, Coherent and Fabrinet all down by around a tenth or more and no company news at Corning.

The verdict

The fiber segment earns its growth outright, and the merchant optics makers confirm the demand is physical rather than narrative — Applied Optoelectronics' chief financial officer Stefan Murry told investors on August 6 that "demand to support next generation AI infrastructure remains so robust that our near term revenue is bounded almost entirely by production capacity and key component availability." What the fiber does not do is set Corning's reported numbers. Consolidated revenue growth decelerated from 20.9% to 16.6% across four quarters and operating-income growth from 95% to 21.8%, because 56% of the company is not optical.

That is visible in the price of the profit stream. Corning trades at 21.4x trailing gross profit against roughly 35.7x at the June high, and 18.9x forward — a compression of about a tenth between the two, where Lumentum goes from 59.2x to 28.7x and Applied Optoelectronics from 52.7x to 30.0x. Consensus is not being asked to believe Corning's gross profit compounds like a pure-play optics maker's; it is being asked to hold a fiber business inside a conglomerate whose consumer end is contracting. Coherent, at 21.6x trailing and 14.5x forward, is the closer comparison and the cheaper one on next year.

So the drawdown is not a mispricing of fiber so much as a repricing of everything wrapped around it. The stake ahead is whether the phone-glass drag is one bad year or the permanent tax on owning the optical franchise. Until Hickory runs at the end of 2027, the AI half of Corning is three signatures and a growth rate — and the other half still writes the guidance.

CrowdStrike, which sells cloud-delivered endpoint, identity and cloud-workload protection as subscriptions to its Falcon platform, reported its July quarter after the close on 26 August: revenue of $1.47bn, up 25.8%, and net new annual recurring revenue of $332.8m, a record and up 51%. The next session, twelve listed security vendors rose together. Palo Alto Networks, which has published no financial statement since April, rose 11.4% of them.

That is the stake this week. A sector that has roughly doubled since the start of May had, until Wednesday night, almost no fresh audited evidence underneath it. CrowdStrike has now supplied some. Palo Alto supplies its own after the close on Tuesday, 1 September.

What CrowdStrike actually disclosed

Revenue growth accelerated for a fourth straight quarter, from 22.2% three quarters ago to 25.8%. Gross profit grew faster than revenue, up 27.7%, as gross margin widened 111 basis points to 74.6%. The number that matters most against the long shadow of the July 2024 outage is the subscription line: GAAP subscription gross margin was 78%, up from 77% a year earlier. The discounting embedded in post-outage customer packages is not visible in the economics.

Underneath, ending ARR reached $5.84bn, up 25%. Commitments through Falcon Flex, the pooled-licensing construct that lets customers draw down across modules, exceeded $2.29bn and more than doubled year on year. Next-generation security information and event management — the log-analytics business CrowdStrike is taking from Splunk-era incumbents — reached $695m of ARR, up 60%. The company said its separately priced AI detection and response module nearly tripled its ARR in a single quarter. Management raised the full-year outlook for net new ARR growth by 630 basis points, to 34% at the midpoint, the second consecutive quarterly raise.

"We are in an arms race," co-founder and chief executive George Kurtz told investors on the 26 August call. "AI is driving more cyber attacks. AI is driving more cyber spending."

The mechanism has a name and a date

The budget shift is traceable. Anthropic's release of its Mythos model, capable of exploiting previously unknown software vulnerabilities, converted AI from a claim on security budgets into a source of them; Kurtz called it "the Mythos moment." Black Hat in early August had already pushed both names to records. On 27 August, 116 companies and entities including OpenAI, Anthropic, Microsoft and AMD signed a letter stating that "We have a limited window to strengthen cyber defenses." Okta reported the same evening as CrowdStrike and rose 27.5%. Microsoft, the largest incumbent competitor in endpoint and security operations, rose 0.8%.

Palo Alto's evidence is four months old

In its April quarter, Palo Alto's revenue rose 31.1% — but $388m of it came from the acquired CyberArk and Chronosphere businesses, leaving organic growth near 14%. Purchase accounting cut gross margin 5.4 points to 67.6%, gross profit grew 21.5% against that 31% headline, and GAAP operating income swung to a loss of $183m from a profit of $219m. The organic evidence that does hold up is contractual: next-generation security ARR grew 28% excluding acquisitions, and remaining performance obligations 23% — both running well ahead of organic revenue, which is what longer contracts look like rather than deeper discounts. "Q3 was a standout quarter," chairman and chief executive Nikesh Arora said on 2 June, "with accelerating organic bookings growth as customers turn to us to secure their AI deployments at scale." Guidance for the quarter now due calls for next-generation security ARR of $8.90bn to $8.95bn.

What the price has done with it

CrowdStrike's four-for-one split on 2 July makes its chart misleading; adjusted, the shares are up 95% since 1 May, in line with the group. Its price to trailing gross profit has gone from 32 times on 3 May to 57 times, against gross-profit dollars up 26% over the past year. Palo Alto's went from 20 times to 41 times on gross profit up 17%, and its forward price-to-earnings ratio is 93 times consensus for fiscal 2027 — a year in which consensus has earnings growing 8.9%, down from 15.3%.

So: CrowdStrike earned its business move and has been paid roughly three times over for it. Palo Alto's re-rating is identically shaped and rests on a print that has not happened. The demand mechanism is common to both and looks real; the distinction now is disclosure. One company has shown the money arriving in ARR, in margin and in raised guidance. The other has shown a backlog from April and a promise.

Tuesday settles half of it. If Palo Alto's fourth quarter holds acquisition-adjusted growth near where April left it, the sector's summer will look like the market pricing a budget cycle early. If it does not, Palo Alto will have doubled on somebody else's earnings call.

Ambarella, which sells low-power system-on-chip processors that run video compression and computer-vision inference inside dashcams, security cameras and robots, has handed back the whole takeover premium the market awarded it a month ago. The shares closed at $70.85 on 27 August, below the $74.09 they fetched the session before the Financial Times reported on 31 July that NXP Semiconductors was in talks to acquire it — a report that named no deal price, only a valuation around Ambarella's own market capitalization. Neither company has commented.

That matters because of what comes next. Ambarella confirmed on 12 August that it will report fiscal second-quarter results after the close on Thursday 3 September. It will be the first fundamental test of a stock whose entire 2026 gain came from deal speculation, and the test lands on a question that separates it cleanly from CEVA, the edge-AI intellectual-property licensor investors treat as its sibling: when does an edge-AI design win actually become money?

Two ways to bill for the same inference

Ambarella gets paid when a chip ships. Edge-AI processors were 80% of its record fiscal 2026 revenue of $390.7m, it has shipped more than 46 million of them cumulatively, and the average selling price recorded last quarter was about $15. That is a clean, immediate line — and it has been slowing. Reported revenue growth has decelerated four quarters running, from 49.9% to 31.2% to 20.1% to 16.9%, the last on $100.4m of revenue in the April quarter. The company's own release put GAAP gross margin at 58.4%, against 60.0% a year earlier. Guidance for the September quarter is $105-111m, a midpoint barely above where consensus already sat.

"Demand signals and the long-term secular growth outlook for edge AI remain very strong," chief executive Fermi Wang told investors on the 28 May call. The order book supports him — a ten-year agreement with South Korea's Hanwha Group worth more than $800m, and a semi-custom 2-nanometer part taped out in January for production in fiscal 2028. The reported line does not yet.

CEVA gets paid twice, and only the first check has cleared. Its June quarter showed licensing revenue up 21% to $18.2m, the best in three years, while royalties — the per-device stream — came in at $10.8m, up 1%. Customers shipped 567 million units, 16% more than a year earlier, which means royalty dollars per unit fell roughly 13%: the volume came from cheap connectivity sockets, and the largest of those, Bluetooth, shipped 295 million units, down 16%. The flagship win, NeuPro-M neural-processing IP selected for custom silicon by a large computing platform company, carries management's own timeline of 1.5 to 2 years to production. It cannot pay a royalty before roughly 2028.

CEVA raised full-year growth guidance to 13-15% from 12% and still fell 13.4% on the day, then 8.4% the next. Management said the shift toward platform-level deals delivers "meaningfully higher" license sizes and royalty per unit than component IP — true, and years away. The whole IP complex is being repriced on this lag: Arm's data-center royalties doubled and it cut its royalty growth guidance anyway.

What the prices now assume

Both companies are loss-making under generally accepted accounting principles, and their gross margins differ by nearly thirty points, so the comparable anchor is price against trailing gross profit. Ambarella sits at 13.2x, up from about 12.1x six months ago but well down from 18.1x at its 26 May peak; on forward earnings it is at 89.9x against consensus revenue growth of 13.2%. CEVA is at 7.8x, compressed from roughly 14.3x at its 3 June high while gross-profit dollars kept rising, though still above February's 6.3x.

The verdict is asymmetric. CEVA's business is doing what the edge-AI thesis requires — accelerating revenue, record licensing, operating margin up eight points — and its shares have priced the royalty delay as though it were a royalty failure; that is the more interesting mispricing of the two, tempered by the fact that the delay is real and management dated it. Ambarella's shares have simply discounted a rumor back to zero, leaving a decelerating top line to justify a forward earnings multiple that requires re-acceleration nobody has yet reported.

One line in next week's print will settle more than the revenue number. Inventory days jumped to 145 from 99 last quarter on a deliberate pre-build ahead of expected memory price increases — the same DRAM squeeze CEVA's customers flagged in handsets. On 3 September the market learns whether those chips were built for customers or for the channel.

Pomerantz LLP said on 25 August that it is investigating whether Pegasystems and certain of its officers or directors engaged in securities fraud, citing the July earnings miss and the company's disclosure that clients were stretching out decision cycles. Pegasystems — which sells workflow, case-management and customer-engagement software to banks, insurers, healthcare payers and government agencies — closed higher in the sessions that followed.

That sequence is the month in miniature. Over the thirty days to 27 August, Pegasystems rose 20.9% and Appian, a low-code automation platform sold into financial institutions, federal agencies, life sciences and insurance, rose 48.4%. Only one of the two put any new gross profit on the board inside that window, and the money at issue is whether the group is being repriced on bookings at all.

One good quarter, one bad one, three weeks apart

Appian's June quarter was the real thing. Revenue rose 19.1% to $203.3m and cloud subscription revenue rose 23% to $131.7m, with net expansion on the cloud base holding at 115%. Adjusted earnings before interest, taxes, depreciation and amortization came in at $16.2m against management's own $5-8m guide, and the full-year outlook went up to $845-853m of revenue. Eighty-five percent of new customers in the quarter bought artificial-intelligence-enabled tiers. The cost shows up above the operating line: gross margin fell to 71.2% from 74.3% a year earlier, which management attributed to the mix of AI products and the compute behind them.

Pegasystems' July quarter was the opposite. Growth in the annualized value of its contracted business slowed to 8% in constant currency; the cloud portion grew 22%, down from 27% the prior quarter. Operating margin fell to 4.0% in the June quarter from 20.7% in December, and net income dropped 55.7% to $13.3m even as revenue rose 9.4% — reported revenue lurches because the company books term licenses upfront, which is why contract value, not the income statement, is the growth measure that matters here. Founder and chief executive Alan Trefler called it a "max confusion moment" on the 22 July call, with buyers whipsawing between spending on AI and controlling its token costs. Management had budgeted a third of the year's net-new contract value into the first half, said the half significantly underachieved, and called the recovery "very difficult."

The month was multiple, not profit

Because Appian carries a trailing net loss, its price-to-earnings ratio measures nothing; gross profit is the comparable base. On that measure Pegasystems went from 3.70x on 29 July to 4.49x — and since it reported on 22 July, before the window opened, its entire advance is the multiple, with the denominator untouched. Appian went from 3.63x to 5.25x while its quarter added roughly 3% to the trailing gross profit base. Roughly nine-tenths of that 48% move is repricing.

The rest is sector weather. Reuters reported on 26 August that software shares have been on a roller coaster driven by AI uncertainty and by momentum and leveraged trading that amplify the swings, with January's drawdown following an Anthropic product launch that some managers feared would obsolete enterprise software — the same substitution fear sitting under both these multiples. The final session of the window was Salesforce's, already the sector's story that day, and it lifted ServiceNow 9.2% alongside it. Over the identical thirty days Salesforce rose 35.9% and Microsoft 26.8%: Pegasystems underperformed both. Appian beat them, and the excess is dated to its 7 August earnings gap, after which DA Davidson, Morgan Stanley and Citi all raised targets — to $34, $32 and $40. The shares have since closed above all three.

What each side has earned

Appian has earned part of its year: trailing gross profit grew 15.7%, so roughly 40% of the twelve-month gain is business and the balance is repricing. What it has not earned is the forward curve. At 38.2x forward earnings it is priced against a consensus that has revenue growth halving to 10.8% in 2027, and against its own guidance implying cloud growth decelerating to 20% from the 23% just printed. Pegasystems has earned none of the last thirty days — but it is not expensive on what it produces: 14.6x forward earnings against 18.9x trailing, on a trailing free-cash-flow yield of 8.5%, with $288m of first-half free cash flow and $360m spent on buybacks in the same six months.

The two are no longer a hedge against each other. A year ago the pair could be held as one bet on process automation; today the cheap name is rising on nothing it reported and the executing name is re-rating past its own forecast. Both need the same thing to hold: a print.

Pegasystems put two-thirds of its full-year net-new contract value into a second half whose first quarter management itself called a "lousy time for turnaround." That number arrives in October, and it will be the first company-specific fact anyone has had since the rally started.