DK Street Journal

A dated test, four days out

CrowdStrike, which sells endpoint, identity and cloud-workload protection as subscriptions through its Falcon platform, reports quarterly results after the close on Wednesday, 26 August. It is the only one of the three largest listed security-software companies with numbers due this week. Palo Alto Networks follows on 1 September; Cloudflare last reported on 6 August.

The business heading into that print is speeding up, which is rare at this size. April-quarter revenue was $1.386bn, up 25.6% — a third consecutive quarter of faster growth, from 21.3% four quarters earlier. Gross margin widened 150 basis points to 75.3%, so gross profit grew 28.1%, ahead of sales. Annual recurring revenue (ARR) ended at $5.51bn, up 24%, with a record first-quarter addition of $255.8m in net new ARR, up 32%. Free cash flow reached $468m, and management raised full-year net new ARR growth guidance by 520 basis points at the midpoint.

The price history was misleading

Anyone scanning CrowdStrike's summer chart saw what looked like a wipeout: about $701 in late June, roughly $186 days later. That was a four-for-one stock split, approved on 3 June and effective 2 July. Adjusted, the shares are up about 98% since 20 February and 61% since early May. CrowdStrike participated fully in the security advance; only the unadjusted series suggests otherwise.

Where the re-rating outran the earnings

All three are loss-making under generally accepted accounting principles (GAAP), so price against gross profit is the lens that survives the comparison. Measured that way, since 22 May Palo Alto went from 27.8 to 38.2 times trailing gross profit — a 37% increase on gross profit that did not change at all, because it has reported nothing since April. Cloudflare went from 44.9 to 57.0 times, up 27%, while its gross-profit dollars grew 6.8%. CrowdStrike went from 47.0 to 51.1 times, up 9%, on gross-profit dollars up 6.4%. One of the three re-rated roughly in line with what it earned.

Cloudflare, which routes and filters traffic for websites from its own global network, has the loudest demand story: revenue up 35.9% to $696.1m, the fastest in two years, dollar-based net retention of 120%, and 4,698 customers spending over $100,000 a year, up 27%. But cost of revenue grew 52.7% as network capacity landed in cost of goods, pulling GAAP gross margin to 71.8% from 74.9%. Management says more than half the traffic on its network is now AI agents rather than people. None of that appears as a disclosed product revenue line; the crawler tolls and payment gateways are announced, not itemized.

Palo Alto, the firewall vendor now bundling security subscriptions around its hardware, reported next-generation security ARR of $8.1bn, up 60% — but $1.6bn of that arrived with CyberArk and Chronosphere. Gross margin fell to 67.6% from 72.9%, and the quarter carried a GAAP operating loss of $183m.

The bond market ended the run, not the customers

All seven of the big security names peaked in the same session, 13 August. Then the 30-year Treasury yield printed above 5.33%, its highest in 19 years, on 18–19 August. Cloudflare fell 15.7% into the 20 August trough and CrowdStrike 15.6%, against 8.9% at Fortinet and 7.0% at Zscaler. The losses sorted by multiple, not by security demand. The partial bounce on 21 August followed the Treasury Department saying it would at least double its debt-buyback size.

What has to print

Sell-side previews put the bar for Wednesday at roughly $292m of net new ARR, more than 3% above CrowdStrike's own guidance. The durable part of the case is consolidation: gross retention near 97%, with endpoint, identity, cloud workload and security-analytics modules sold onto one agent, which is how CrowdStrike takes budget from single-product vendors. The durable threat is Microsoft, which bundles competing security into licences enterprises already buy. Consensus already assumes revenue growth of 23.7% this fiscal year — the acceleration is in the price before it is in the release.

The setup

Where it stands — CrowdStrike's growth and margins improved into a print due 26 August, on a multiple that rose less than its peers' since May.

Would confirm — Net new ARR above $292m with gross margin holding at or above 75%.

Would invalidate — Net new ARR below the roughly $255m added a year ago, or a sequential gross-margin decline.

Watch next — Fiscal second-quarter results after the close on 26 August; Palo Alto's fiscal fourth quarter on 1 September.

Valuation — 51.1x trailing gross profit and 43.8x forward, against 47.0x in late May; 38.4x trailing sales.

Every AI data center begins as paperwork — an interconnection study, a substation layout, a water-withdrawal permit, a stamped set of drawings. Jacobs Solutions, a Dallas-based engineering and consulting firm of about 47,000 people, is the only large firm in that business that has told investors how much of the work it now holds.

AI data-center build-out reached 11% of Jacobs' adjusted net revenue — the fee base that remains after subcontractor and equipment costs pass through — in its fiscal third quarter, up roughly a point from the previous three months. Management said the related backlog has doubled and the pipeline tripled, with client visibility stretching from six-to-nine months out to two or three years. Engineering News-Record ranks Jacobs the largest data-center design firm. In the quarter it took a sole-source engineering, procurement and construction-management award from Hut 8 for Beacon Point, a 1GW campus in Texas due for first energization in 2027, following the River Bend campus in Louisiana. Company-wide backlog hit a record $29bn, up 27%, and Jacobs raised full-year adjusted earnings guidance for a third consecutive quarter, to $7.20-7.30 a share.

The shares are worth what they were a year ago.

The fee and the pass-through

Part of the answer is in how the AI money arrives. Jacobs' revenue grew 34.6% to $4.08bn last quarter while gross profit grew 7.3%. Gross margin fell from 24.9% to 19.9%, and net income dropped 26.7%. Construction-management dollars inflate the top line and carry almost no margin with them.

Tetra Tech, a Pasadena consulting firm of about 25,000 people split between US government agencies and commercial water, energy and mining clients, shows the same arithmetic more starkly. Revenue rose 13.5% to $1.31bn in its June quarter while gross profit fell 3.3% to $243.2m, a 323 basis-point margin compression. Across three quarters revenue growth accelerated — 1.1%, then 10.6%, then 13.5% — and gross profit fell in every one. Its data-center practice runs at roughly $60m annualized; its Ukraine reconstruction work for the US Department of State billed $66m in the quarter alone. The federal-cuts story told about Tetra Tech is also backwards: US federal revenue grew 12% while US commercial grew 1%, offshore-wind cancellations offsetting data-center and mining gains. The company raised fiscal 2026 guidance on 30 July to revenue of $4.315-4.365bn and earnings of $1.56-1.59 a share.

The firms that refuse the construction risk

Stantec, an Edmonton-based design and environmental consultancy of about 34,000 people, takes the opposite route: fees for drawings and studies, no construction exposure. Net revenue rose 11.5% to C$1.8bn in the second quarter, of which 3.7% was organic, and adjusted earnings per share rose 18.4% to C$1.61. Adjusted margin reached a record 18.7% for a second quarter. Backlog set a record at C$9.2bn, up 17.5% and 7.0% organically, about thirteen months of work; its US book grew to C$5.5bn from C$4.6bn. Its buildings team was selected to design Meta's $13bn campus in Sturgeon County, Alberta — the largest named AI-campus design mandate any of these firms has disclosed. The soft spot is domestic: US organic growth was flat, as a Navy environmental job and a Western utility project ramped slower than planned.

AECOM, a 51,000-person infrastructure consultancy, demonstrated in one session what construction risk costs a design firm. It booked a $337m pretax charge on a 2018-vintage public-private design-build contract, attributed in analyst notes to the JFK Airport modernization program. June-quarter revenue fell 14.2% to $3.59bn, the company posted a net loss of $86.7m, and adjusted earnings of -$0.50 a share missed a $1.51 consensus. Free-cash-flow guidance was cut to $300m from $400m, with about $500m of cash draining out through the first half of fiscal 2027. Orders said the reverse: record backlog of $27.8bn, up 13%, at a 1.6x quarterly book-to-burn. Management has said it will not bid public-private design-build again, and that construction management is 6-7% of net service revenue.

Demand is not the constraint. Pennsylvania's governor removed data centers from the state's permit fast-track program on 18 August, requiring any project drawing 25MW or more to sign a consent agreement with the environmental regulator — mechanically adding review and interconnection work of exactly the kind these four firms sell.

What the prices say

AECOM sits near 10.5x consensus fiscal 2027 earnings of $6.17, and 10.8x management's own ex-charge figure of $6.00 — against 22-24x forward as recently as May. Stantec is 23.0x trailing, down from 28.95x in early May, roughly 21.5x forward on a like-for-like currency basis, with a 5.7% free-cash-flow yield. Tetra Tech's 23.5x forward sits above its 22.2x trailing, meaning analysts model earnings lower than the past year's. Jacobs is 20.7x forward — the cheapest forward multiple of the three that disclose growing data-center work, and the only one with a quantified AI number behind it.

Over twelve months AECOM fell 47% and Stantec 33%, while Jacobs and Tetra Tech went nowhere. The contractors that build what these firms draw compounded: Comfort Systems rose 140%, Quanta Services 72%, EMCOR 27%. The market has priced the design rung as one class and has not yet separated the firm with the biggest disclosed AI book from the ones describing it in adjectives.

The setup

Where it stands — Jacobs is the only large design firm quantifying AI exposure, at 11% of fee revenue, and carries a lower forward multiple than Tetra Tech's 1.4%. Would confirm — AI data centers exceeding 12% of adjusted net revenue in the October-quarter report with adjusted margin holding. Would invalidate — data-center backlog growth stalling, or fee margin falling further as pass-through revenue keeps climbing. Watch next — Jacobs' fiscal fourth-quarter results in November; Stantec's CEO handover to Susan Reisbord on 1 October. Valuation — Jacobs 20.7x forward against a distorted 52.3x trailing; Tetra Tech 23.5x forward, Stantec 23.0x trailing, AECOM ~10.5x fiscal 2027.

Taiwan Semiconductor Manufacturing builds the chips that Nvidia, Apple, Broadcom and Qualcomm design and cannot get made competently anywhere else. In the June quarter that position produced a gross margin of 67.7%, up from 58.6% a year earlier, on revenue of NT$1,270bn — growth of 36% year on year, per the company's 6-K filing. Gross profit grew 57% and operating income 65%, both far ahead of sales. July was a record month at NT$467.6bn, up 44.7% from a year earlier, with essentially all available wafer capacity already sold.

What is actually being sold

High-performance computing — accelerators, networking silicon, server processors — rose 20% sequentially to 66% of revenue on the July call. Smartphones fell to 22%. Processes at 7 nanometers and below accounted for 77% of wafer revenue, with 3nm and 5nm together about 63%. Management called the gap between demand and supply "very large" and said rivals face "no shortcuts": five to seven years to develop, build and ramp a competitive node.

The company also spent against that. Capital expenditure for 2026 was lifted 15% to $60-64bn, with a further $100bn committed to roughly four Arizona fabs at 2nm and below. That carries a visible cost: the 2nm ramp dilutes gross margin by three to four points, and overseas fabs by two to three points early on. September-quarter margin was guided to a 66% midpoint. TSMC's long-term target is about 68%, and it has said publicly it will not push through the 4-5x price increases that memory suppliers have taken.

The valuation moved the other way

TSMC now trades at 27.9x trailing earnings and 21.9x trailing gross profit, against roughly 25.7x gross profit three months ago. Its forward multiples are not directly quotable — consensus estimates are published in Taiwan dollars against a US-dollar listing — but analysts expect earnings to rise about 65% this year. The multiple compressed while the earnings base compounded.

It was also the weakest of the large foundries over the past year: up 84%, against 172% at United Microelectronics and 345% at Tower Semiconductor. Every one of them peaked in the second half of June and has fallen since — TSMC 12% from its 30 June high, and it dropped out of its strongest uptrend at the end of July, though its 50-day average remains above its 200-day.

The mature-node half

The two trailing-edge names tell a different story about the same demand. GlobalFoundries, a specialty foundry making radio-frequency, power-management and silicon-photonics chips from Malta, New York, expanded gross margin by 411 basis points to 28.3% and grew communications and data-center revenue 62% year on year — and has fallen 41% in three months, to 13.85x trailing gross profit. United Microelectronics, the Hsinchu contract foundry that dominates 22 and 28nm work, lifted utilization to 85% and guided above 90%, and for the first time quantified AI-linked sales at roughly $300m in 2026; it costs 19.04x trailing gross profit, more than GlobalFoundries, on far less disclosed AI content.

Mature-node pricing has stopped deflating: TrendForce reports foundry prices rose 5-15% between the first and second quarters, with a third round being prepared for 2027. What broke the shares was not the fabs. Chip stocks lost more than $1trn in late July on doubts about the return on AI capital spending, and a second leg came when the 30-year Treasury yield hit 5.33% on 18 August, a 19-year high. Long-duration compounders repriced with the discount rate.

The setup

Where it stands — TSMC's margin and revenue set records in the June quarter while its multiple on trailing gross profit compressed by roughly four turns since May.

Would confirm — September-quarter gross margin printing at or above the 66% guided midpoint despite the 2nm ramp dilution.

Would invalidate — Monthly revenue growth slowing below 30% year on year, or a cut to the $60-64bn capital-spending plan.

Watch next — Monthly revenue for August and September, then third-quarter results and 2027 capital-spending commentary in October.

Valuation — 27.9x trailing earnings, 21.9x trailing gross profit, down from about 25.7x in May; forward estimates are quoted in Taiwan dollars.

Dell Technologies recognized roughly nine times as many dollars of artificial-intelligence server revenue in its most recent quarter as a year earlier, and the gross margin on the whole company fell by more than three points. The segment that shipped those machines still earned the highest operating margin in its history.

That pairing is the story the assembler rung of the AI buildout has been telling all year, and it runs against the intuition. Dell, the Round Rock, Texas group that builds servers, storage and PCs and sells them through the broadest enterprise channel in the industry, reported January-to-May quarter revenue of $43.8bn, up 87.5%. Gross margin came in at 17.75%, down 336 basis points, as memory prices doubled. But its Infrastructure Solutions Group — servers, storage and networking for enterprises and hyperscalers — grew revenue 181% to $29.0bn and lifted segment operating margin 80 basis points to a record 10.5%. Group operating income tripled. The company also generated record operating cash flow of $4.1bn, which is the number that matters most, because the bear case holds that passing through graphics-processor cost swallows working capital before it reaches the bank. It ended the quarter with a $51.3bn AI server backlog.

The network didn't do the work

Hewlett Packard Enterprise, the Houston maker of ProLiant and Cray servers that spent $14bn on Juniper Networks to buy itself a richer revenue mix, shows the same thing more starkly. Its Cloud and AI segment — the servers — grew 22.9% to $7.7bn at a 12.4% operating margin, against 6.6% a year earlier. Networking revenue rose 148% on the acquisition, but its margin fell to 21.6% from 25.0%. The mix got richer and the acquired business got less profitable; the gross-margin expansion came substantially from the box.

The mechanism is scarcity, and it is specific. TrendForce expects server DRAM contract prices to rise 13–18% quarter over quarter in the third quarter, with registered-memory bit supply growing 15–20% a year against faster server shipment growth and cloud providers locking supply away under long-term agreements. When the scarce input is memory and graphics processors, the assembler holding allocation sets the delivery date — and can price it.

Allocation is not evenly held. IDC puts Dell first in the first-quarter server market at about 10% share on $12.6bn of revenue, with HPE fifth at 3.0%. These two are not peers in this socket. At Barron's midyear roundtable, the investor Scott Black argued Dell's hyperscaler servers beat HPE's on total cost of ownership and serviceability, and named that gap as HPE's central competitive risk.

NetApp is guiding the other way

NetApp, the San Jose maker of all-flash storage arrays and the ONTAP data-management software that runs inside Microsoft Azure and Amazon Web Services, was supposed to be the high-margin counter-case. It held gross margin at 70.1% in the April quarter and grew operating income 53%, on record all-flash revenue of $1.2bn, up 18%. But full-year revenue grew only 5.4% to $6.9bn, and management has guided fiscal 2027 gross margin down to 68.5–69.5% from 71.3%, citing memory and flash costs. Where Dell and HPE absorbed the input shock and expanded segment profitability, NetApp has told investors it will not.

What the shares have already paid for

All three have held an uptrend since spring, their 50-day averages above their 200-day — Dell since 31 March, HPE since 23 April, NetApp since 21 May — with no interruption. The moves are large: Dell closed at $442.08 on 21 August against $164.13 at the end of March. This was not sector beta. Supermicro, the second-largest AI server builder, still trades below where it stood a year ago.

How much was earned differs sharply. HPE's price per dollar of trailing gross profit is 5.53x against 2.94x in March, on gross profit that grew 35% — roughly half the move came from the business, and the shares sit at 13.2x fiscal 2027 consensus earnings. Dell's is 11.49x against 4.85x in March, on gross-profit dollars up 21%; at 23.5x forward earnings it has left behind the 14–17x forward range it occupied in May. NetApp is the furthest stretched, at 7.70x trailing gross profit versus 4.33x in March on gross profit that grew 6.2%, and 21.3x forward earnings against a company-guided 8% revenue growth.

The test arrives inside 48 hours. Dell reports on 1 September; NetApp and HPE both report on 2 September.

The setup

Where it stands — Segment profitability at Dell and HPE rose through a memory-cost shock; NetApp has guided its gross margin down instead.

Would confirm — Dell's infrastructure segment operating margin holds at or above 10.5% in the 1 September print.

Would invalidate — HPE's server segment margin falls back toward the 6.6% of a year ago, or Dell's operating cash flow turns negative on working capital.

Watch next — Three reports in 48 hours: Dell on 1 September, NetApp and HPE on 2 September.

Valuation — Dell 34.5x trailing and 23.5x forward earnings, against roughly 14–17x forward in May; HPE 13.2x fiscal 2027; NetApp 21.3x forward.

Salesforce, which sells the Customer 360 suite of sales, service and marketing software billed largely per licensed user, raised $25bn of debt in March and handed it to an accelerated share repurchase, half of a $50bn authorization approved in February. The arithmetic arrived quickly. Diluted shares fell to 871m in the April quarter from 970m a year earlier. Diluted earnings per share reached $2.42 against $1.59.

It was not free. In late May the company cut its fiscal 2027 operating and free-cash-flow growth guidance to 4-5% from 9-10%, specifically to carry the interest.

The business underneath is accelerating

Strip the financial engineering and Salesforce is still growing faster each quarter: revenue up 8.6% year over year in October, 12.1% in January, then 13.3% in April to $11.13bn. Gross margin held at 76.92%, against 76.96% a year earlier — notable in a year when smaller application vendors handed margin back to the cost of running inference. Operating margin widened to 21.8% from 19.8%.

The AI disclosure is more mixed than the headline suggests. Agentforce and Data 360 annual recurring revenue reached nearly $3.4bn at the May print, with Agentforce alone at $1.2bn, up 205%. But $1.1bn of that pool is Informatica Cloud, bought rather than built, and consensus has group revenue decelerating to about 11% next year as the acquisition anniversaries.

The skeptical case has not gone away. Salesforce is a system of engagement with dominant share and no infrastructure underneath it; Klarna publicly left it for an AI-assembled stack, and rival platforms are now indexing Salesforce and Slack records into their own context graphs, which moves the coordination value to whoever holds the graph rather than the seat.

Against that, the price. Salesforce trades near 5.15x trailing gross profit, down from roughly 7.9x a year ago and about 5.5x in February — the shares rose and the multiple did not, because gross-profit dollars grew 11% while the share count shrank 10%. Forward earnings are 14.8x against 24.1x trailing.

ServiceNow won the demand argument and paid for it in margin

ServiceNow sells the Now Platform, workflow automation for IT service management, security and human resources, anchored on a configuration database that most of the Fortune 500 runs its operations through — which is why rip-and-replace is rare and renewals sit at 98%. Its June-quarter results were the strongest demand evidence the sector produced: revenue up 24% to $3.99bn, current remaining performance obligation of $13.2bn, and annual contract value from AI products past $1bn for the first time, up more than 40% sequentially. Management counted 123 deals above $1m of net-new contract value and said half of new business is no longer seat-based, with AI editions carrying a 30% price uplift.

The profit line contradicts it. Reported gross margin fell to 70.68% from 77.48%, so gross profit grew 13.1% while revenue grew 24%. Operating income fell 55% to $162m. Hyperscaler hosting for AI features and three consolidated acquisitions sit in cost of revenue, and the company guides subscription gross margin to 81% for the year on an adjusted basis, expecting scale to relieve the pressure.

ServiceNow also cut about 1,000 roles, near 3% of staff, targeting flat headcount for 2026 — a vendor whose customers buy seats removing its own. Its CRM line now carries more than $2bn of contract value, but that is 1-2% of a market where Salesforce holds over 19%: an attack from a small base, not displacement.

ServiceNow's multiple went the other way from Salesforce's. It sits near 12.06x trailing gross profit against roughly 10.5x in February — a 25% expansion bought with 13% gross-profit growth — though still far below the 19x it carried a year ago and the 40x-plus forward earnings of its history. Forward earnings are 31.6x.

The money came out of chips, not out of the Fed

The re-rating was not gradual and it was not about rate cuts. On 27 July Nvidia's Jensen Huang told Bloomberg the chip industry must grow roughly tenfold to serve "100 billion agents and billions of robots", extending an argument he had made earlier that markets "got it wrong" on AI's threat to software. Salesforce gained 11.3% and ServiceNow 15.9% across the following week while the broad market fell. On 19 August the pair added 5.1% and 6.5% as the Philadelphia semiconductor index fell 2.2%. Over the month Advanced Micro Devices fell 14% and Broadcom 7%. Meanwhile the 10-year Treasury yield sat at 4.67% and Goldman Sachs was arguing a September rate hike had become unlikely — the debate was tightening, not easing.

Both names were followers, not leaders. Atlassian roughly doubled over the month, Workday rose on a reported $51bn Silver Lake take-private approach, and SAP and Adobe both outran the pair. Salesforce only crossed into an uptrend on 19 August, its 50-day average rising above its 200-day; ServiceNow did so on 7 August. Both remain below where they traded a year ago — Salesforce by 15%, ServiceNow by 27%.

The cleaner read: at Salesforce the business is doing the work and the multiple has not been paid up; at ServiceNow the contract value is real and the multiple has already moved ahead of the gross profit.

The setup

Where it stands — Both have re-rated on a rotation out of semiconductors, with Salesforce's advance earned in gross-profit dollars and ServiceNow's in multiple. Would confirm — Salesforce revenue growth holding at or above 13% with gross margin near 77% in the July quarter. Would invalidate — ServiceNow's reported gross margin falling further below 70.7% while current remaining performance obligation growth slips under 20%. Watch next — Salesforce reports second-quarter fiscal 2027 results on 26 August 2026, after the close. Valuation — Salesforce 5.15x trailing gross profit against roughly 7.9x a year ago; ServiceNow 12.06x against 10.5x in February.

Three companies that supply the machinery makers of the chip industry — not the tools themselves, but the plumbing, chemistry and test hardware consumed inside them — each reported expanding margins over the past month and lifted their outlooks. Their shares were then sold down roughly twice as hard as those of Applied Materials, the largest buyer of what they make.

The trigger was not in any of them. On 18 August the 30-year US Treasury yield touched 5.33%, its highest level since 2007, on inflation and government-spending concerns and a wave of long-dated supply; Bloomberg reported the same session that global sovereign yields had reached multi-year highs. Over the four sessions to 21 August, MKS fell 13.7%, FormFactor 17.7% and Ichor 17.9%, against 8.9% at Applied Materials. None of the three released news in that window. A discount-rate shock falls hardest on earnings sitting in a 2027 order book, and that is where much of this layer's profit is booked.

What each one actually sells

MKS Inc. supplies vacuum and pressure control, gas delivery, radio-frequency power and lasers, and — through its 2022 Atotech acquisition — the electroplating chemistry and laser drilling used to build advanced circuit boards. Its Electronics & Packaging segment grew 44% year over year to $381m in the June quarter, faster than the $554m semiconductor segment. But MKS disclosed that, stripping out currency and palladium pass-through, chemistry sales rose 21%. The gap is lower-margin equipment: the recurring-consumable story is real, but it is currently being led by the machines that will consume the chemistry later. Group revenue rose 28.3% to $1,248m while gross profit rose 31.3% and operating income 85.9%. Net leverage is 3.0x trailing EBITDA, down a full turn in a year after a $100m voluntary term-loan prepayment.

FormFactor makes probe cards, the consumable contact arrays that test chips electrically while they are still on the wafer. It is the cleanest read here on advanced-packaging intensity: DRAM probe-card revenue rose 48.9% to $85m in the June quarter, with high-bandwidth memory roughly two-thirds of it, after two customers adopted its full-wafer contactor for HBM4 testing. Gross margin went from 37.2% to 50.7% in four quarters. The caveat is in the guidance: September-quarter DRAM revenue is flat, with mix shifting toward conventional DDR memory as customers chase better DDR pricing. The incremental growth this quarter comes from data-center logic and from co-packaged optics test, now expected to exceed $20m by quarter-end against $10-20m guided for the full year.

Ichor builds fluid-delivery subsystems — gas panels, chemical modules, welded components — largely to other companies' designs. Its gross margin has run 4.6%, 9.4%, 11.6% and 13.9% over four quarters, and management puts internal proprietary content at about 25% of what it ships, targeting 35% by year-end. It raised 2026 revenue growth guidance to at least 30%. That is a genuine inflection off a very low base — and it is still a broad wafer-equipment recovery rather than a packaging one: lithography customers are working down inventory into the December quarter.

What is being paid for the gross profit

MKS trades at 9.83x trailing gross profit, against 10.77x in early May, and at 21.3x forward earnings versus 42.7x trailing. FormFactor is at 21.6x trailing gross profit against 34.35x in early May — a 37% de-rating while its gross margin rose thirteen points — and 37.5x forward earnings versus 77.1x trailing.

Ichor is the outlier. At 20.4x trailing gross profit it is fractionally more expensive than the 19.70x it carried in May, the only one of the three whose multiple has not compressed. Its forward price/earnings ratio of 38.3x is the highest of the three, earned on a 13.9% gross margin against FormFactor's 50.7% and MKS's 47.7%. The gross profit Ichor booked over the last four quarters, near $103m, is about 7% below the prior four — so its 232% twelve-month share gain is multiple, not earnings. Operating cash flow was minus $15.9m in the June quarter on inventory build. And it completed a $200m at-the-market share sale, issuing 2.5m shares at an average $80.70, into a stock that closed at $60.20 on 21 August.

The demand is not in doubt. Applied Materials has raised its 2026 advanced-packaging outlook to more than 70% growth, and Lam Research expects its own packaging revenue to grow more than 50% in a wafer-equipment market it now sizes at $150bn. SK Hynix is projected to hold 54% of the HBM4 market next year. The open question is which supplier converts that spending into gross profit per dollar of revenue, and on what durable basis — a proprietary consumable that must be requalified to replace, or an assembly contract that can be rebid.

All three have slipped this summer from a strong uptrend to a milder one, their 50-day averages still above their 200-day. Over three months each fell about 12% while Applied Materials rose 12.8% and Micron 27.9%: the suppliers have lagged both their customers and their end market.

The setup

Where it stands — MKS and FormFactor got cheaper on gross profit while margins expanded; Ichor's multiple held as its trailing gross profit shrank.

Would confirm — Ichor's September-quarter gross margin printing inside its guided 14.5-15.5% range, with proprietary content reaching 30%.

Would invalidate — FormFactor's December-quarter DRAM revenue falling as DDR mix displaces high-bandwidth memory, or MKS chemistry growth slowing below 20% ex-pass-through.

Watch next — Third-quarter results from all three in late October and early November; FormFactor's Farmers Branch plant starts ramping in the fourth quarter.

Valuation — Trailing gross profit: MKS 9.83x, FormFactor 21.6x, Ichor 20.4x, against 10.77x, 34.35x and 19.70x in early May.

Atkore makes the steel and plastic conduit, cable management and metal framing that electrical contractors install inside every data hall, under brands including Allied Tube & Conduit, AFC Cable Systems and Unistrut. Its investment case had one moving part: pandemic-era conduit prices, which collapsed after 2022 and took the margin down with them, would eventually stop falling, and the data-center volume underneath would become visible.

In the quarter ended 26 June it stopped falling. The company's own sales bridge shows average selling prices adding $22.4m and volume adding $65.7m, with revenue up 8.1% to $794.8m. Growth had been negative two quarters earlier. Gross margin fell year on year regardless: input costs rose $48.9m, more than double the price recapture. US hot-rolled coil sat near $947 a ton in mid-2026 with a 50% Section 232 duty on steel articles in force since June 2025. The inflection is in the revenue line, not the margin line: operating margin of 8.1% is a quarter of the 31.6% Atkore earned in fiscal 2022.

What a strategic buyer will pay for conduit

Two days after that print, Prysmian agreed to acquire the company for $95.00 a share in cash, an enterprise value of about $3.8bn and a 30% premium to the prior close, with roughly $150m of expected run-rate synergies. Prysmian bought Encore Wire in 2024 and General Cable before that. Its stated view of what it was buying is unusually blunt: Atkore has not captured data-center growth to the same extent as Encore.

The shares gapped from $72.96 to $93.55 on 8.9m shares — against typical daily volume in the low hundreds of thousands — and have since traded inside a 65-cent range for thirteen sessions, closing at $93.60, some 1.5% under the contract price. Nothing about that price is a market opinion on conduit. It does, though, fix a number: against maintained fiscal 2026 adjusted EBITDA guidance of $340-360m, Prysmian paid roughly 10.6-11.2x. EMCOR, the largest US mechanical and electrical contractor, trades at 14.9x trailing EBITDA; IES Holdings at 22.8x; Comfort Systems at 29.0x. A strategic buyer with synergies would not stretch for the material. The installation layer is where the campus money is being priced.

It is also where it is being earned. Comfort Systems grew June-quarter revenue 50.3% with gross margin widening to 25.9% from 23.5%. IES Holdings grew 39.6% at a 27.4% gross margin. EMCOR grew 19.8% and reported record remaining performance obligations of $17.14bn, up 44%, with mechanical data-center revenue more than doubling.

Mueller's copper is running faster than Mueller

Mueller Industries — copper tube, fittings and PEX for plumbing wholesalers, brass rod, and valves and heat exchangers for air-conditioning makers — is the counter-case, and it half-fails. Second-quarter revenue rose 25.5% to $1.43bn with unit volume growth in all three segments, so this is not pure metal inflation. But COMEX copper averaged $6.16 a pound in the quarter, up 30.6%. Gross profit rose only 12.0%, gross margin narrowed 332 basis points to 27.7%, and reported operating income rose 1.9% — 15.7% adjusting for a $36.3m insurance gain a year earlier. One quarter before, the spread was genuinely widening: revenue up 19.3%, gross profit up 31.8%, operating income up 51.4%. The conversion margin now grows at less than half the speed of the metal.

Neither release quantifies any data-center or liquid-cooling revenue; the demand commentary is generic. The shares, adjusted for the two-for-one split effective 1 July, are down 7.1% over three months and 11% from their 10 August high, against copper's roughly 3% retreat from an all-time COMEX record of $6.77 set on 7 August. That is a de-rating with a reason. It leaves Mueller at 16.2x trailing earnings, above roughly 14.6x a year ago, and at 11.0x trailing gross profit against Comfort Systems' 20.2x — cheap against the installer, expensive against its own history.

A final caution on grouping: the electrical balance-of-plant names that rose most over the past year did so on their own accidents. Babcock & Wilcox's 442% twelve-month gain is a $1.51 share round trip, and its 59% three-month fall came with securities class actions and a $200m equity raise at $18.50. There is no shared materials trade here — only one company being bought and one being repriced against copper.

The setup

Where it stands — Atkore is a merger stub at $93.60 under a $95.00 cash bid; Mueller's volumes are growing while its spread over copper narrows.

Would confirm — Mueller's third-quarter gross margin holding at or above 27.7% with revenue growth still above 20%.

Would invalidate — A third straight quarter of gross-profit growth below revenue growth at Mueller, or the Prysmian deal failing antitrust review.

Watch next — Mueller's third-quarter release in late October; Prysmian targets closing the Atkore acquisition by calendar year-end 2026.

Valuation — Mueller 16.2x trailing and 15.2x forward earnings, versus about 14.6x trailing a year ago; Prysmian paid roughly 10.6-11.2x guided EBITDA for Atkore.

Three contractors reported record quarters this summer for the same unglamorous job: getting power, water and cooling into a shell building before a hyperscaler can switch on its chips. The results were emphatic. The reaction, four weeks later, was a synchronized selloff none of them announced anything to deserve.

What the money is paid for

EMCOR Group, a 44,000-person mechanical and electrical contractor that installs power distribution, piping, ventilation and low-voltage cabling in data centers, hospitals and industrial plants, posted June-quarter revenue of $5.15bn, up 19.8% and almost entirely organic. Operating income rose faster, up 31.8%. Remaining performance obligations — signed work not yet performed — hit a record $17.14bn, a 44% increase, led by the network-and-communications category that carries data centers. Management raised full-year earnings guidance about 10% at the midpoint.

Inside that result is a split worth more than the headline. EMCOR's electrical construction margin expanded 210 basis points to 13.9% on 24% revenue growth. Mechanical construction grew faster, at 31%, and lost 110 basis points, to 12.5%. The reason is contract form: roughly a tenth of mechanical work has shifted to guaranteed-maximum-price and construction-manager arrangements, where the contractor procures chillers and switchgear on the owner's behalf and marks them up thinly. Revenue earned on passed-through equipment is not the same business as revenue earned on scarce electricians. Electrical work accounts for 45-70% of data-center construction cost, and cost per megawatt has climbed from $7.7m in 2020 to $10.7m in 2025 — labor scarcity is the binding constraint, and it is the part of the job that prices.

The concentrated one

Comfort Systems USA does the same installation work plus off-site modular assembly of complete data-center plant rooms. Revenue grew 50.3% to $3.27bn, with same-store growth of 44%, and backlog reached a record $14.1bn, up 73%. Gross margin expanded 242 basis points to 25.9%. Technology work reached 58% of first-half revenue, from 40% a year earlier: the mix is concentrating, not diversifying. Management sells directly to two large hyperscalers, books $510m of modular orders in a quarter, and is allocating all new modular capacity to existing customers.

Two qualifiers travel with that margin. The quarter included $7.7m of favorable estimate revisions, double the prior year, after $43m of change-order gains in March that management told investors to exclude from the baseline. And second-half same-store growth was guided to the high-20s or low-30s, against 44% in the June quarter. Free cash flow of $999m was roughly 2.5x earnings, about a third of it advance cash on bookings — customers pre-funding the backlog, a benefit that reverses if orders pause.

The control case

IES Holdings, a smaller electrical and technology installer whose Communications segment builds network infrastructure inside data centers, is the test of whether this demand reaches past the two large caps. It does: revenue rose 39.6% to $1.24bn, accelerating from 16.8% growth in March, operating income rose 59.6%, and backlog reached $4.5bn, up 91% since September. Communications revenue grew 51%. Its 27.4% gross margin is the highest of the three. Its consensus forecast rests on a single analyst, so forward estimates carry little information; a two-for-one split goes ex on 24 August.

What the twelve months were made of

EMCOR's shares rose 27.3% over the year while trailing earnings per share went from $24.10 to $32.01 — a 32.8% gain. The multiple contracted, from about 25.3x to 24.3x. Comfort Systems rose 140%, and while earnings covered most of that on a price-to-earnings basis, gross profit grew 66.7%, lifting price per dollar of gross profit from 14.1x to 20.2x. IES rose 100.8% on 27% gross-profit growth: 8.5x to 13.2x. Because gross margins differ — 19.8%, 25.9%, 27.4% — that ratio is the cleaner comparison across three firms selling essentially the same physical work.

All three have de-rated since May, when EMCOR traded at 31.4x trailing earnings and Comfort Systems at 56.8x. Forward, EMCOR is 23.6x against consensus 2027 revenue growth of 10.9%; Comfort Systems is 33.8x against 20.2%, roughly half its 2026 rate.

Four sessions

From 17 to 21 August the group gave back a month: EMCOR fell 9.5%, Comfort Systems 11.2%, IES 12.5%, and neighbors including Sterling Infrastructure and Dycom fell with them. No company disclosed anything. On 18 August the 30-year Treasury yield topped 5.33%, a 19-year high, and Pennsylvania removed data centers from its permit fast-track program, conditioning a sales-tax exemption on grid commitments — a schedule penalty, not a ban. The rates channel runs through customers: hyperscalers had issued nearly $223bn of bonds by 20 August, more than double all of 2025. Meanwhile only about a third of the capacity announced for 2026 completion has broken ground. None of the three has disclosed a deferral.

The setup

Where it stands — Records at all three, multiples compressing since May, and a four-day August drawdown driven by long rates rather than any disclosed order change.

Would confirm — Comfort Systems' September-quarter backlog rises again in dollars and gross margin holds above 25% without one-off estimate revisions.

Would invalidate — Any of the three reports sequentially lower backlog, or discloses a data-center project deferral or cancellation.

Watch next — EMCOR's third-quarter report in late October, the first read on whether 44% backlog growth converts at rising electrical margin.

Valuation — EMCOR 24.3x trailing and 23.6x forward earnings, 9.4x gross profit; Comfort Systems 40.7x, 33.8x and 20.2x; IES 30.0x and 13.2x.

Elastic closed its fiscal year in May with revenue of $1.74bn, up 17%, and told investors the pace from here would be slower. Its guide for the quarter it reports on 27 August is $469-470m, about 13% growth at the midpoint, with the full year set at roughly 15%. Since then the company has published no financial statement — and the shares have risen from $58.68 in late July to $85.94.

What Elastic sells into the AI stack

Elastic's product is the Elastic Stack: Elasticsearch, a distributed search and analytics engine, plus the Kibana, Beats and Logstash tools around it, sold for search, logging, observability and security work. Elasticsearch also stores vectors, the numeric representations that let a model retrieve a company's private documents at query time — the technique known as retrieval-augmented generation (RAG). The bull case is that every AI agent needs a retrieval layer, and that this traffic arrives as Elastic Cloud subscription revenue.

The disclosed numbers do not yet show it. Net expansion — what existing customers spend versus a year earlier — was 112%, and gross margin slipped to 75.4% from 76.7%. The April quarter carried a GAAP operating loss of $16.4m. MongoDB, which sells a document database with built-in vector search into the same socket, grew revenue 25.2% to $687.6m in its April quarter, with its Atlas cloud service up 29.4%. Elastic also faces Amazon's OpenSearch, a fork of its own open-source code, sold as a managed service by the largest cloud vendor. On the evidence to date it is following the data layer, not leading it.

What actually moved the shares

The advance was a grind, not a gap: Elastic cleared $62 in late July, $75 by 7 August and peaked at $87.34 on 19 August, its largest single session a 9.5% gain on 14 August. Its 50-day average crossed above its 200-day at the end of July. The proximate catalysts were an upgrade to its security-operations product toward an autonomous, agent-run threat desk, after which Truist, Oppenheimer, Jefferies and Stifel all raised targets, and a broad rotation into enterprise AI software as Wall Street's attention moved from training-cluster spending to inference and agent monetization. MongoDB rose 41% over the same thirty days.

One caution on the headline multiple: Elastic's trailing price/earnings of 24x is an artifact. The April quarter booked net income of $435.9m on revenue of $450.7m — a tax valuation-allowance release, not operations. On gross profit, the honest lens for a company at this margin, Elastic has gone from 3.85x in early May to 4.91x in late July to 6.75x now, 5.89x forward. That is a 75% re-rating while the guided growth rate fell. It is also still the cheapest name in the layer, against MongoDB at 18.5x and Datadog at 26.6x, and the shares sit 9% below their twelve-month high with a twelve-month return of only 11%.

The company where the business agrees

Snowflake, whose Data Cloud consolidates corporate data for analytics and now sells AI credits on top of it, is the mirror image. Product revenue grew 34% in the quarter reported on 27 May, accelerating from 30%, with net retention of 126% and remaining performance obligations up 38%. Gross margin has slipped two quarters running, 67.8% to 66.6%, consistent with the cost of serving AI queries landing in cost of revenue; stock compensation, at 41% of revenue in fiscal 2025, fell to 34% and is guided to 27%. Price to trailing gross profit has moved from 15.66x in early May to 34.13x. Snowflake reports on 2 September, not in August.

Datadog, which monitors software infrastructure and is the only one of the four to have published results since May, is the warning. It grew 35.6% in the June quarter, with more than 750 AI customers and agent tool calls up 22 times against late 2025 — proof the metered traffic exists. Its largest AI account then renewed at reduced usage, guidance for the September quarter came down to 28-29%, and the shares fell 4% over the same thirty days in which Elastic rose 47%. The one company that showed its numbers was the one that got marked down.

The setup

Where it stands — Elastic has re-rated 75% on gross profit since May while its own guidance points to growth slowing to 13%. Would confirm — Cloud revenue growth above 20% on 27 August with net expansion rising above 112%. Would invalidate — Total revenue at or below the $469-470m guide with net expansion flat or lower. Watch next — Elastic reports first-quarter fiscal 2027 after the close on 27 August; Snowflake follows on 2 September. Valuation — 6.75x trailing gross profit, 5.89x forward, against 3.85x in early May and MongoDB's 18.5x.