DK Street Journal

A group of seven companies that sell workflow software, cloud data tools, network security and endpoint protection to large enterprises has climbed steadily over the past month — not in a single violent spike, but in a grind that coincided almost exactly with their July earnings reports and with a separate, unrelated collapse in semiconductor stocks. The move looked unremarkable against a "flat for a year" backdrop, but that backdrop was partly an illusion: one member, CrowdStrike, split its stock 4-for-1 in early July, and unadjusted price data made it look like it had fallen 55% over 12 months when, split-adjusted, it was actually up about 78%. Corrected for the split, the group's average 12-month gain is closer to 22%, not the roughly 3% the raw numbers implied.

The seven and what they sell. ServiceNow, which runs the software many large companies use to route internal IT and HR requests, has built an AI add-on called Now Assist. Salesforce, the dominant customer-relationship-management platform, sells an AI agent layer called Agentforce. Snowflake rents cloud storage and computing for corporate data warehouses. Cloudflare operates the network infrastructure — content delivery, security, and serverless computing — that a growing share of the internet, including AI applications, runs on. Palo Alto Networks sells firewalls and cloud-security software and recently closed a $21.1 billion acquisition of CyberArk, an identity-security specialist. Zscaler sells cloud-based "zero trust" network security. CrowdStrike sells endpoint and identity protection software.

The fundamentals mostly back the move. Over the past 30 days, six of the seven rose — Snowflake +18.2%, Cloudflare +16.6%, Salesforce +11.9%, ServiceNow +7.4%, Zscaler +4.8%, CrowdStrike +4.4% — while Palo Alto was roughly flat. None appeared on the platform's list of the 30-day's most extreme movers, consistent with a gradual re-rating rather than a speculative spike. The AI revenue behind the story is dated and recognized, not just announced: ServiceNow's Now Assist AI bookings crossed $1 billion in the quarter reported July 22, with net-new AI bookings up 40% from the prior quarter and a 98% renewal rate; Salesforce's Agentforce booked revenue reached $1.2 billion, up 205% year-over-year, alongside accelerating overall revenue growth per its May earnings report; Snowflake's product revenue grew 34% year-over-year with retention actually rising to 126%, the strongest dollar growth in company history; CrowdStrike posted a record $256 million of net-new annual recurring revenue and $468 million of free cash flow, detailed in its earnings release. Not every number is clean: Palo Alto's much-touted 60% growth in its next-generation security business included $1.63 billion contributed by the CyberArk and Chronosphere acquisitions, meaning organic growth is materially slower than the headline figure, and its forward price-to-earnings multiple has roughly doubled to 84.4x. Zscaler's total annual recurring revenue grew 25%, but stripped of its Red Canary acquisition, organic net-new revenue growth was only 14% — a real deceleration. On valuation, Salesforce (13.1x forward earnings, down from 21.4x trailing) and ServiceNow (28.0x forward, well below its 40-50x historical range) look like earnings catching up to depressed multiples; CrowdStrike (164.5x forward earnings) and Cloudflare (35.7x forward sales) remain priced for continued perfection. Verdict: CONFIRMS for ServiceNow, Salesforce, Snowflake and CrowdStrike, whose disclosed metrics support the rally; INCONCLUSIVE for Palo Alto and Zscaler, where acquisition-inflated growth clouds the organic picture.

A trend-signal split that mirrors the fundamentals. Cloudflare has held an uninterrupted bullish trend signal (its 50-day average above its 200-day) for 91 straight sessions since May 5, and Snowflake and Palo Alto have done the same since late May and June. ServiceNow and Zscaler were both just upgraded out of the platform's most bearish trend category in the last week, a repair also visible in adjacent enterprise-software names Atlassian and Figma. Salesforce is the outlier: its price is up nearly 12% in a month, but its trend signal has not repaired and remains in the most bearish category — a genuine divergence between price and trend worth watching. The same 30 days that lifted this group saw a sharp semiconductor selloff — Micron down 15%, Marvell down 21%, Qualcomm down 15% — tied to weak memory pricing and AI-spending scrutiny. That is a real, concurrent rotation, but it does not explain away the software group's move: all seven companies also reported beat-and-raise earnings in the same window, making this both a rotation beneficiary and an earnings-backed re-rating.

The setup

Where it stands — Six of seven names are higher over 30 days on earnings that show AI features converting to booked or recognized revenue. Would confirm — Salesforce's trend signal repairs to match its price gain, or its next quarter shows accelerating constant-currency growth. Would invalidate — Zscaler's or Palo Alto's organic (ex-acquisition) revenue growth decelerates again next quarter. Watch next — Snowflake, Cloudflare and CrowdStrike's next earnings reports, due within the current quarter, for continued net-new revenue and retention trends. Valuation — ServiceNow trades at 28.0x forward earnings versus a 40-50x historical range; CrowdStrike trades at 164.5x forward earnings, the cohort's richest.

Companies that assemble the physical hardware inside AI data centers — the racks, power shelves and network switches hyperscale cloud providers design but don't build themselves — pulled back over the past month even as their underlying businesses kept accelerating. Six contract manufacturers in this niche fell an average of 7.6% over 30 days, a shallow and orderly retreat with none of the six landing on either end of the market's list of the period's biggest movers. That pullback followed a 12-month run that averaged 85.8% across the group. The timing matters: the stock declines came before each company reported earnings in late July, and every one of them beat estimates and raised guidance. That sequence — valuation cools first, results confirm demand second — is the opposite of what a genuine order slowdown would look like.

The six companies. Celestica (CLS), a Canadian contract manufacturer that builds servers, 800G/1.6T network switches and full racks for hyperscale cloud customers. Flex (FLEX), a diversified contract manufacturer whose Cloud and Power Infrastructure unit — including its Anord Mardix busway and power-pod business — sits alongside legacy automotive, industrial and health-solutions work. Jabil (JBL), whose Intelligent Infrastructure segment builds AI servers and networking gear next to a broader consumer- and industrial-electronics base. Sanmina (SANM), which acquired AMD's ZT Systems server-manufacturing operation and now derives most of its growth from AI compute builds. Plexus (PLXS), a smaller manufacturer weighted toward aerospace, defense and medical electronics with an emerging data-center power business. Benchmark Electronics (BHE), a similarly sized manufacturer serving aerospace/defense and medical customers whose newer AI-programs segment is growing fastest.

Business versus tape: a real divergence. For five of the six, the fundamentals argue against the price move rather than for it. Celestica's second-quarter revenue rose 62% year-over-year with its cloud/AI segment up 167%, ahead of its own guidance, and it disclosed new multi-year 2027 programs with OpenAI and AMD/Broadcom. Flex's power-infrastructure segment grew 35% with management citing over 90% of bookings already visible for the next three quarters, and its board has approved spinning off that unit into a standalone company by early 2027. Sanmina's AI infrastructure revenue jumped 173% to 62% of total sales, driving a near-tripling of full-year earnings guidance. Plexus posted record revenue and its best cash-conversion cycle in five years, and Benchmark's AI-programs segment grew 71% while its cash cycle improved to 59 days from 90. Jabil is the partial exception: its AI segment guide was raised to 46% growth, but total company revenue growth decelerated to 11.8% in its most recent quarter as its non-AI base drags on the group figure — a genuine, if narrower, business-side caution flag.

Valuation: the multiples did the work, not demand. Forward price-to-earnings ratios compressed sharply across the group even as prices fell, because guidance rose faster than stocks did — Sanmina's forward P/E dropped to 16.0x from a trailing 33.9x, and Flex's roughly halved to 25.0x from 45.3x. Celestica is the exception: its pullback was the shallowest (-2.4%) but its multiple is the richest, at 16.5x forward price-to-gross-profit versus roughly 9x-15x for peers, consistent with a first-phase re-rate from single-digit EMS multiples that this desk's own prior notes describe as largely complete. Its customer concentration has also worsened, with its top three customers now around 65% of revenue, and its free-cash-flow yield, at 1.3%, is the lowest in the group amid heavy capex.

The tape, briefly. All six names — including non-starred Benchmark and Plexus — stepped down from a strong-uptrend reading to a milder one within a three-week window in late June and July, ahead of earnings, which is the technical fingerprint of a valuation pause rather than a fundamentals break. That agrees with the earnings sequencing above rather than contradicting it.

A broader risk sits above this entire group: the hyperscale cloud customers funding this buildout are projected to spend roughly $725 billion on capex in 2026, up 77% from 2025, but three of the four largest hyperscalers lost market value after their own recent earnings calls even as spending guidance rose — a sentiment risk that has not yet shown up as an order-book signal at any of these six suppliers, but bears watching.

The setup

Where it stands — Six AI-hardware contract manufacturers fell 7.6% on average over 30 days, then all beat earnings and raised guidance in late July. Would confirm — Backlog, bookings visibility or segment guidance (e.g., Flex's 65-75% CPI growth target) holding or rising at the next quarterly report. Would invalidate — A guidance cut or customer-concentration loss at any of the six, or rising inventory/cash-conversion-cycle days signaling order digestion. Watch next — Jabil's next earnings report (its last transcript on file predates this pullback) for whether total revenue growth stabilizes or keeps decelerating. Valuation — Group forward P/E ranges roughly 16x (SANM) to 30x (PLXS), down sharply from trailing multiples of 34x-55x across the cohort.

Nine chipmakers that sell analog, power-management and radio-frequency components to automakers, factories and phone makers have fallen together over the past 30 days, part of a broader semiconductor repricing that wiped out more than $1 trillion in chip-sector market value around July 29 as investors reassessed whether near-term revenue justifies artificial-intelligence infrastructure spending. But the shared price decline conceals two very different stories underneath: at least one company is reporting its strongest order visibility in years while its stock fell harder than almost anyone else in the group, and another looks cheap on next year's earnings only because this year's are collapsing.

NXP's order book is growing while its stock is not. NXP Semiconductors (NXPI), which makes microcontrollers, RF chips and security processors sold mostly to automakers, reported second-quarter revenue up 19.5% year over year, with management explicitly stating it saw "no restocking" behind the growth and extending backlog visibility to 18 months. Book-to-bill is above 1.0 and climbing, gross margin rose to 58%, and its physical-AI design-win pipeline — chips for data centers and robotics — grew to $1.5 billion across 200-plus customers, more than double last year's total. Despite that, NXP's stock fell 19.8% over the past month, the second-worst decline in the group, leaving it at 19.0x trailing and 14.8x forward earnings — the cheapest multiple among these nine names even as consensus expects 28% earnings growth this year. That combination of accelerating fundamentals, a falling share price and the group's lowest multiple is the clearest case here of the tape pricing in cyclical weakness that the company's own numbers don't show.

Analog Devices and Microchip are recovering, unevenly. Analog Devices (ADI), a maker of analog and mixed-signal chips used across industrial, automotive and communications equipment, has posted five straight quarters of accelerating revenue growth, hitting 37% year over year in its most recent quarter, with gross margin expanding to 67%. But the recovery is lopsided: industrial bookings are running well above 1.0 book-to-bill while automotive orders remain below 1.0 and guided under seasonal levels — industrial demand is outrunning autos within ADI's own order book. Citi analysts have stayed buyers of ADI, Microchip and ON Semiconductor through the pullback, noting analog suppliers are largely shipping at consumption levels rather than still correcting channel inventory. Microchip Technology (MCHP), which makes microcontrollers and analog chips for industrial and auto customers, is rebuilding off a severe 2025 correction — revenue fell 42% that year before rebounding 35% in its latest quarter — but operating margin of 16% remains well below the 30%-plus it posted in 2022-2024, and its 23.6x forward earnings multiple prices in continued improvement that hasn't fully shown up yet.

Monolithic Power's AI exposure is real; Navitas's isn't holding. Monolithic Power Systems (MPWR), which makes power-management chips including for AI data-center server racks, posted record quarterly revenue up 48% year over year with its AI-linked Enterprise Data segment up 45% sequentially, and raised its full-year guidance for that segment. Its stock still fell roughly 9% in a single session after Broadcom's soft AI guidance in June triggered a broader rotation out of high-multiple AI-adjacent chip names, leaving MPWR at 82x trailing and 49.5x forward earnings — among the richest multiples here. By contrast, Navitas Semiconductor (NVTS), a maker of gallium-nitride and silicon-carbide power chips competing for the same Nvidia data-center power sockets against Infineon and ON Semiconductor, has seen revenue fall 27-59% year over year across recent quarters with negative gross margin in three of the last five, and its stock was cut from an uptrend to a downtrend signal in late July — a much cleaner case of an AI-premium name losing ground than Monolithic Power.

Skyworks and Qorvo: cheap against what? Skyworks Solutions (SWKS), which supplies RF chips mostly for smartphones led by Apple, has posted declining revenue in three of its last four quarters, with operating margin collapsing to 5.2% and net income down 68% year over year. Its 12.2x forward earnings multiple looks inexpensive only relative to earnings that are still shrinking. Skyworks is merging with Qorvo (QRVO), another Apple-concentrated RF supplier, to form a $22 billion combined company expected to close in early 2027; Qorvo's own operating margin has recovered to 12.3% from a 2025 trough near zero, and its 8.2% trailing free-cash-flow yield is the highest in the group.

ON's SiC business is still working through an EV slowdown. ON Semiconductor (ON), a maker of power and sensing chips including silicon-carbide components for electric vehicles, saw annual revenue fall 15% and gross margin compress from 45% to 32% in 2025 as U.S. electric-vehicle sales growth slowed sharply after federal tax-credit termination; margins have only partly recovered since. Lattice Semiconductor (LSCC), which makes low-power programmable logic chips, is genuinely re-accelerating — revenue grew 42% last quarter — but trades at 70x forward earnings, the richest multiple in the group by far. indie Semiconductor (INDI), a small automotive driver-assistance chipmaker with a $716 million market value, remains unprofitable and fell 26.7% over the past month, the worst decline among these names, though its size means it moves the group average far more than it reflects broad demand.

On the tape, most of these stocks were cut from strong uptrends to milder ones in early July, and NXP's and Skyworks' declines have been sharpest; NXP's price move looks disconnected from what its own order book shows, while Skyworks' weaker signal lines up with its shrinking revenue.

The setup

Where it stands — NXP's fundamentals are accelerating while its stock lags the group; Skyworks and Navitas show deteriorating businesses matching their weaker charts. Would confirm — NXP's book-to-bill stays above 1.0 and automotive revenue keeps growing next quarter without a channel correction. Would invalidate — NXP's backlog visibility shortens or automotive orders reverse below 1.0 book-to-bill in the next earnings report. Watch next — NXP Semiconductors' next quarterly report, expected late October 2026. Valuation — NXP trades at 19.0x trailing and 14.8x forward earnings, the cheapest multiple in the group despite the fastest-growing backlog.

What happened

Eight companies that sell IT consulting and outsourced software work — from Accenture, the Dublin-based professional-services giant, to Grid Dynamics, a $553 million engineering boutique — rose 9% to 31% in a single month. The move wasn't triggered by anything specific to the industry. In mid-July, a more than $1 trillion selloff in AI-chip and semiconductor stocks sent global investors rotating into software and services names seen as insulated from the AI-hardware cycle. India's Nifty IT index, which includes Infosys and Wipro, gained roughly 16.7% in July — its best month in years, and its widest margin over chip stocks since 1999. That is a rotation, not proof the underlying businesses turned. Pulling the eight names apart shows why the distinction matters.

The two names with numbers to match the tape

Cognizant, a New Jersey-based consultancy built around banking and healthcare technology work, rose 31% in the month — the largest move of the eight — and it is the one large-cap where the business backs up the stock. Its financial-services division grew revenue 12% year-over-year for a second straight quarter, the company has now expanded adjusted operating margin for six consecutive quarters to 16.0%, and it raised full-year earnings guidance in its July 29 call while signing seven deals worth over $100 million each. It trades at 9.6 times forward earnings, the cheapest multiple among the three mega-caps in the group. Grid Dynamics, a small engineering firm serving Fortune 1000 clients, is the other confirmed story: revenue grew 7% while headcount fell 3%, AI-related work reached 30.7% of revenue and grew 55% year-over-year, and management said there is no pricing pressure — meaning artificial intelligence is adding productivity rather than deflating what clients pay. Both are CONFIRMS: business and tape agree.

Where the tape is ahead of the business

Accenture rose 21% in the month, but this is the same company that posted its worst single-day stock decline on record on June 18, cutting full-year guidance after U.S. federal-spending cuts and Middle East disruption — not AI substitution. Its bookings fell 2% year-over-year to $19.3 billion even as revenue grew 6%, and the company has stopped disclosing discrete AI-bookings figures altogether, so there is no way to check what fraction of its AI narrative is converting to recognized revenue. Its trend band has not upgraded and remains in the same downtrend it entered March 6. Infosys, a Bengaluru-based outsourcer, told investors on its July 23 call that it is seeing real price "deflation" on large contract renewals and cut its already-modest 2027 revenue growth guidance to roughly 1.5%-3%, with true organic growth closer to 0.5% once acquisitions and mix effects are stripped out — a rare instance of a company directly confirming the AI-pricing-pressure thesis this whole group has traded against all year. Its stock stayed in its downtrend, which is at least consistent. Wipro, its Bengaluru-based peer, is the sharpest divergence in the set: its trend indicator upgraded to neutral on July 28, twelve days after it guided to a sequential revenue decline for the current quarter and described large deals as "slipping" rather than closing. That is a chart moving one way while the business guides the other.

The mid-sized names: mixed and murky

CGI, the Montreal-based firm whose book leans on Canadian and European government contracts, rose 10% but organic growth is essentially flat once acquisitions are excluded — a "stopped falling" story at an 8x forward multiple, not an accelerating one. Globant, a Luxembourg-domiciled digital-engineering firm pivoting toward AI subscription-style delivery, rose 14% despite revenue shrinking year-over-year in four of its last five quarters; its AI Pods initiative targets just $60-100 million in annualized revenue by year-end against a $2.45 billion total business, and its forward P/E of roughly 6 times reflects a shrinking base as much as a bargain. EPAM, a digital-platform engineering specialist, saw its trend indicator flip to neutral on the same July 28 date as four peers — but its last earnings call was May 7, meaning the market re-rated it with no fresh company disclosure at all, pointing to sector rotation rather than a fundamental trigger.

The tape, read carefully

Five of the eight names — EPAM, Grid Dynamics, CGI, Globant and Wipro — flipped from bearish trend bands to neutral on the identical date, July 28, which is itself evidence of a rotation event rather than five separate company stories; two of those flips (CGI, Grid Dynamics) came before those companies had even reported earnings. Accenture, Cognizant and Infosys, the three largest names, never joined the upgrade and remain in downtrends that began March 6 and March 19, now 135-148 sessions long.

The setup

Where it stands — Cognizant and Grid Dynamics show fundamentals matching their rally; Accenture, Infosys, Wipro and Globant show a chart move ahead of, or contradicting, recent guidance. Would confirm — Infosys or Wipro reporting constant-currency organic growth reaccelerating above 3% with book-to-bill over 1.0 next quarter. Would invalidate — Cognizant's operating margin contracting or bookings falling below the $27 billion trailing level cited entering 2025. Watch next — Accenture's fiscal Q4 2026 results, due late September, for whether bookings return to growth after the 2% Q3 decline. Valuation — Cognizant trades at 9.6x forward earnings versus 11.8x trailing; Accenture at 12.0x forward versus 13.1x trailing; Infosys at 15.4x forward, above its own trailing 14.8x.

Four networking-equipment makers that sell the switches and routers moving data inside AI data centers and corporate campuses have all rallied over the past year, but the reasons behind each stock's gain are almost unrelated to one another — and the weakest 12-month performer is the one most investors would call the AI-networking pure play.

Arista Networks (ANET), which designs high-speed Ethernet switches sold mainly to cloud giants for AI and cloud data centers, is up 56.8% over the past year — the smallest gain in the group despite raising its 2026 AI-networking revenue target twice this year, most recently to $3.5 billion as Ethernet continues to take share from Nvidia's InfiniBand interconnect inside cloud clusters. Quarterly revenue growth reaccelerated to 35.1% year-over-year in the March quarter, but gross margin has slid from 65.2% to 61.9% over four straight quarters as white-box switch makers and Nvidia's own Spectrum-X networking gear compete for the same sockets. IDC data show Nvidia has overtaken Arista as the top vendor by revenue in datacenter Ethernet switching, a shift analysts say reshapes the competitive map Arista built its lead on. Its forward earnings multiple has expanded to 50.8x from the roughly 41-45x cited in the user's own notes as recently as July — growth is real, but the stock has outrun the anchor it was priced against even as margins compress. CONTRADICTS-leaning: business growth confirms, valuation does not.

Cisco Systems (CSCO), the longtime enterprise-networking and security giant, is up 72.8% over 12 months on genuinely accelerating results — quarterly revenue growth stepped up from 7.5% to 9.7% to 12.0% over three straight quarters with operating margin expanding to 25%. Management guided hyperscaler AI-infrastructure orders to roughly $9 billion for the fiscal year, but only about $4 billion is expected to convert to recognized revenue in fiscal 2026, with $6 billion more pushed into fiscal 2027 — orders are outrunning recognized revenue. The stock's trend was downgraded from a strong to a milder uptrend on July 20, and it was flat over the past 30 days even as its forward earnings multiple expanded from roughly 18x to 24.2x since May. CONFIRMS on fundamentals, STRETCHED on price.

Hewlett Packard Enterprise (HPE), which sells servers and — since absorbing rival Juniper Networks in 2026 — wired and wireless networking hardware, is the cohort's standout: up 152.8% over 12 months off a depressed February low tied to acquisition writedowns. Networking segment revenue grew 151.5% year-over-year in the March quarter, now over 30% of company revenue but more than half its profit, and the company now expects to hit its 2028 Juniper cost-savings targets two years early. Net income swung from a $1.05 billion loss to $624 million profit in a year. Its forward P/E rose only to 14.7x from roughly 10-13x in May notes despite the stock nearly tripling, because consensus earnings estimates were revised up in step; free-cash-flow yield sits at a cohort-best 8.5%. CONFIRMS on both counts — though the user's own notes flag risk if hyperscalers in-source GPU capacity or if rival Dell's server economics keep beating HPE's.

Extreme Networks (EXTR), a small-cap campus and Wi-Fi networking vendor selling mainly to schools, hospitals and government agencies, is up 73.3% over 12 months but has rolled over 4.8% lower in the past 30 days after a run built on Wi-Fi 7 and AI-automation product launches. Revenue growth has decelerated for four straight quarters, from 19.6% to 11.4%, while the stock trades at a rich 63x trailing enterprise value to EBITDA. CONTRADICTS: the tape is now catching down to slowing fundamentals.

Sector-wide, the four largest hyperscalers are guiding to roughly $630 billion of combined 2026 capital spending, up from about $388 billion in 2025, a tailwind supporting continued networking-port growth broadly — but it is not lifting all four names evenly. None of the four appear among the market's most extreme 30-day movers in either direction; this is a gradual, not violent, divergence. HPE alone has held an unbroken uptrend, its 50-day average above its 200-day, for roughly 100 straight trading sessions since April 23.

The setup

Where it stands — Four networking stocks share a headline 86% average 12-month gain, but HPE and Cisco fundamentals confirm it while Arista's margin and Extreme's growth are cooling. Would confirm — Arista gross margin stabilizes above 62% and Cisco's AI order backlog converts to recognized revenue at guided fiscal 2027 rates. Would invalidate — Extreme Networks revenue growth decelerates below 10% next quarter, or Arista margin falls below 60%. Watch next — Cisco's fiscal fourth-quarter 2026 results (guided $16.7-16.9 billion revenue) and Arista's next quarterly report for gross-margin trend. Valuation — HPE trades at 14.7x forward earnings vs. roughly 10-13x in May notes; Arista at 50.8x forward vs. roughly 41-45x cited in July notes.

Companies that staff call centers, back offices and data-analytics teams on behalf of other businesses have spent the past year being priced as artificial intelligence's most direct casualty — every chatbot that answers a customer call is a job these firms used to sell by the hour. Five of the largest publicly traded names in that business rose an average of 12.5% over the past 30 days even as their one-year return remains deeply negative. But the rally is not evidence the market has decided AI fears were overdone across the board: it is concentrated in two companies that actually reported accelerating, AI-led revenue growth, while the sector's weakest and strongest fundamental performers each moved in ways a simple "AI reprieve" story wouldn't predict.

The AI-disruption backdrop is real. Swedish fintech Klarna disclosed that its AI assistant now handles the workload of roughly 700 full-time contact-center agents, and shares of European outsourcer Teleperformance fell 29.3% in a single day on the read-through. Research firm Gartner projects contact centers will need 30-40% fewer human agents for the same call volume by 2026, cutting an estimated $80 billion from industry labor costs, and consulting giant Capgemini closed a $3.3 billion acquisition of outsourcer WNS last October — a consolidation move that only makes sense if standalone scale in this business is eroding. That is the backdrop against which July's bounce has to be judged.

ExlService Holdings (EXLS), a data-analytics and digital-operations firm serving insurance, healthcare and banking clients, led the group with a 27.6% 30-day gain after its July 29 second-quarter report. Revenue grew 16% year over year to $594.8 million, beating estimates, and the company raised full-year guidance to $2.39-$2.415 billion. Crucially, data- and AI-led services now make up 61% of total revenue and grew 30% year over year — a faster clip than the legacy analytics base, and EXL is closing an acquisition of AI data-annotation firm iMerit to extend that lead. Revenue growth has accelerated for four straight quarters (12.2% to 15.6% year over year), and forward price-to-earnings of 14.96x sits well below its trailing 21.53x, implying the market expects more of this growth to show up in earnings. The business explains the move.

Genpact (G), a Bermuda-based business-process and IT-services firm serving banking, insurance and manufacturing clients, gained 22.1% in price over 30 days on genuinely strong numbers: first-quarter revenue grew 6.7% to $1.296 billion with operating margin expanding to 16.1% from 15.1%. Its Advanced Technology Solutions segment — data, AI and "agentic" automation — grew 24% year over year and now represents 27% of revenue; first-quarter agentic bookings alone nearly matched all of 2025's total contract value, a year that closed with a record $5.5 billion in new bookings. At 8.63x forward earnings and 8.17x trailing enterprise value to EBITDA, Genpact remains the cohort's cheapest name even after the rally.

NIQ Global Intelligence (NIQ), a Nielsen spinoff that sells consumer-shopping and retail-measurement subscriptions, rose 17.0% on Buy ratings from Stifel and Barclays, a new AI-focused executive hire, and a new retail-data agreement with Circle K — but its next earnings report isn't due until August 10, so this move is running ahead of any confirming numbers. The company's most recent reported quarter showed revenue growth decelerating sharply to 1.3% year over year from 23.4% a year earlier, the cohort's clearest organic slowdown, even as a 2025 refinancing of $3.4 billion in debt cut annual interest expense by nearly $100 million.

Concentrix (CNXC), the largest pure contact-center and back-office outsourcer in the group, is the name most directly exposed to "AI answers the call we used to staff," and its stock managed only a 5.7% 30-day gain — the weakest in the cohort, and still down roughly 46% over the past year. The numbers explain why: constant-currency revenue grew just 0.6% last quarter, gross margin fell to 33.4% from 35.1%, operating income dropped 35.7%, and the company cut full-year guidance citing accelerating client offshoring. Its AI platform, IX Suite, is targeting only about $120 million in annual recurring revenue against roughly $10 billion in total revenue. A forward price-to-earnings ratio near 2x looks cheap only because heavy debt from its Webhelp acquisition and a prior $1.48 billion impairment have crushed the equity value underneath it.

Innodata (INOD), a data-engineering firm that builds AI training and evaluation datasets for hyperscale cloud customers, is the cohort's inversion: the strongest fundamentals, the weakest recent stock. Revenue grew 54.4% year over year last quarter with gross margin expanding to 43.4%, and business from customers outside its top client grew 453% year over year, chipping at a customer-concentration risk that still sees one client generate roughly 57% of revenue. Yet the stock fell 7.9% over the past 30 days after nearly doubling earlier in 2026, and trades at 58.6x forward earnings and 34.3x trailing EBITDA — three to five times richer than any peer here. This looks like valuation digestion after a run, not a fundamental reversal.

On the tape, the divergence between price and trend is the story: EXLS and NIQ both flipped from a locked-in downtrend to neutral around their catalysts, a genuine multi-step repair for NIQ (strong downtrend to mild downtrend to neutral across July). Genpact's trend indicator, by contrast, has stayed pinned in its worst downtrend category continuously since March 9 despite the 22% price gain — a clean divergence between the business, which is improving, and a chart that hasn't caught up. Concentrix's downtrend never broke at all.

The setup

Where it stands — EXL and Genpact show accelerating, named AI-led revenue and re-rated multiples; Concentrix's core business is still shrinking; Innodata and NIQ await confirmation. Would confirm — EXL's data/AI-led revenue mix keeps growing above 25% year over year for two more quarters while total company margins hold or expand. Would invalidate — Genpact's Advanced Technology Solutions growth decelerates back toward its legacy blended rate, or its trend indicator fails to exit its downtrend despite the price gain. Watch next — NIQ reports second-quarter results August 10, the first real test of whether its 17% rally has fundamental support. Valuation — EXL trades at 14.96x forward earnings versus 21.53x trailing; Genpact at 8.63x forward versus 8.17x trailing EV/EBITDA, both near multi-year lows.

Six smaller software companies that investors have spent the past year pricing as the most vulnerable to AI-driven disruption — the ones selling content-delivery, search, developer tools, digital banking, point-of-sale and call-center software — jumped an average of 15% over the past month after reporting July earnings. But the move is not one story. Two of the six posted genuine acceleration in growth and margins; one rally is riding unconfirmed takeover chatter; one company's AI-product growth is still losing the race against its shrinking legacy base; and the name with arguably the best quarter of the six — Lightspeed — fell anyway.

The six, and what they actually do. Fastly (FSLY) runs an edge-computing and content-delivery network that speeds up and secures web traffic for media and e-commerce sites. Elastic N.V. (ESTC) makes the Elasticsearch search-and-analytics software, now marketed as a database for AI applications. Q2 Holdings (QTWO) sells cloud-based online-banking software to regional and community banks. GitLab (GTLB) sells a platform developers use to write, test and deploy code, including an AI assistant called Duo. Five9 (FIVN) sells cloud software that runs corporate call centers, including AI tools that can handle calls without a human agent. Lightspeed Commerce (LSPD) sells point-of-sale and payments software to retailers and restaurants.

Where the numbers back the move. Q2 Holdings' July 29 quarter showed revenue up 13% and adjusted EBITDA up 37% with margin expanding 510 basis points; subscription annual recurring revenue rose 15% and backlog climbed 17% to $2.8 billion, and management raised full-year guidance and authorized $350 million more in buybacks after retiring its convertible debt. The stock is up 17% over 30 days. Elastic's fiscal 2026 revenue rose 17%, with net revenue retention near 112% and gross margin at 76%; forward price-to-sales of 3.6x is the cheapest in the group against mid-teens growth, and shares are up 19%. Both moves trace to disclosed results, not sentiment — the business explains the tape here (CONFIRMS).

Where it's murkier. Fastly's security and compute revenue grew 47% and 67% in the first quarter and management raised full-year guidance, which explains part of its 27% rally, the largest in the group. But that guidance implies growth decelerating to roughly 15% from 20% in the prior quarter, and Fastly does not report its next quarter until after the close on August 5 — two days from now — so a meaningful share of the run-up, including a one-week gain some trackers put near double digits, is positioning ahead of a print, not a reaction to one. The company also benefits from surging AI-crawler and agent traffic that grew roughly 30% from January through May, a genuine tailwind, but forward price-to-sales of 5x and a 71x forward earnings multiple leave little room for a soft print (INCONCLUSIVE pending Wednesday's results).

GitLab's quarter showed revenue up 23%, but full-year guidance was cut to 16-17% growth from a prior 26%, and only about $20 million of its roughly $1 billion revenue base is disclosed AI-assistant run-rate, while a fifth of its accounts face seat contraction from customer layoffs. Its 9% rally coincides instead with reports that Datadog is working with Morgan Stanley on a renewed takeover approach that could exceed $60 a share, unconfirmed by either company. Forward sales multiples have jumped from 4x to 5.4x in three months against a guide that got slower, not faster — the valuation move has outrun the business (CONTRADICTS).

Five9's AI-driven contact-center revenue grew 68% year over year and now makes up 13% of subscription revenue, but total company revenue growth has decelerated for four straight quarters, from 12.4% to 9.2%, meaning AI gains are not yet clearly outrunning erosion in its traditional per-agent business. Its forward sales multiple has nearly doubled to 1.8x even as the stock still trades at nine times forward earnings — cheap by absolute standards, ambiguous by trend (INCONCLUSIVE).

The clean divergence. Lightspeed's core North American retail and European hospitality segments grew 21-24%, EBITDA turned positive, and the company concluded a strategic review by staying public while upsizing its buyback and repurchasing $86 million of stock in the June quarter. None of that stopped the stock from falling 3% over the same 30 days — the one name in the group where the business and the tape moved in opposite directions, and, at 1.1x forward sales, the cheapest stock in the cohort.

Sector context. This is not an isolated pocket. The broader software complex that had been priced as AI casualties also re-rated in late July: ServiceNow, Atlassian and Figma all flipped from deeply bearish to mildly bearish trend readings within the same week, part of a broad rate-cut-driven software rebound that left ServiceNow still down roughly 45% over the trailing year even after bouncing. Datadog, by contrast, held an uptrend continuously since June with no bearish reading at all. The smaller, lower-moat names in this story are the higher-beta expression of the same rotation — some of their gains are sector-wide relief, not company-specific proof.

Technicals. Five of six names flipped from bearish to bullish trend bands between mid-July and early August, but only Five9 and Fastly have held that reading since June; Elastic, GitLab and Q2 Holdings flipped only on the final trading day of July, with no multi-week confirmation yet, and Lightspeed has not reached a strong uptrend at all.

The setup

Where it stands — Five of six lower-moat software names rallied on July earnings, but only Q2 Holdings and Elastic show fundamentals clearly outrunning their valuation re-rating. Would confirm — Fastly's August 5 guidance holds or raises its implied ~15% growth rate rather than confirming further deceleration. Would invalidate — GitLab's forward sales multiple stays above 5x with no confirmed takeover bid and guidance is cut again next quarter. Watch next — Fastly reports fiscal Q2 2026 results after market close August 5, 2026. Valuation — Elastic trades 3.6x forward sales versus 4.1x trailing; GitLab 5.4x forward versus roughly 4x three months ago on a cut guide.

Every AI data center needs someone to write the check before the chips arrive. Over the past year, the businesses doing that writing have split into two camps with opposite stock-market outcomes. Investment banks Goldman Sachs and Morgan Stanley, plus infrastructure owner Brookfield Infrastructure Partners, have rallied 27% to 47%. Meanwhile the six largest private-equity and credit managers actually underwriting the debt and equity behind hyperscaler and "neocloud" data-center buildouts — Blackstone, KKR, Apollo Global Management, Brookfield Asset Management, Ares Management and Carlyle — have fallen 12% to 29%, even though every one of them told investors this summer that their fee income from managing money is growing 20% to 38% a year.

The lenders, in brief. Blackstone (BX) is the world's largest alternative-asset manager, with a data-center platform now valued at $185 billion including its QTS colocation business and a new $35 billion financing partnership with Broadcom for AI-chip deployment. KKR runs Helix, a $10 billion-plus vehicle backed by Nvidia and Vistra built to fund hyperscaler power and compute needs. Apollo Global Management (APO) co-anchored that same $35 billion Broadcom financing platform. Ares Management (ARES) operates Ada Infrastructure, seven data-center campuses totaling roughly 1 gigawatt of compute, with a dedicated digital-infrastructure fund closing later this year. Carlyle (CG) and Brookfield Asset Management (BAM) are diversified private-equity and infrastructure managers with smaller, less-detailed AI-financing disclosures. Brookfield Infrastructure Partners (BIP) directly owns roughly 50 data centers plus fiber and power assets. Goldman Sachs (GS) and Morgan Stanley (MS) are investment banks earning advisory and financing fees on the buildout. Hannon Armstrong (HASI) is a smaller climate-infrastructure lender.

Fundamentals versus the tape: a real divergence. For the six PE-style managers, the numbers do not match the stock reaction. Blackstone's fee-related earnings rose 22% and infrastructure assets under management 40% in its latest quarter; KKR's fee-related earnings per share rose 34% with $72 billion of committed capital still to start earning fees; Ares posted a record $36 billion fundraising quarter and 20% fee-earnings growth. None of that shows up in the share prices. Instead, the selloff traces to a separate scare: Apollo-managed business-development company MidCap Financial cut its dividend and marked down its portfolio roughly 3% on weak software loans, and fears that AI could obsolete private-equity-owned software companies hit the whole group this year. A separate scare in neocloud financing — CoreWeave's funding partner Blue Owl stumbling on a Pennsylvania data-center project — spread investor anxiety about AI-infrastructure lending broadly, dragging even firms with growing, well-disclosed data-center books. This is the pattern the desk should flag most: business fundamentals accelerating while the stock discounts credit stress that, so far, sits elsewhere in the portfolio.

Where the credit risk is real. It isn't imaginary. KKR's own management flagged that hyperscaler data-center loan spreads have recently widened and large deals are showing "indigestion," prompting more selectivity. Separately, tenant quality is starting to price differently: loans against data centers anchored by investment-grade hyperscaler tenants price tighter and need less equity per megawatt than those anchored by speculative-grade neoclouds. Ares, for its part, reports direct-lending non-accruals below 2% with underlying portfolio earnings still growing — evidence the deterioration is early and selective, not systemic, at least in disclosed vehicles.

The banks and the landlord look supported. Goldman and Morgan Stanley both posted record quarters built partly on AI-financing activity — Goldman's advisory revenue rose 17% on $1.2 trillion of announced deals, Morgan Stanley raised its 2026 AI data-center capital-spending forecast to $850 billion — and both trade at forward price-to-earnings multiples roughly flat to trailing (14-17x), meaning the market isn't pricing aggressive further upside, just current results. Brookfield Infrastructure's Data segment funds from operations rose 36%, but roughly $1.2 billion of its year-to-date cash came from selling stakes in a colocation IPO rather than new leases — a genuine but partly one-off contribution to the number behind its recent upgrade to a stronger uptrend.

The thin one. Hannon Armstrong's public disclosures mention data centers only in passing, tied to clean power rather than any specific committed lending; it was itself downgraded from a strong to a mild uptrend in the past month, suggesting its inclusion in any "data-center financing" grouping does more marketing work than balance-sheet work.

Technicals. As of August 3, the six PE-style managers all carry bearish trend readings (Ares strongly so, the other five mildly) despite 30-day bounces of 6-14% tied to rate-cut optimism rather than data-center news specifically, as seen when Apollo and KKR both surged more than 4% in a session on easing private-credit stress. Goldman, Morgan Stanley and Brookfield Infrastructure carry strong uptrends; Hannon Armstrong a milder one. Valuation gaps are stark: Apollo's price-to-earnings falls from 38x trailing to 15x forward, Ares from 61x to 24x, Carlyle from 32x to 13x — pricing in continued double-digit fee growth that, so far, is exactly what these firms have delivered.

The setup

Where it stands — Six AI data-center lenders trade near multi-year-low fee-growth-adjusted multiples while fee income and disclosed committed capital both expanded last quarter. Would confirm — Fee-related earnings growth stays above 20% and non-accruals in affiliated private-credit BDCs remain below 2% next quarter. Would invalidate — Non-accrual or PIK-income ratios rise materially in KKR, Ares or Blackstone credit vehicles tied to data-center or neocloud borrowers. Watch next — Q3 2026 earnings calls (October-November) for updated data-center AUM, spread, and non-accrual disclosures across BX, KKR, ARES, APO. Valuation — Forward P/E of 13-24x across the six managers versus 30-61x trailing, both well below their own recent-year highs.

Six companies that sell, distribute, install or operate enterprise computer systems have moved together over the past month, but the businesses behind the tickers just told six different stories — one of them a genuine earnings crash. International Business Machines (IBM), the mainframe-to-consulting giant, plunged roughly 25% intraday on July 22 after reporting Z-series mainframe revenue down 42% and cutting its full-year revenue-growth outlook to 4-5% from more than 5%, a move covered by Briefs.co. Chief Executive Arvind Krishna attributed part of the mainframe collapse to customers pulling purchases forward ahead of an expected price increase, according to Forbes — a pull-forward hangover, not new competitive damage. IBM is still down about 11.7% over the trailing 30 days, the worst showing in the group, which matters because IBM's $214 billion market value is large enough to swing a cap-weighted average — as a drag, not a lift.

The name that actually led was CDW, a $19 billion reseller that sells hardware, software and integrated data-center and networking solutions to corporate and government IT buyers. Its trend signal flipped from a mild uptrend to a strong one on July 28, the only genuine bull-signal upgrade among the six — but it happened with no earnings call of its own; CDW last reported May 6. The move instead rode a broad market rally after the Federal Reserve held rates steady, when the S&P 500 rose 1.7% and the Nasdaq 2.8% on July 30, ending a six-day Nasdaq losing streak, plus a read-through from Ingram Micro's print two days earlier. On the fundamentals, CDW's own Q1 revenue grew 9.2% year-over-year to $5.68 billion, an acceleration from 6.8% growth for all of 2025, but gross margin compressed to 20.95% from 21.59% and operating margin slipped slightly — consistent with a hardware-heavy sales mix rather than higher-margin services upsell. CDW trades at 13.7 times forward earnings versus 17.9 times trailing, a gap that only closes if 2026 profit grows roughly a third from 2025's $8.08 a share.

Ingram Micro, the $6.4 billion distributor whose Xvantage platform and CloudBlue marketplace route hardware and cloud subscriptions to resellers worldwide, is the name CDW's rally leaned on — yet Ingram's own signal was already at its strongest reading for most of July, before any "turn," and was downgraded to neutral by month-end even after it beat second-quarter estimates on July 30 (revenue $14.5 billion versus $13.9 billion expected, adjusted profit $0.82 a share versus $0.78 expected), a beat detailed by Investing.com. The stock popped into the print and gave the gain back the next day. Revenue grew 13.6% and operating income 38.6%, but gross margin guidance implies roughly 6.8% — a distributor's thin, AI-hardware-pressured spread — and trailing free cash flow is negative on working-capital swings. Ingram trades at 8.3 times forward earnings versus 15.2 times trailing, and is up about 43% over the past year, the only positive one-year return in a group otherwise tagged a shared downtrend.

DXC Technology, a $1.8 billion outsourcer that runs IT infrastructure and business-process work for large enterprises, showed the most credible improvement, moving from a strong downtrend in early July to neutral by month-end. Its July 30 quarter backed that up: company-wide bookings-to-billings ran 0.99 times, the best first quarter in three years, and infrastructure-outsourcing bookings jumped 35% year-over-year even as that segment's revenue fell 11% and margin held at just 2.6%. Total revenue was still down 5.1%, and management kept guidance for a 3-5% organic decline; free cash flow of $314 million included a one-time $214 million litigation settlement.

Kyndryl, the $3.1 billion managed-infrastructure provider spun off from IBM in 2020, and N-able, a $950 million maker of remote-monitoring software for IT service providers, both rallied on no fresh print of their own — Kyndryl reports August 5, and N-able last reported May 7. Kyndryl's full-year revenue was roughly flat but gross margin expanded to 21.8% from 20.9%, a genuine post-spin margin story; the stock is still down 64% over the past year, the group's steepest drawdown. N-able's stock rose the most of the six over 30 days, about 57%, while its gross margin has compressed for three straight fiscal years, from 83.8% to 76.8%, a trend one market note tied to AI substitution risk in MSP software — margin direction moving opposite the share price.

Verdict: The business case CONFIRMS for DXC (bookings genuinely improving) and partially for CDW (revenue reaccelerating, though on a lower-margin mix); it CONTRADICTS for IBM (price cratered despite still-positive, if decelerating, fundamentals) and for N-able (price up sharply while margins erode); it is INCONCLUSIVE for Kyndryl and Ingram Micro pending confirming prints. Valuation is not the constraint — every name's forward multiple sits well below its trailing one — but that gap reflects earnings expectations the smaller names haven't yet proven.

The setup

Where it stands — CDW alone carries a fresh bull signal on no earnings news; IBM, the group's largest member, is still recovering from a mainframe-driven crash. Would confirm — Kyndryl's August 5 report shows signings growth and margin expansion continuing without revenue re-accelerating. Would invalidate — CDW's next quarter shows gross margin still compressing even as revenue growth decelerates back toward its 2025 pace. Watch next — Kyndryl reports fiscal first-quarter results on August 5, 2026. Valuation — CDW trades at 13.7x forward vs. 17.9x trailing P/E; Ingram Micro 8.3x forward vs. 15.2x trailing; both below their own recent norms.

In mid-July, Iranian strikes knocked out roughly 17% of Qatar's liquefied natural gas (LNG) export capacity, and QatarEnergy's chief executive said full repairs could take three to five years, according to reporting on the outage. The shock sent global gas benchmarks sharply higher and rippled through America's four major LNG export companies — but not evenly. One captured a spot-market windfall, two kept collecting largely fixed contract fees, and the smallest, which still has no revenue, fell on a disappointing earnings report the same month.

A cohort average that hides a divergence

Over the 30 trading days through August 3, Cheniere Energy (LNG), the country's largest LNG exporter and operator of the Sabine Pass and Corpus Christi terminals, gained 3.7%. Cheniere Energy Partners (CQP), the limited partnership that actually owns and runs the Sabine Pass terminal, gained 5.8%. Venture Global (VG), which builds modular LNG plants at Calcasieu Pass and Plaquemines and is completing a third facility, CP2, jumped 21.4%. NextDecade (NEXT), still building its Rio Grande LNG terminal in Texas and carrying zero revenue in every quarter on record, fell 13.4% after its July 30 earnings report. Averaged together, the four names look like a quiet 4.5% drift — the kind of move that might suggest a sector-wide re-rating. It isn't one; it's a single stock's spike offsetting another's earnings-driven slide, with the two Cheniere entities moving on business fundamentals in between.

Why Venture Global jumped

Venture Global's stock move — it rose as much as 13.7% in a single session, per Qatar-outage coverage — is a spot-price story. Unlike Cheniere, which locks most of its capacity into long-term contracts, Venture Global sells a larger share of cargoes on the spot market, so it benefits directly when Asian and European gas benchmarks spike; the Japan-Korea Marker stood at $21.38 per million British thermal units and European TTF at €58.32 per megawatt-hour in late July against a Henry Hub price of just $2.95. But the rally hasn't obviously priced an unresolved legal overhang: an arbitration panel ruled against Venture Global in October 2025 in a dispute with BP over delayed cargo deliveries, with BP seeking more than $1 billion in damages at a hearing expected in 2026, according to arbitration coverage — even as a judge separately rejected Shell's attempt to overturn a prior award Venture Global had won against it, per a related ruling. Revenue growth is also decelerating fast, up 58.9% year over year in the first quarter of 2026 versus 191.7% the prior quarter, and gross margin compressed to 35% from 55.9% a year earlier as heavy construction spending continues.

Cheniere and its partnership: real growth, already re-rated

Cheniere's case is the strongest of the four: first-quarter 2026 revenue rose 24.8% year over year to $6.65 billion, an acceleration, and the company raised full-year 2026 EBITDA guidance to $7.25 billion-$7.75 billion and distributable cash flow guidance to $4.75 billion-$5.25 billion. Its Corpus Christi Stage 3 expansion is nearly finished, with a sixth train producing LNG and expected to enter commercial service this summer, per industry reporting; the company is also targeting a final investment decision on a Sabine Pass expansion in 2026 or 2027 that would push combined capacity above 90 million tonnes a year, according to FID coverage. Cheniere Energy Partners grew revenue a steadier 20.4% the same quarter. Both booked large non-cash GAAP losses from Henry Hub-linked hedging derivatives, muddying headline earnings even as operating margins held. The catch: the multiple has already moved. Cheniere's trailing enterprise-value-to-EBITDA ratio is 12.0x today versus roughly 7.3x-8x three months ago — the growth is real, but much of the re-rating this kind of setup would look for has already happened.

NextDecade: still an option, not yet a business

NextDecade has never booked a dollar of revenue. Its first Rio Grande LNG train is 74% complete and targeting first LNG in the first half of 2027; the company just closed $4.5 billion of project debt, including an inaugural $3.5 billion bond offering, to fund construction without diluting shareholders — but its net loss widened to $65.4 million in the second quarter and the stock fell roughly 20% on the July 30 print, per earnings coverage. Trains 4 and 5 were financed with project debt rather than equity, but the company's own disclosures point to roughly $2 billion-$2.5 billion of fresh equity needed per train for Trains 6 through 8 — dilution risk the current price does not obviously reflect.

The tape agrees with the split, not the average

None of the four names appear on any 30-day list of extreme movers; even Venture Global's 21% gain falls short of the roughly 39%-plus threshold used to flag violent moves. That's consistent with a name-by-name story rather than a coordinated sector breakout — the technicals confirm the fundamentals split, not the cohort-level average.

The setup

Where it stands — Cheniere and its partnership grow on contracted fees at a richer multiple; Venture Global rides a spot spike with unresolved litigation; NextDecade remains pre-revenue. Would confirm — Venture Global's next quarter shows contracted (not just spot) revenue growth while the JKM-Henry Hub spread stays wide. Would invalidate — The JKM/TTF-Henry Hub spread narrows sharply toward shipping cost, or the BP arbitration damages hearing produces a large award. Watch next — NextDecade's Train 1 first-LNG target in the first half of 2027; Cheniere's Sabine Pass FID expected 2026-2027. Valuation — Cheniere trades at 12.0x trailing EV/EBITDA versus ~7.3x-8x three months ago; NextDecade has no meaningful multiple, pre-revenue.