DK Street Journal

The six companies that build, lease and power the physical data centers AI computing providers rent by the megawatt have fallen together over the past month, giving back a chunk of a run that made this corner of the market one of the best-performing over the past year. But their actual leases, revenue and balance sheets tell three separate stories: two Chinese wholesale hosts trading at a steep discount to U.S. peers while revenue accelerates, one U.S. operator that already worked off its richest multiple through real contract growth rather than a falling stock price, and three much smaller, newer companies ranging from a credible growth story to a disputed contract under litigation.

Applied Digital (APLD), which builds and leases high-performance-computing data centers to AI companies in North America after abandoning crypto mining in 2022, is the group's largest by market value at roughly $7.8 billion. Fiscal 2026 revenue tripled to $611.3 million as leases with CoreWeave and an undisclosed investment-grade hyperscale customer began paying off — a 150-megawatt lease expansion finalized this year brought total CoreWeave-related capacity to 400 megawatts and roughly $11 billion of contracted revenue across three roughly 15-year leases, with the first 100 megawatts reaching full service in November 2025, on schedule rather than delayed. That growth cut the trailing price-to-sales multiple from 28.7x three months ago to 12.8x today (9.6x forward) — the stock's 30-day, 17% pullback looks more like a re-entry into an already-cheaper name than a re-rating of an expensive one. The company still lost $244 million on an operating basis funding that buildout, and convertible notes priced in 2024 carry dilution protection capped at $14.72 a share — no cushion now that the stock trades at $27.39, a real dilution risk if shares stay above that level into conversion.

GDS Holdings (GDS) and VNET Group (VNET), which build and operate wholesale and colocation data centers across China for cloud providers, internet firms and banks, are the clearest case of a business outrunning its stock. GDS revenue growth reaccelerated to 23.6% year over year in the first quarter after three quarters of near-flat growth, with area utilization rising to 75.5%; VNET revenue has grown 18%-24% every quarter for more than a year. Both trade at 1x-3x sales versus roughly 10.5x for U.S. data-center REITs, a gap tied to audit-compliance and delisting overhang on their China listings rather than operating weakness. GDS also pushed a $550 million convertible note's maturity from 2029 to 2032, reducing near-term refinancing risk, and both stocks rallied roughly 8% in early July on reports China is weighing a five-year, roughly $295 billion data-center investment program — a sector catalyst, not a company-specific one.

The three smaller names are far less alike than their grouping suggests. KEEL Infrastructure, the renamed and U.S.-redomiciled parent of former crypto miner Bitfarms, is pivoting toward a 2.2-gigawatt power pipeline across Pennsylvania, Washington and Quebec, but first-quarter revenue fell 44.7% year over year with a $145 million net loss, and consensus expects a further 59% revenue decline in fiscal 2026 even as its forward price-to-sales multiple rises to 22.5x from 12.7x trailing — a shrinking business getting more expensive, not less. SharonAI Holdings (SHAZ), a GPU-cloud company that went public via SPAC merger in December, generated just $294,000 of revenue in the first quarter against a $513 million market value; its forward multiple only looks reasonable if a disputed $1.25 billion contract materializes, and an April short-seller report alleges the counterparty earned just $5.8 million in its own last fiscal year while separately flagging litigation accusing the CEO of self-dealing at his prior company. WhiteFiber (WYFI), spun off from Bit Digital last August, is the exception: revenue grew 66% to $79.2 million in fiscal 2025 at a 62% gross margin, with consensus projecting a 45% EBITDA margin in fiscal 2026 — a credible path to profit, though only a year of trading history leaves no multi-year range to judge its 7.9x forward sales multiple against.

Financing conditions across this industry are genuinely tightening — short-term, GPU-collateralized debt now carries interest above 12% industry-wide, with roughly $7.5 billion maturing within an 18-month window in 2026, prompting Nvidia to introduce backstop financing. That risk lands unevenly: Applied Digital's dilution cushion is gone at current prices while GDS just extended its own maturity wall by three years.

On the tape, the pullback isn't uniform: five of six names are down over the past 30 days, but GDS's price is actually up 8.2% even as its trend indicator has read bearish continuously since late June — a case where the momentum signal and the fundamentals disagree, and the fundamentals here (accelerating revenue, an extended debt maturity) are the more persuasive read.

The setup

Where it stands — The group is down for 30 days after a divergent 12-month run; GDS and VNET show accelerating revenue at low multiples while APLD, KEEL and SHAZ diverge on credibility and dilution risk. Would confirm — GDS/VNET quarterly revenue growth stays above 20% for two more quarters while their price-to-sales multiples remain below 3x versus ~10.5x U.S. peers. Would invalidate — SharonAI's disputed $1.25 billion contract fails to show up as billed, disclosed revenue by year-end, or Applied Digital's shares trigger convert dilution above $14.72. Watch next — Next quarterly reports for GDS, VNET and Applied Digital, due within roughly eight to ten weeks, for updated utilization and lease disclosures. Valuation — GDS trades at 13.75x trailing EV/EBITDA and ~2.7x sales versus ~10.5x for U.S. REIT peers; Applied Digital's trailing P/S has fallen to 12.8x from 28.7x three months ago.

A basket of five industrial and clean-energy stocks tied to the physical guts of AI data centers and battery microgrids -- conduit, copper tubing, boilers, storage software and solar inverters -- has returned 141% over the past year. That number suggests a coherent buildout story. It isn't one. Pulling the average apart shows one company's near-collapse restructuring accounts for almost the entire gain, while the other four are moving for reasons that have nothing to do with each other: a buyout offer, a copper-price whipsaw, and a real slowdown in home solar demand.

The five businesses. Babcock & Wilcox (BW) builds industrial boilers and gas-fired power equipment and is emerging from a near-collapse restructuring. Atkore (ATKR) makes electrical conduit and cable management used on construction sites, utility-scale solar farms and data centers. Mueller Industries (MLI) makes copper tube, fittings and valves used in plumbing and cooling systems. Enphase Energy (ENPH) makes solar micro-inverters and home batteries sold mainly to homeowners, not data-center operators. Stem (STEM) sells software that manages and dispatches batteries for commercial and grid customers.

One stock explains the average. Adjusted for Mueller's July 2026 stock split, the twelve-month returns are roughly: BW +920%, Stem -54%, Mueller +60%, Enphase +25%, Atkore +23%. Babcock & Wilcox alone contributes nearly five times more return than the other four names combined. Its backlog jumped 483% year over year to $2.7 billion in the first quarter, anchored by a $2.4 billion contract to build four natural-gas power plants for data-center operator Applied Digital — but that contract contributed only $31 million to actual first-quarter revenue, and the company still posted a $79.6 million loss, driven mostly by a non-cash markup on customer warrants. Babcock & Wilcox also diluted shareholders more than 30% overnight in a $200 million equity raise priced May 15, and now faces multiple securities class-action suits alleging it misrepresented the Applied Digital contract and that its largest shareholder had undisclosed ties to the counterparty. Its trailing valuation multiples are negative (equity is still underwater); its forward P/E of 74x prices in a turnaround that hasn't yet shown up in cash flow.

Atkore is now merger arbitrage, not a buildout bet. Prysmian agreed to buy Atkore for $95.00 per share in cash, announced August 2-3 — a 30% premium to its prior close. With the stock at $93.55, upside is now capped by deal terms rather than by conduit demand. That's a shame, because Atkore's own fundamentals were genuinely improving: fiscal third-quarter adjusted EPS rose to $1.92 from $1.63 a year earlier after a rough fiscal 2025, in which revenue fell 11% and net income swung to a loss. The company will not update guidance while the deal is pending, so the conduit/data-center volume question this basket was built to answer is now moot for Atkore as a standalone trade.

Mueller is the real divergence. Second-quarter revenue rose 25.5% year over year to $1.428 billion, yet the stock (split-adjusted) fell from roughly $139 in late June to a low of $61 in mid-July, flipping its trend signal from a strong uptrend to a strong downtrend in about three weeks. Gross margin compressed to 27.7% from 31.0%, consistent with copper-spread compression on the COMEX exchange rather than a volume problem. Mueller trades at 17.0x trailing and 16.0x forward earnings and 11.2x trailing EV/EBITDA — the most conventionally reasonable multiple in the group for a business still growing revenue at a double-digit clip, which makes the price collapse the clearest case in this basket of the tape moving against, not with, the business.

Enphase's downgrade tracks real demand, not rotation. Its trend signal was cut from a strong uptrend to a mild downtrend on July 8, then cut again to a fuller downtrend on July 30. That matches the fundamentals: second-quarter revenue fell 19.6% year over year, and consensus estimates call for a further 19.3% revenue decline in fiscal 2026 following the U.S. residential solar tax-credit step-down. Channel inventory is described as normal for batteries but still elevated for microinverters. Enphase's forward P/E of 19x looks cheap next to its trailing 39x, but that's because trailing earnings collapsed, not because forward earnings are accelerating.

Stem is a drag, not ballast. Annual recurring revenue was effectively flat at $61.2 million in the first quarter, cash fell $12.3 million in the quarter to $36.6 million, and reported profitability in fiscal 2025 depended on a one-time non-operating gain that masked a $55.7 million operating loss. At a sub-$100 million market capitalization, Stem is small enough that its near-55% decline barely dents a simple five-name average even as it signals real distress.

Verdict. There is no single cohort answer. Mueller's business CONFIRMS a fundamentals-tape divergence worth watching; Babcock & Wilcox's backlog growth is real but swamped by dilution and litigation risk, an INCONCLUSIVE case at best; Enphase's decline CONFIRMS genuine demand softening, not sector noise; Atkore's story is now CONTRADICTED as an ongoing thesis by the pending buyout; Stem CONTRADICTS the idea that this is an investable group by remaining too small and too fragile to matter. None of the five appeared among the market's most extreme 30-day movers, underscoring that the group's -1.1% 'flat' 30-day average is arithmetic cancellation between Atkore's deal-driven surge and Mueller's, Enphase's and Babcock & Wilcox's declines — not a pause.

The setup

Where it stands — A five-name group's 141% year is mostly one distressed stock's restructuring, with Atkore now capped by a pending cash buyout. Would confirm — Mueller's gross margin recovers toward 30%+ while revenue growth holds above 15% for two more quarters. Would invalidate — Babcock & Wilcox's Applied Digital contract converts to backlog-to-revenue below 10% for another two quarters. Watch next — Atkore-Prysmian deal closing timeline and any regulatory review outcome, expected later in 2026. Valuation — Mueller trades at 16x forward earnings and 11.2x EV/EBITDA versus Babcock & Wilcox's 74x forward P/E on negative trailing earnings.

A group of seven small and mid-cap chipmakers — makers of power semiconductors for electric vehicles and data centers, industrial lasers, and the substrate materials that go into optical networking — has been treated by the market for a year as a single AI-infrastructure trade. Their combined 12-month gain is enormous. But the earnings reports landing this quarter show these are not seven versions of the same business, and the past 30 days have started to price that difference: three names rose, four fell, with declines as steep as 38% for one member even as the stock market elsewhere stayed calm.

The real growth story: AXT, nLIGHT, Vishay. AXT, which makes indium-phosphide and gallium-arsenide substrate wafers used in optical transceivers, reported record second-quarter revenue of $47.6 million, up 165% from a year earlier, with non-GAAP gross margin scaling from 8% to 45% in a year and a backlog above $100 million backed by signed supply agreements with Coherent and Casella. That is a genuine operating turn, not a promotional headline — but the stock is up roughly 35-fold over 12 months and trades at 16 times forward sales and 80 times forward earnings, and it was cut from a bullish to a bearish technical rating in the last week of July even as the earnings came in strong, consistent with profit-taking after a parabolic run rather than a fundamental break.

nLIGHT, which builds fiber and semiconductor lasers for industrial cutting and, increasingly, for military directed-energy weapons, posted the cohort's cleanest combination of growth and profitability improvement: revenue up 55% year-over-year with gross margin expanding from 2.4% to 33.1% in five quarters. In July it was awarded a Department of Defense laser-weapon contract with a $627 million ceiling and $162 million already funded — a demand driver with nothing to do with AI data centers. The stock still trades at 13.6 times trailing sales.

Vishay Intertechnology, a broad-line maker of power transistors, diodes and passive components sold mostly into industrial and automotive customers, reported a 1.34 book-to-bill ratio and backlog up 21% to $1.6 billion in its most recent quarter, with revenue growing 17% and margins recovering — real evidence of an industrial-electronics restock rather than an AI-only narrative. Yet Vishay's stock fell 23% over the past 30 days heading into its August 5 earnings report, a clear divergence between what the order book says and what the tape has done.

The narrative story: Wolfspeed and POET. Wolfspeed, which emerged from Chapter 11 bankruptcy in September 2025 with debt cut from $6.7 billion to roughly $2 billion, makes the silicon-carbide and gallium-nitride power chips used in EV inverters and, in the bull case, in next-generation 800-volt data-center power systems. Its most recent quarterly revenue fell 19% year-over-year with negative operating margin, and Chinese suppliers have pushed 6-inch silicon-carbide wafer prices down to roughly $400-500, less than a third of prices two years ago. Wolfspeed's spike to nearly $73 in May was, per prior desk research tied to an AI-power narrative note, largely a short squeeze rather than a demand signal, and the stock has since fallen 38% in 30 days heading into Tuesday's earnings report — the real test of whether any operating turn exists behind the post-bankruptcy story.

POET Technologies, developing an unproven "optical interposer" chip-packaging technology for data-center optics, generated just $503,389 in quarterly revenue against a market capitalization near $1 billion. Its stock fell after Passive Foreign Investment Company class-action lawsuits were filed and reports surfaced that a confidentiality breach by its CFO led a customer, Celestial AI, to cancel purchase orders — a specific, negative event, not generic valuation reversion.

In between: Photronics, which makes the photomasks used to pattern chips and displays for foundry customers, posted flat revenue but is investing $330 million to expand advanced-node mask capacity for AI-related demand; its trailing-to-forward P/E widened from 11.5x to 16.9x even as its technical rating flipped bearish in mid-July while the price itself rose — a signal-versus-tape mismatch worth watching. IPG Photonics, a fiber-laser maker, grew revenue 17% but posted a small operating loss and trades at 33 times EV/EBITDA, rich for its thin margins.

The setup

Where it stands — The cohort's 12-month gain reflects three genuinely growing businesses (AXT, nLIGHT, Vishay) mixed with a bankruptcy-recovery story (Wolfspeed) and a pre-revenue speculation (POET). Would confirm — Wolfspeed's August 4 and Vishay's August 5 earnings show sequential revenue and margin improvement, not just flat or declining guidance. Would invalidate — Wolfspeed's FQ4 revenue guidance falls further below the ~$758 million FY2026 consensus already baked in. Watch next — Wolfspeed reports fiscal Q4 earnings August 4, 2026; Vishay reports Q2 earnings August 5, 2026. Valuation — AXT trades at 16x forward sales versus roughly 1x a year ago; Vishay trades at 46x forward earnings versus a not-meaningful trailing multiple, reflecting a low profit base.

Four companies that supply the water handling, sand and wellhead equipment shale drillers need to complete a well have been treated as a single quietly-compounding group over the past year. They are not. Pulling apart each company's actual results shows one name whose rally is backed by a genuine shift in its business model, and three others whose gains — or losses — tell a very different story once the fundamentals are checked against the tape.

Select Water Solutions (WTTR), which manages, recycles and disposes of the wastewater that comes up alongside oil and gas production, is the standout. Its stock is up 106.6% over the past year but only 3.45% over the past 30 days — a slow, unbroken grind rather than a spike. The business explains it: Water Infrastructure segment revenue grew 33% year-over-year in the first quarter of 2026 at a record 56% gross margin, and the company signed a wave of long-term contracts this year — three minimum-volume-commitment deals, two acreage dedications and two rights-of-first-refusal agreements — while recycled and disposed volumes hit a record 1.4 million barrels a day. Total revenue was roughly flat (-2.25%) but operating income rose 54%, and the company raised full-year growth guidance to 25-30%. Trailing EV/EBITDA of 9.95x and a forward price-to-earnings ratio of 34.9x (down from a distorted 92.6x trailing, as forward EPS is projected to more than double) look reasonable against that growth. Verdict: CONFIRMS — the shift from spot water hauling to contracted infrastructure is real; valuation is a supported advance, not a stretch.

Atlas Energy Solutions (AESI), which mines and delivers the sand used to prop open fractures during well completions, moved in the opposite direction: down 12.8% over the year and down a sharp 24.6% in just the past 30 days, its price band rolling from a strong uptrend into a downtrend between March and July. The fundamentals agree: first-quarter 2026 revenue fell 10.8% year-over-year, gross margin turned negative for the first time (-0.64%), and the company posted a $47.3 million net loss as sand prices slid toward roughly $18 a ton in an oversupplied Permian market. Its electric sand-conveyor system, the Dune Express, is hitting record utilization and the company touts a longer-term power-supply pivot, but free cash flow yield is a thin 1.37% and EV/EBITDA sits at 14.2x despite the earnings decline. Verdict: CONFIRMS — both the business and the stock are deteriorating together, the opposite of the "quiet compounder" framing.

Cactus (WHD), which makes and rents the wellhead and pressure-control equipment used to cap and control wells, shows a violent 26.5% move in just the past 30 days atop a 60.7% one-year gain — hardly the gradual pattern the cohort was supposed to share. Second-quarter revenue jumped 64% year-over-year to $449.5 million, beating consensus, but most of that growth came from a January 2026 joint venture that folded in Baker Hughes's Surface Pressure Control business, roughly 85% of which is generated in the Middle East — not US completion activity. Net margin compressed to about 6% from 16% a year earlier even as revenue surged. Trailing price-to-earnings of 53.6x (22.0x forward) is rich for a business whose organic US segment faces the same headwinds as its peers. Verdict: CONTRADICTS — the headline growth is real but mostly acquired, masking margin dilution underneath.

Smart Sand (SND), a smaller frac-sand miner, posted the biggest one-year gain in the group at 110.7% but fell 8.1% in the past 30 days, and its trend band has round-tripped between strong and mild uptrends rather than holding a steady line. A single outsized quarter (24.9% net margin in mid-2025) distorts its trailing price-to-earnings to 7.7x; full-year 2025 net income was just $1.35 million on $330 million of revenue, and consensus now expects FY2026 revenue to fall roughly 25%, pushing the forward P/E to 43.6x. Verdict: INCONCLUSIVE, leaning CONTRADICTS — the rally has outrun a shrinking earnings base.

The shared backdrop explains the split. Haynesville-region gas rigs have climbed to a multi-year high near 60, but the top five producers there are banking 149 drilled-but-uncompleted wells rather than finishing them, and Appalachian rig counts are down 29% even as production rises 10% on efficiency gains. Completion activity — the thing that drives sand and pressure-control demand — is lagging headline drilling. Select Water's contracted, fee-based infrastructure model is comparatively insulated from that lag; the sand and completion-linked names are not.

The setup

Where it stands — WTTR's contracted water-infrastructure model is expanding profitably while AESI, WHD's organic base, and SND face a completion-activity slowdown. Would confirm — WTTR's Water Infrastructure segment revenue growth stays above 25% and gross margin holds near 56% next quarter. Would invalidate — WTTR's forward EPS growth guidance is cut below the ~118% currently projected for FY2026. Watch next — AESI's second-quarter 2026 results, guided to roughly $50 million EBITDA, due in the coming weeks. Valuation — WTTR trades at 9.95x trailing EV/EBITDA and 34.9x forward P/E versus 92.6x trailing, cheap relative to its own growth reset.

Nine software companies whose products range from text-messaging pipes to hospital-grade clinical databases have rallied together since late May 2026, unwinding a months-long selloff that was built on a single fear: that AI would let big companies build their own tools instead of paying for licensed software. Investors concluded enterprise IT budgets were reallocating toward AI-capable platforms rather than shrinking outright, and software stocks caught what one market write-up called an indiscriminate bid starting the weekend of May 29. Seven of the nine names now sit in an uptrend by the market's trend-following measure, and their businesses tell three different stories at once — real acceleration, real deceleration, and one stock priced as if nothing had changed at all.

The nine and what they sell. Twilio lets businesses send texts, calls and AI-driven voice interactions to customers through a pay-per-use platform. Bill Holdings (BILL) automates bill payment and invoicing for small and midsize businesses. Dynatrace (DT) sells software that monitors the health of a company's cloud applications. Samsara (IOT) sells sensors and software that track vehicle fleets and industrial equipment. JFrog (FROG) sells the software pipeline companies use to build, store and secure their code. Nutanix (NTNX) sells private-cloud infrastructure software, an alternative to VMware. Veeva Systems (VEEV) sells cloud software built specifically for pharmaceutical companies' clinical and regulatory work. Atlassian (TEAM) makes Jira and Confluence, workplace tools historically priced per employee seat. Salesforce (CRM) is the largest customer-relationship-management software vendor, also priced mostly per seat.

A broad move, not one stock. Five names — Twilio, JFrog, Nutanix, Dynatrace and Samsara — are in the strongest uptrend classification simultaneously, and two more (Bill Holdings, Veeva) flipped to a milder uptrend in the past two weeks. The average one-month gain across all nine is roughly 14%, with Atlassian (+30%) and Bill Holdings (+29%) leading and JFrog (-10% over the past month) giving back part of an earlier 65% three-month run. Over the trailing year the group is split down the middle: Twilio (+50%) and JFrog (+89%) are sharply higher while Atlassian (-49%), Salesforce (-30%) and Veeva (-29%) remain well below where they traded a year ago, even after the recent bounce.

Where the businesses actually confirm the move. Twilio's organic revenue growth accelerated to 16% last quarter, its fastest since 2022, with its dollar-based expansion rate rising to 114% from 107% and operating margin at a record roughly 20% — a result strong enough that the stock's uptrend, now running 106 consecutive sessions since April 17, still trades at roughly 4.7 times forward revenue, below its own 10-year median near 7.2 times. Dynatrace's annual recurring revenue grew 16-18% and next year's guidance points to further acceleration. JFrog's revenue grew 26% with cloud revenue up 50%, though its stock's price-to-sales multiple nearly doubled to over 16 times in three months before pulling back — a valuation the desk had already flagged as needing to settle. Atlassian beat and raised guidance, with cloud revenue growth guided to roughly 26.5% and large deals nearly doubling, a result that cuts against fears of a shrinking per-seat business; but its trend-band upgrade from a downtrend to a mild uptrend happened just three sessions before this snapshot, with no separate news behind it — consistent with a stock that simply stopped falling.

Where price and fundamentals disagree. Samsara's revenue grew 26-28% last year, but next year's guidance calls for growth to slow to 21-22% — yet the stock trades at 12.6 times sales in its strongest uptrend band, leaving little room for further slippage. Nutanix's headline annual-recurring-revenue growth of 15% outran actual reported revenue growth of just 10% last quarter on supply-chain delays, while its price-to-sales multiple expanded roughly 42% to 5.8 times over the same stretch — the stock re-rated faster than the print did.

The clearest divergence: Salesforce. Salesforce trades at roughly 20 times trailing earnings, cheaper than its own spring-2026 level near 23-25 times, with revenue up 11-13% and non-GAAP earnings per share up roughly 50%. Its Agentforce AI product has grown its combined booked revenue to about $3.4 billion, and the company says it closed 8,000 Agentforce deals last quarter — yet the stock is still down roughly 26-30% over the trailing year. Despite a 19% bounce in the past month, Salesforce's trend classification has not moved out of its downtrend band once since April, the only member of the group where that is true. That gap — cheap multiple, real earnings growth, no technical confirmation — is the most concrete unresolved question the group presents.

Fundamentals verdict. For Twilio, Dynatrace and JFrog, the rally CONFIRMS a real re-acceleration in the underlying business, though JFrog's valuation had briefly outrun it. For Samsara and Nutanix, rising prices CONTRADICT guidance that points toward slower growth ahead. For Atlassian and Salesforce, the picture is INCONCLUSIVE: Atlassian's guidance beat is real but its band flip is too fresh to separate from a simple stop-falling bounce, and Salesforce's cheap multiple and earnings growth have not yet been rewarded by the tape at all.

The setup

Where it stands — Seven of nine names are in a confirmed uptrend after a sector-wide reversal that began in late May 2026; Salesforce alone remains in a downtrend despite a 19% monthly gain. Would confirm — Salesforce's trend classification finally flips out of its downtrend band, or Samsara/Nutanix revenue growth reaccelerates toward their ARR growth rates next quarter. Would invalidate — Atlassian's cloud growth guidance is cut in its next report with no fresh catalyst behind the recent band upgrade. Watch next — Nutanix and Samsara's next quarterly reports, due within the coming reporting cycle, for whether reported revenue closes the gap with ARR and billings growth. Valuation — Twilio trades near 4.7x forward revenue versus its own 7.2x ten-year median; Salesforce trades near 20x trailing earnings versus its own 23-25x spring-2026 level.

What happened

Seven chipmakers that supply the radio-frequency, optical and precision-timing components used to move data inside artificial-intelligence data centers and wireless networks fell together in July, each losing between 17% and 40% of its value in a single month. That looks alarming next to a trailing 12-month average gain of 146% across the group. But the underlying businesses did not deteriorate in the same window for most of these companies — orders, backlog and guidance largely held up. The one clear exception is Qualcomm, the mobile chipmaker trying to diversify away from smartphone modems and processors into automotive and industrial chips, which took a real, dated hit when Apple's in-house modem arrived faster than expected.

A shared sell-off, not a shared slowdown

The trigger was not any single company but a broader debate over whether cloud giants are overspending on artificial-intelligence infrastructure. Applied Optoelectronics, which makes optical transceivers that plug fiber-optic cables into data-center switches, fell alongside Coherent and Lumentum, two other optical-component makers, on four separate dates in July, with coverage citing "AI spending doubts" rather than any company-specific news. The flashpoint was Alphabet's quarterly report, which showed capital spending roughly doubling to $44.9 billion while free cash flow swung negative, stoking doubts about the payback on AI hardware spending across the whole supply chain. Applied Optoelectronics had no negative company news in the window and remains up about 150% year-to-date; its next earnings report is due August 6, after this drop.

MACOM, which sells RF, analog and optical chips into data-center, telecom and defense markets, gave standing guidance in May for roughly 35% sequential growth in its data-center segment and 59%-60% gross margin, a forecast that has not been revised despite the stock's decline. Its most recent quarter carried a 1.3x book-to-bill ratio and record backlog, with roughly half its revenue from data centers and about a quarter from telecom and satellite links, which grew 81% year over year — a broader mix than a pure interconnect play. SiTime, which makes MEMS-based precision-timing chips (micro-electromechanical systems) used to synchronize data-center and networking equipment, guided for at least 80% full-year revenue growth after an eighth straight quarter of triple-digit growth in its data-center business; its next results land August 5, meaning July's decline came before, not after, any fresh number. Semtech, which sells analog chips including LoRa wireless modules and high-speed data-center cable interconnects, posted record quarterly sales with data-center revenue up 92% year over year, its sixth straight quarter of growth. For these three, the tape moved and the business didn't — a de-rating, not a deterioration.

The one real casualty: Qualcomm's Apple problem

Qualcomm's July 29 earnings did contain genuine bad news: management said Apple's in-house C2 modem is arriving faster than planned, cutting Qualcomm's Apple-related revenue roughly in half sequentially into the next quarter, even as automotive revenue hit a record $1.59 billion (up 61%) and internet-of-things revenue (IoT, chips for connected devices outside phones and cars) rose 9%. Qualcomm also lifted its long-term non-handset revenue target to $40 billion by fiscal 2029 from $22 billion previously. Because Qualcomm's trailing 12-month return was roughly flat (+2.3%) before this drop, it did not drive the cohort's 146% average gain — that came almost entirely from Applied Optoelectronics (+318%), MaxLinear (+340%) and SiTime (+173%). Qualcomm now trades near the bottom of the group on any valuation measure, at roughly 18.7x trailing earnings and 4.1x trailing sales, cheaper than every peer here.

indie Semiconductor's lone upgrade

indie Semiconductor, which designs radar and driver-assistance chips for automakers, was the only name in the group upgraded rather than downgraded on its medium-term trend reading even as its longer-term reading was also cut. That divergence is backed by real revenue: the company holds a $7.4 billion strategic backlog and landed a dated $25 million production order for its eighth-generation radar chipset tied to two automaker programs. First-quarter revenue was $55.5 million with a 38% gross margin, and the stock trades at roughly 3.7x trailing sales — the cheapest multiple in the group by a wide margin. MaxLinear, which makes RF and optical-interconnect chips for broadband and data-center networks, showed new storage and optical-signal products at a late-July trade show, but no second-quarter revenue or cash-burn figures have surfaced yet to confirm whether its nearly-quadrupled six-month stock move is backed by recognized sales; its trailing gross margin, estimated from reported figures, runs near 57%.

Valuation: the give-back hasn't reached the multiple

Despite the month-long drop, valuations for most of the group remain rich by their own recent history. MACOM's trailing price-to-sales ratio has actually risen this year, from about 19.5x in early May to roughly 26x now, even as the stock fell — meaning earnings estimates, not the price decline, moved the multiple. SiTime's trailing price-to-sales, near 47x, is the richest in the cohort with no visible compression. Applied Optoelectronics is the exception: its trailing price-to-sales has fallen from about 31x to roughly 22x, a genuine multiple reset alongside the price. Set against this, indie Semiconductor's sub-4x sales multiple and growing backlog stand out as the one name where valuation and business trend both point the same direction.

The setup

Where it stands — Six of seven names show intact backlog and guidance behind a shared July price drop; Qualcomm has a confirmed, company-specific Apple revenue hit. Would confirm — MACOM, SiTime and Semtech data-center revenue keeps growing double digits sequentially in results due through September 2026. Would invalidate — Any of the group cuts guidance, flags distributor inventory build, or reports book-to-bill below 1.0x in its next report. Watch next — SiTime reports August 5 and Applied Optoelectronics August 6, 2026 — both after this drawdown, testing whether it was de-grossing or deterioration. Valuation — MACOM trades near 26x trailing sales versus roughly 19.5x in May; indie Semiconductor trades near 3.7x, the cohort's cheapest, against its own backlog growth.

Natural gas demand from artificial-intelligence data centers and new gas-fired power plants is showing up in company results — but the money attached to that demand is flowing almost entirely to the pipes and compressors that move the gas, not to the drillers who produce it. Over the past year, gas-pipeline operators and compression-equipment lessors have rallied 20%-55%, while six Appalachian gas producers are essentially flat, even though several just posted their best quarters in years and raised production targets. That split is the story: a documented gap between what these businesses are doing and what their stocks are doing.

Midstream: contracts are signed, and it's showing up in the numbers. Kinder Morgan (KMI), which owns the largest natural-gas pipeline network in North America, ended its second quarter with a $9.6 billion project backlog, 92% of it gas-related and more than 60% tied to power-generation demand, and now expects full-year adjusted earnings before interest, taxes, depreciation and amortization (EBITDA) at least 5% above budget — a swing chief financial officer David Michels pegged at more than $430 million, driven by the natural-gas buildout. Its Trident pipeline, a $1.7-1.8 billion project moving gas from Katy, Texas to Port Arthur, has its first phase targeted for early 2027 in-service. Williams Companies (WMB), owner of the Transco pipeline serving the Southeast and mid-Atlantic, is committing over $5 billion to power-linked projects including the $1.6 billion Project Socrates, targeted for late-2026 completion, and is separately in advanced talks on a roughly $5.5 billion acquisition of gathering assets serving ten liquefied-natural-gas (LNG) facilities and 26 power plants. Energy Transfer (ET), a diversified pipeline operator, signed a 450 million-cubic-feet-per-day supply contract with CloudBurst Data Centers near Austin and will feed a 1.2-gigawatt power plant near San Marcos starting the third quarter of 2026. Oneok (OKE), Enterprise Products Partners (EPD), Enbridge (ENB), TC Energy (TRP) and MPLX round out the group; all eight names sit in bullish trend bands, and the cohort's still-modest 12-13 times EBITDA multiple keeps it below merchant power-generation comparables even after the run.

Compression: real utilization, but the tape has started giving some back. Archrock (AROC), which rents natural-gas compression units to producers and pipelines, ran its fleet at 95% utilization on 4.53 million operating horsepower in the first quarter, is funding $250-275 million of 2026 growth capital expenditure without issuing new equity, and keeps debt under 3.5 times EBITDA, according to its own earnings materials. USA Compression Partners (USAC), a similarly sized fleet operator sponsored by Energy Transfer, is funding its capital plan from free cash flow, but its first-quarter gross profit fell 22.9% even as revenue grew 35% — a margin-compression signal inside a growth story — and the stock has slipped to a mild-bear trend band, the one name in this layer moving against its peers. Kodiak Gas Services (KGS) has climbed roughly 78% year-to-date on real, largely contracted distributed-power capacity, but sell-side coverage has downgraded it to Hold on valuation, and Archrock's own price target was trimmed in late July — a sign the group's easiest gains may be behind it even where the underlying contracts are genuine. Natural Gas Services Group (NGS) is the smallest name in the set.

Producers: better results, cheaper stocks, worse trend. All six Appalachian gas producers — EQT (EQT), the region's largest gas producer; Expand Energy (EXE), the Haynesville- and Appalachia-focused driller formed from Chesapeake's merger with Southwestern; Antero Resources (AR), an unhedged gas and natural-gas-liquids producer; Range Resources (RRC), a Marcellus operator with liquids-export exposure; CNX Resources (CNX); and Gulfport Energy (GPOR) — sit in the same bearish trend band, and all six are down over the trailing 90 days despite modest 30-day bounces. Yet EQT beat first-quarter consensus, posted a $5.07-per-thousand-cubic-feet realized price after hedges, and raised full-year production guidance; Range delivered its highest gas premium to benchmark futures in over a decade and a record natural-gas-liquids premium, with a favorable 2026 basis guide, not a blowout, per its results. The selling instead tracks a soft government forecast — the Energy Information Administration expects Henry Hub gas near $3.5-3.7 per million British thermal units through 2027, versus a $4.15 estimate from Goldman Sachs. That flat strip has priced these producers at 4.5-to-12.3 times trailing earnings, roughly a third of the compression group's multiple and a fraction of midstream's, even as EQT and Antero run largely unhedged into any late-decade repricing. Expand Energy has also been without a permanent chief executive since February, adding company-specific uncertainty on top of the sector-wide markdown.

What the tape says, and what it's missing. The trend bands read cleanly: midstream bullish across all eight names, compression bullish in three of four, and every one of the six producers in the same bearish band since mid-April — a cohort-wide signal, not one company's problem. But bearish price action next to earnings beats and raised guidance is a genuine divergence between the tape and the business, and it's the producers, not the pipelines, carrying it.

The setup

Where it stands — Midstream and compression have re-rated on signed, dated gas contracts; Appalachian producers beat estimates but trade at 4.5x-12.3x earnings versus 12x-100x for pipelines and compressors. Would confirm — Henry Hub futures for 2027-28 rise toward Goldman's $4.15/MMBtu case while EQT, Range and Antero hold or raise 2026 production guidance again. Would invalidate — Producer forward EV/EBITDA rises to converge with midstream multiples without any change in the futures strip, or USAC-style margin compression spreads to Archrock and Kodiak. Watch next — Expand Energy's permanent CEO appointment, expected within its original six-to-nine-month search window from February 2026. Valuation — E&P trailing P/E averages roughly 8x versus midstream's 12-34x and compression's 23-100x; midstream's 12-13x EV/EBITDA sits below merchant-power comparables.

A landlord layer that hasn't kept pace with its tenants

Data centers are the buildings that house the servers running artificial-intelligence workloads, and the companies that own and lease that real estate are structured as real estate investment trusts (REITs), which must pay out most of their taxable income as dividends. Over the past year, companies that make the memory chips, packaging and networking gear inside those buildings have rallied more than 100%, in some cases nearly 280%, on AI-driven demand. The eight REITs and REIT-like landlords that own the buildings themselves have moved a combined average of roughly flat to slightly negative over the same period. That gap raises an obvious question: are the landlords a lagging piece of the AI trade that simply hasn't caught up, or is data-center real estate structurally unable to capture AI economics the way chipmakers do? The answer splits the group in two.

The US pair is executing — and priced for it

Equinix (EQIX), the largest global operator of interconnected data centers where cloud providers and networks exchange traffic, rose 29.5% over the past year and reported its second-highest quarterly bookings on record plus a record backlog of signed-but-not-yet-installed capacity, prompting what the company called its largest single guidance raise in its history — 2026 funds-from-operations-per-share growth (AFFO, the REIT-industry profit measure that adjusts for real estate depreciation) now guided to 10-12%. Digital Realty (DLR), the largest wholesale data-center landlord serving hyperscale cloud tenants, is up 7.9% over the year and posted renewal spreads above 25% in cash terms and an all-time-high $1.9 billion leasing backlog, meaning existing customers are paying markedly more to renew — a reversal after a decade of pricing pressure. Verdict on the business: CONFIRMS acceleration. Verdict on valuation: STRETCHED. Equinix trades near 74-78x trailing GAAP earnings and roughly 25-30x forward AFFO, the highest multiple in the sector; Digital Realty sits around 51-52x trailing earnings and 22-25x forward AFFO. Both are executing, but further re-rating now depends on continued delivery rather than a valuation gap being closed.

NextDC, GDS and VNET look like the real dislocation

NextDC (NXT.AX), an Australian data-center and interconnection operator expanding into Malaysia, New Zealand and Singapore, fell 5.5% over the year even as contracted utilization jumped 60% quarter-over-quarter to 667 megawatts and its forward order book rose 83%; it is funding growth with a fresh A$2.2 billion capital raise, and its gross margin briefly collapsed to 0.4% last quarter, a reminder that scaling capacity ahead of revenue has near-term costs. GDS Holdings (GDS) and VNET Group (VNET), Chinese data-center operators serving domestic cloud and enterprise customers, fell 7.6% and 11.4% respectively despite GDS posting record quarterly bookings of roughly 200 megawatts and VNET guiding wholesale-segment earnings to grow 56% through 2026. Both trade at 0.2-0.5x sales versus roughly 11x for Equinix and Digital Realty — a discount the user's own research notes attribute mainly to US-listing audit-compliance risk and refinancing exposure rather than demand weakness. Verdict on the business: CONTRADICTS the flat tape — fundamentals are accelerating while prices fell. Verdict on valuation: this is the group's genuine possible dislocation, tempered by real geopolitical and funding risk.

Towers and paper storage are dragging, and diverging, not confirming

American Tower (AMT), the largest global wireless-tower landlord that also owns the CoreSite data-center business, fell 18.2% over the year even as its data-center segment grew 13.4% and management raised guidance for the second time — this is a tower-lease-churn recovery story wearing a data-center label. Crown Castle (CCI), a US tower REIT, fell 28.2% after an $8.5 billion sale of its fiber and small-cell businesses — an explicit retreat from the AI/edge-computing convergence this basket is meant to capture. Iron Mountain (IRM), a records-storage and document-management company expanding into data centers, rose 27.4% on 47% data-center segment growth but trades above 140x trailing earnings, a records-storage base carrying a growth-stock multiple. None of these three is primarily a data-center bet, and excluding them lifts the remaining five-name average from roughly flat to about +6.7% — still an order of magnitude below the 100%+ moves in chips, packaging and networking.

What's actually rationing capacity

Digital Realty's new developments are pre-leased at an 11.5% stabilized yield, comfortably above typical REIT cost of capital, so new supply remains profitable to build. The scarce input is no longer land or capital — it's power. US and European grid interconnection queues now average more than six years, and Sightline Climate estimates 30-50% of 2026's planned capacity could be delayed. Hyperscale cloud companies are increasingly funding their own power generation, a channel that bypasses landlords on that layer — but the largest AI-cloud renter, CoreWeave, still leases essentially all of its 49 data centers from third parties rather than self-building, meaning demand hasn't broadly left the REIT model yet.

On the tape, American Tower's trend flipped to mild bearish from strong bearish over the past month as DISH-related headwinds fade, Iron Mountain remains in a strong uptrend, and the broader eight-name group still carries a bullish trend label despite averaging near flat — a reminder that trend labels here reflect direction, not magnitude.

The setup

Where it stands — Equinix and Digital Realty are executing at record levels but priced near decade-high multiples; NextDC, GDS and VNET show accelerating fundamentals despite falling prices. Would confirm — Digital Realty's cash renewal spreads stay above 20% and NextDC's contracted utilization keeps rising in its next quarterly update. Would invalidate — Equinix's AFFO-per-share growth guidance is cut below 10% or GDS/VNET utilization rates decline sequentially. Watch next — Equinix and Digital Realty's Q3 2026 earnings, expected late October 2026, for continued bookings momentum. Valuation — Equinix trades near 25-30x forward AFFO and Digital Realty near 22-25x, both above their own decade averages for the sector.

Investors who piled into the most successful artificial-intelligence-adjacent stocks over the past year have spent the last month selling them, and the size of the sell-off tracks the size of the earlier rally almost name for name. Memory-chip maker Micron, radio-frequency component supplier MaxLinear, and glass-and-fiber manufacturer Corning — all among the biggest winners of the past twelve months — have each fallen 20% to 40% since late June. Meanwhile stocks that lagged badly over the same period, including IT-consulting firms and identity-security software vendors, have risen 6% to 27% in the last 30 days. That symmetry points to investors rotating out of crowded winners rather than a wave of bad news — except at two companies where guidance actually did deteriorate.

The names at the center of this are Micron (MU), which makes DRAM and NAND memory chips used in AI servers; MaxLinear (MXL), which makes radio-frequency and data-center connectivity chips; MACOM Technology (MTSI), a maker of high-speed optical and RF components for data centers; Corning (GLW), which supplies optical fiber and specialty glass to telecom and data-center customers; contract electronics assemblers Celestica (CLS), Jabil (JBL) and Flex (FLEX), which build servers and networking gear for cloud customers; Accenture (ACN), the largest IT-consulting and systems-integration firm; identity-security software vendors Okta (OKTA) and SailPoint (SAIL); cloud-communications software provider Twilio (TWLO); optical-networking equipment maker Applied Optoelectronics (AAOI); silicon-carbide chipmaker Wolfspeed (WOLF); cloud-communications platform Bandwidth (BAND); and IT-services firm EPAM Systems (EPAM).

At the hardware leaders, the business kept accelerating even as the stock fell. Micron's fiscal third-quarter revenue hit a record $41.5 billion, up 74% from the prior quarter, with gross margin expanding to nearly 85% and management saying DRAM and NAND supply will trail demand into 2027 on AI-driven demand. MACOM booked a record 1.3-times book-to-bill ratio with backlog still building. Celestica, Jabil and Flex all raised AI and data-center revenue guidance in their most recent quarters, and MaxLinear raised its 2026 optical data-center revenue outlook to $210-$230 million even after a one-off operating loss. None of that squares with a 20%-to-40% share-price decline on fundamentals alone — it looks like a crowded trade unwinding after chip stocks logged an 80% first-half rally before shedding more than $1 trillion in value in a single week in late July, alongside a broader wobble after the largest cloud providers raised 2026 capital-spending guidance to roughly $700 billion combined and investors reacted by questioning the payoff rather than the spending itself.

Two names break that pattern. Corning fell about 15% on July 28 after issuing a third-quarter sales forecast below Wall Street's estimate, tied to the pace of building new optical-fiber production lines for data centers — a real, if modest, guidance miss, not pure rotation. Accenture cut its full-year revenue growth guidance to 3%-4% from 3%-5% and reported bookings down 2%-3% year over year, citing weak US federal spending — a genuine deterioration that sits oddly next to Accenture's own peers, which merely lagged the market rather than falling on fresh bad news and have since bounced.

The bounce side is also not one story. Okta's revenue grew 12% last quarter, a continued deceleration from the mid-teens pace of prior years, even as free cash flow stayed strong at $211 million — a real business-maturation trend, not just sentiment. SailPoint, by contrast, grew revenue 20% and SaaS annual recurring revenue 38%, crossing $1 billion in total ARR, but is still posting losses. Twilio's organic revenue growth accelerated to 20%, its best pace in three years, with guidance raised twice in 2026.

On the tape, the pattern is directional but not perfectly linear: Micron's 30-day drawdown (-20%) is smaller than MaxLinear's (-40%) despite a far bigger prior-year gain, and Wolfspeed's -47% drawdown reflects a distressed, post-restructuring share count more than crowding. Database records show a uniform strong-bull-to-mild-bull downgrade stamped July 31 across Micron, MaxLinear, MACOM, Corning, Celestica, Jabil, Flex, Applied Optoelectronics and several smaller peers, while Accenture alone moved to a strong-bear reading — consistent with its being the one name in the group facing an actual guidance cut rather than a crowded-trade exit.

On valuation, the hardware leaders still carry cycle-sensitive multiples: Celestica, Jabil and Flex trade at trailing price-to-earnings ratios of roughly 40x-60x, and Micron's ~12x forward earnings looks inexpensive only if AI-driven memory pricing holds near current, historically elevated levels. Twilio's ~33x-35x forward earnings and Photronics' 11x trailing earnings sit at opposite ends of that spectrum, underscoring that the group's re-rating has room to run in either direction depending on which multiple an investor anchors to.

The setup

Where it stands — Chip and optical-hardware leaders fell 20%-47% in 30 days despite mostly intact guidance; Corning and Accenture fell on genuine guidance cuts. Would confirm — Micron, MACOM, Celestica, Jabil and Flex maintain or raise guidance again in their next quarterly reports without further share declines. Would invalidate — Book-to-bill or hyperscaler capex commentary turns down at MACOM or Micron's next print, confirming a fundamental, not technical, rollover. Watch next — Micron's fiscal Q4 FY26 report and Corning's Q3 2026 results, both due within the next 60-90 days. Valuation — Micron trades near 12x forward earnings versus a cycle-average that has ranged from single digits to over 20x across past memory cycles.

What happened

Bitcoin, the largest cryptocurrency by market value, and ether, the second-largest and the native token of the Ethereum network, both moved from a steep downtrend to a milder one on the same two time horizons this week — the trailing 30 days and the trailing 180 days — a genuine multi-month improvement after both coins fell roughly 44-46% over the past year. But the three companies whose stock prices are supposed to track crypto most closely — Strategy Inc. (formerly MicroStrategy), the software firm that converted its balance sheet into the world's largest corporate bitcoin hoard; BitMine Immersion Technologies, a Las Vegas company built around the largest corporate stash of ether; and Coinbase, the largest U.S. cryptocurrency exchange — all got worse on the three-month and one-year charts over that same stretch. The gap is not tape noise. Each company's own numbers explain why its stock is diverging from the coin it is supposed to track.

The coins: stabilizing, not confirmed

Over the past 90 days bitcoin fell about 19%, and both bitcoin and ether are down roughly 44-46% over the full year. That drawdown has left bitcoin trading above its "realized price" — the average cost basis of every coin ever moved on-chain — near $53,000-$55,000, but still below the short-term-holder cost basis of roughly $69,000-$72,000 and the longer-run "true market mean" of about $76,000-$79,000. That is consistent with a market that has stopped falling but has not yet drawn in fresh buyers. Confirming that, U.S. spot bitcoin ETFs took in just $205 million in July, the smallest monthly haul on record, after $2.43 billion and $4.52 billion left the funds in May and June. The Federal Reserve has also held rates at 3.5%-3.75% since late 2025 with fewer cuts priced for the rest of 2026 than bulls expected, a headwind that has pushed bitcoin's correlation with the S&P 500 up to 0.6-0.8 — a macro drag shared by coins and equities alike, which is part of why the equities have not simply mirrored the coins' improvement.

The equities: a capital-structure story, not a coin-price story

Strategy's stock fell 47% over the past 90 days and 75% over the past year — far worse than bitcoin's 19% and 44% declines over the same windows — and the reason sits on its balance sheet, not in the bitcoin market. The company's "mNAV," the ratio of its market value to the value of its bitcoin holdings, has compressed from a 2.6x-2.8x premium at the 2024 peak to roughly 1.08x-1.16x, briefly dipping below 1.0x in late June — meaning the market now values the company at or below the coins it holds. That premium collapse forced Strategy to sell 3,588 bitcoins in June, its first sale since 2022, to help fund the roughly $876 million annual dividend on its preferred stock, a cash obligation that exists regardless of where bitcoin trades. A separate risk sits ahead: a $1.01 billion convertible note lets holders demand repayment in September 2027 if the stock stays below the conversion price. One tail risk did clear, however — MSCI confirmed it will keep Strategy in its major indices, averting a forced sale by passive funds.

BitMine shows the same mechanism. Its own "realized mNAV" sat at 0.84 in early July — a 16% discount to the value of its ether holdings, not a premium, after shares outstanding grew 149% over the past year through continuous stock sales. The company's cost basis in its ether — $19.05 billion — now exceeds the $10.86 billion fair value of those coins, an unrealized loss documented in its own disclosures. Staking revenue reached $45.7 million in its latest quarter, 98% of total sales, but the company still posted an $83.6 million net loss, and its 30-day bounce (+22%) looks more like a dead-cat rally off a beaten-down base than a reversal.

Coinbase is the one name in this group that is not simply a leveraged bet on coin prices, and its second quarter shows why that distinction matters less than hoped: revenue fell to $1.2 billion from $1.5 billion a year earlier and the company posted a $359 million net loss, with both transaction revenue and subscription/services revenue — the fee-light, steadier business Coinbase has spent years building — down year over year. Average USDC balances held on the platform hit an all-time high of $20 billion, yet interest income from that stablecoin fell $17 million as rates declined — proof the business has its own rate sensitivity, but not one working in its favor right now. Trading market share hit a record 10.3% in the quarter, yet that share gain did not translate into revenue growth.

Verdict

On valuation, Strategy trades around 75x sales and BitMine around 170x trailing sales (falling to roughly 79x on forward estimates) — multiples that make sense only through the mNAV lens for this group, and both are now compressed versus their own multi-quarter ranges. That compression is CONFIRMS: it traces to real dilution, unrealized treasury losses and cash-dividend obligations, not simply to falling coin prices — Strategy's mNAV fell from 2.6x-2.8x toward parity even as bitcoin itself only round-tripped from a 2025 peak. Whether spot crypto is genuinely bottoming is INCONCLUSIVE: the band improvement is real, but ETF flows have not turned and bitcoin remains below the cost-basis levels associated with confirmed demand-led recoveries.

On the tape, bitcoin and ether both moved from a steep downtrend to a milder one on both the 30-day and 180-day views, while Strategy, BitMine and Coinbase all slid from a milder downtrend to a steep one on both the 90-day and 365-day views — the five-name split is consistent across every proxy in the group, not an artifact of one stock.

The setup

Where it stands — Bitcoin and ether have improved to a milder downtrend while Strategy, BitMine and Coinbase have all worsened to a steep one. Would confirm — Strategy's mNAV holds above 1.0x and monthly spot bitcoin ETF net flows turn positive for two consecutive months. Would invalidate — Bitcoin re-enters a steep downtrend on the 30-day view or BitMine's realized mNAV falls further below 0.84. Watch next — Strategy's next quarterly preferred-dividend payment and BitMine's ETH-per-share count at its next 10-Q filing. Valuation — Strategy trades near 75x trailing sales and BitMine near 170x trailing/79x forward sales, both below their own multi-quarter mNAV ranges.