Direct-to-Chip Liquid Cooling Supplier AI Data Center Risk / Opportunity Ranking

Thesis: Centamil has direct evidence of a production-ready AI thermal-management product aimed squarely at next-generation accelerator power density, with Hayagreeva HEx validated at 5,000W+ and framed as direct-to-chip liquid cooling across Nvidia, AMD, and Intel architectures.

Why now: All core evidence centers on the May 28, 2026 product launch, including a published-at external confirmation the same day. The product is positioned for chips expected to head toward 4,400W+ within 24 months, which supports a timely multi-quarter relevance argument.

Caveats: Evidence is concentrated in one launch event replicated across outlets, not multiple independent commercialization datapoints. No contracts, customers, or financing disclosures are provided. As a private company, business durability is hard to assess from the reviewed sources alone.

Thesis: ZJK shows credible opportunity through a combination of strong FY2025 financial growth, a manufacturing-product launch with claimed efficiency/cost advantages, and industry recognition in liquid-cooling connectors for data centers, suggesting expanding relevance to AI infrastructure supply chains.

Why now: The sequence matters: FY2025 results were disclosed on May 2, 2026, a next-generation dies launch came on May 13, 2026, and the liquid-cooling supplier award followed on June 2, 2026. That progression suggests operational momentum and growing recognition over the last two months of the recent evidence.

Caveats: The liquid-cooling-specific evidence is recognition rather than a disclosed revenue-bearing order. Vietnam expansion is mentioned without a dated timeline in the evidence, so recency is less certain. Cooling relevance is narrower component exposure versus full-system suppliers.

Thesis: Airsys has credible long-horizon opportunity as a private mission-critical cooling supplier scaling into AI/data-center demand, supported by a new $60M headquarters/manufacturing campus and leadership expansion aimed at surging global demand.

Why now: Two dated announcements in May 2026 show organizational and physical scaling: executive team expansion on May 6, 2026 and a new global headquarters campus opening on May 13, 2026, with manufacturing slated for Q1 2027, which fits a 1 year+ horizon better than a short-term one

Caveats: Manufacturing at the new campus is slated for Q1 2027, so operational payoff timing extends beyond immediate quarters. Evidence is mainly self-reported corporate communications.

Thesis: ChemTreat has a positive opportunity case because it was selected as a strategic national provider in Dow's Coolant Care Network for AI data center cooling, giving it a direct role in coolant supply, testing, analysis, and field support across a large installed base.

Why now: Why now is the May 18, 2026 announcement joining Dow's Coolant Care Network, which is recent and directly linked to AI data center cooling demand

Caveats: Single primary article drives the thesis. This is service/coolant-management exposure rather than direct hardware or platform exposure. No financial contribution, contract size, or margin impact is disclosed.

Thesis: CoolIT has strong strategic validation from its $4.75B-$4.8B sale and ongoing technical progress in single-phase DLC, including a 15kW coldplate milestone, indicating durable relevance in AI thermal infrastructure.

Why now: Two recent developments support the timing: a 15kW coldplate milestone was announced on June 1, 2026, and multiple June 2026 sources still discuss the Ecolab transaction at about $4.75B and associated revenue scale, suggesting continuing strategic importance over the next year. See

Caveats: Private/owned status means this is better framed as a watchlist or strategic-asset ranking, not a public tradable idea. Some supportive evidence comes through acquirer/owner commentary rather than CoolIT standalone filings. Market-growth report rows are thematic support, not direct company performance proof.

Thesis: Danfoss has credible direct-to-chip and CDU relevance through company and product reporting pages describing comprehensive CDU components and liquid-cooling connectivity solutions, while recent operating evidence shows adjacent product validation, strategic M&A in fluid conveyance, and local data-center-related job growth.

Why now: Recent June 2026 evidence shows Danfoss moving strategically in fluid conveyance through the Alfagomma exclusivity agreement, while supporting source pages within the recent evidence describe Danfoss CDU and liquid-cooling component offerings. The timing supports a 1 year+ strategic thesis, though direct monetization evidence remains less explicit.

Caveats: The clearest cohort-fit evidence is supporting source context, not fully mirrored in direct events. Recent direct event evidence is partly adjacent to electrification and industrial systems rather than purely data-center cooling revenue conversion.

Thesis: Iceotope has a credible opportunity profile based on fresh Series B funding and positioning around precision liquid cooling for AI hardware, with the reviewed sources also showing favorable thematic market growth.

Why now: The key dated catalyst is the $26M Series B announced on May 14, 2026, which can fund product development, patent expansion, and ecosystem partnerships over the next year.

Caveats: Most thesis support beyond the financing event is contextual rather than hard operating proof. No direct customer, backlog, revenue, or deployment metrics are provided in the core reviewed sources.

Thesis: LiquidStack shows meaningful opportunity through clear product relevance in direct-to-chip/CDU systems plus recent commercialization and order-scale context, but most of the strongest company-specific evidence comes from supporting article context rather than direct events in the core reviewed sources.

Why now: Later-dated supporting context in the reviewed sources points to commercial availability of the GigaModular CDU platform on May 21, 2026 and a 300MW CDU order on June 28, 2026, which, if durable, would matter over a 1 year+ horizon; however these are article-context rows and should be treated more cautiously than core direct event evidence.

Caveats: The only direct positive event in the core reviewed sources is largely market-level rather than LiquidStack-specific. Most attractive company-specific evidence is from supporting article context, which is lower priority than direct fact/events.

Thesis: MiTAC has the strongest direct positive evidence in this screen: repeated dated launches of liquid-cooled AI infrastructure across multiple events in April, May, and June 2026 suggest an expanding portfolio rather than a single isolated announcement. The most material evidence is the June 2026 debut of a 52U high-density liquid-cooled AI rack integrating 12 G4826Z5 AI servers with 96 AMD Instinct MI355X GPUs at COMPUTEX 2026.: MiTAC also showcased liquid cooling innovations and new server platforms at ISC 2026 on June 23, 2026. Earlier, on April 29, 2026, it launched the C2811Z5 OCP liquid-cooled server with AMD EPYC 9005 processors, showing the portfolio build started earlier in the recent evidence.

Why now: The evidence sequence has improved over the recent evidence: OCP liquid-cooled server launch on April 29, 2026, AI Expo Korea product launches on May 8, 2026, COMPUTEX rack-scale launch on June 3, 2026/June 4, 2026, and ISC 2026 liquid-cooling showcase on June 23, 2026/June 24, 2026. That progression supports a currently maturing liquid-cooled AI infrastructure push with potential 1 year+ relevance.

Caveats: Most evidence is company-issued or syndicated press release content. No financial impact quantified; product showcases may not translate into material revenue. Same-story duplicates should not be treated as independent confirmation.

Thesis: Vertiv has the strongest 1 year+ opportunity profile in the screen because recent evidence shows sustained order momentum, a roughly $15B backlog, Q1 2026 beat-and-raise execution, and multiple 2026 acquisitions expanding liquid-cooling and thermal-chain capabilities for AI data centers.

Why now: Recency is strong: Q1 2026 results and raised FY2026 guidance were reported on April 25, 2026, Strategic Thermal Labs was acquired around late April 2026, ThermoKey acquisition completed on June 12, 2026, and the next earnings update was noted as late July 2026, keeping the story active within the forecast window. See

Caveats: Several supportive rows are undated or supporting article context and are weaker than the direct Q1/guidance/acquisition evidence. Competitive references in relation rows are context-only and not independent counterparty propagation evidence.

Thesis: ACT has direct evidence of U.S. manufacturing expansion for AI data-center cold plates, including a 50,000-square-foot facility and initial annual capacity above 500,000 cold plates, which supports a durable supply-side scaling thesis.

Why now: The expansion was reported with exact reporting on May 19, 2026, and cold-plate manufacturing capacity is a long-duration build-out that fits a 1 year+ horizon better than a short catalyst thesis

Caveats: Single-article evidence base. No direct customer, order, or revenue conversion data in the reviewed sources. Private-company visibility is limited.

Thesis: Shinwa Controls has a credible multi-year opportunity because it just formalized a strategic collaboration with Wiwynn to deliver advanced liquid cooling for high-density data-center racks after three years of joint engineering. That points to a move from development into production-oriented positioning for global CSPs, which fits a 1 year+ opportunity horizon

Why now: Why now is strong because the formal collaboration was announced on June 3, 2026 and is framed as building on three years of validation, suggesting the relationship has matured into go-to-market phase

Caveats: Private company with low coverage confidence. Two main articles describe the same announcement and should not be treated as separate corroboration.

Thesis: OptiCool has direct evidence of relevance in high-density AI cooling through its partnership with Belden and two-phase rear-door heat exchanger offering supporting up to 120 kW per rack, which is meaningful for AI rack retrofit and close-coupled cooling demand.

Why now: The dated catalyst is the April 16, 2026 Belden partnership announcement, positioning OptiCool in integrated high-density AI infrastructure for enterprises and colocations. Recency beyond that is limited.

Caveats: Coverage confidence is low because only one direct article is present. No direct customer deployment evidence is provided. External CDU relevance exists, but the cited OptiCool CDU page has a May 29, 2025 reporting and is outside the direct recent event set, so it is more contextual than timely proof.

Thesis: Airsys has credible long-horizon opportunity as a private mission-critical cooling supplier scaling into AI/data-center demand, supported by a new $60M headquarters/manufacturing campus and leadership expansion aimed at surging global demand.

Why now: Two dated announcements in May 2026 show organizational and physical scaling: executive team expansion on May 6, 2026 and a new global headquarters campus opening on May 13, 2026, with manufacturing slated for Q1 2027, which fits a 1 year+ horizon better than a short-term one

Caveats: Manufacturing at the new campus is slated for Q1 2027, so operational payoff timing extends beyond immediate quarters. Evidence is mainly self-reported corporate communications.

Thesis: Alloy appears strategically valuable in AI/data-center thermal management because Johnson Controls agreed to acquire it, with the article stating the deal gives the buyer immediate access to advanced thermal management technology for data center cooling and high-performance computing; that points to durable strategic relevance over a 1 year+ horizon if the close and integration proceed as described.

Why now: The why-now is the April 14, 2026 report that Johnson Controls agreed to acquire Alloy Enterprises, with expected closing in Q3 2026, making ownership and strategic positioning time-sensitive. The same article also describes Alloy's focus on heat exchangers and fluid systems for data center cooling and HPC. Article timestamp is April 14, 2026 via exact reporting timestamp.

Caveats: Positive thesis is concentrated in a single article; same-article rows are not independent confirmation. No direct revenue, customer, contract, capacity, or funding evidence is provided in the reviewed sources. The article says the deal is expected to close in Q3 2026; completion was not separately confirmed in later evidence.

Thesis: Faster has direct evidence of entering the data-center thermal-management market with OCP-compliant liquid-cooling couplers and quick-disconnect products, creating a credible niche opportunity as liquid-cooling connector demand grows.

Why now: The relevant launch was reported on May 6, 2026 and specifically frames Faster's move as entry into a high-growth data-center market, making this a fresh strategic expansion for a 1 year+ horizon

Caveats: Only one article in reviewed sources. No direct revenue or order-size disclosure tied to the launch. Subsidiary evidence should not be over-propagated beyond what is directly stated.

Thesis: INVT has direct relevance to AI data center cooling through AI-ready power and cooling solutions, including an SC Series coolant distribution unit and broader integrated air-and-liquid cooling data center offerings.

Why now: June 2026 launch materials show INVT unveiling a new strategy and product portfolio, including liquid-cooling and data-center-adjacent products, which supports a 1 year+ watchlist opportunity but with modest conviction because commercialization evidence is thin.

Caveats: Evidence is mostly PR-style and lacks revenue, contract, or adoption metrics. Limited independent confirmation reduces confidence. Cooling relevance is real, but the magnitude of business impact is unclear.

Thesis: Apex has direct evidence of becoming Dow's national service provider for heat-transfer fluids used in data-center liquid and direct-to-chip cooling, which can support a multi-quarter services and maintenance niche in AI infrastructure.

Why now: Both included articles are from late May 2026 and describe a new service-provider designation tied directly to expanding data-center DTC cooling capabilities, making the commercial positioning recent for a 1 year+ horizon

Caveats: Evidence is based on two low-credibility press-style sources. No financial terms, backlog, customer count, or revenue contribution were disclosed.

Thesis: Johnson Controls has the strongest positive reviewed sources among the public names because it combines direct liquid-cooling product relevance, acquisition-led portfolio expansion, and powerful operating execution. Q2 FY2026 showed sales growth, 30% order growth, record $20B backlog, EPS growth, and raised FY guidance, while the company also completed the Alloy Enterprises acquisition to strengthen data-center thermal management and launched a 1MW Silent-Aire CDU platform according to June 28 external coverage

Why now: Why now is especially strong because the positive evidence is both recent and sequenced: Q2 beat and guidance raise on May 6, Alloy acquisition completion on May 13, Armada factory agreement on May 19, and June 28 external evidence of a 1MW CDU launch, showing continued cooling-platform buildout across the quarter

Caveats: The June 28 CDU launch is supporting context, not yet merged serving evidence, though it is dated and company-specific. Some positive rows in the reviewed sources are broader sentiment or institutional-flow items and were not used as primary thesis drivers.

Thesis: Aeroflex has direct evidence of a pivot into AI data center liquid cooling, specifically manifold and skid assemblies, with a planned capacity expansion from 6,000 to 15,000 units and a target for skid assemblies to become a meaningful share of FY27 revenue.

Why now: Recent May 2026 reporting ties Aeroflex directly to India's data center buildout and specifies near-term capacity expansion over the next two quarters with utilization goals by March 2027, making the thesis timely and relevant for a 1 year+ horizon.

Caveats: Most core evidence comes from a small number of Financial Express articles, so corroboration depth is limited. Some positive flow and multibagger references are market/sentiment context, not direct operating proof. Export concentration creates additional macro and regional sensitivity.

Thesis: Main risks are execution and cyclicality rather than a current adverse event: no financial terms were disclosed for the Adani platform, and some reviewed sources items note geopolitical headwinds, peer underperformance in some periods, and insider selling context.

Why now: On June 17, 2026, Jabil reported fiscal Q3 results that beat expectations and raised FY2026 revenue, margin, EPS, free cash flow, and AI revenue outlook, citing strong AI-driven data-center demand. Just before that, on June 15-16, 2026, Jabil and Adani announced an AI/data-center manufacturing platform in India covering high-density and liquid-cooled AI racks, PDUs, CDUs, thermal management, servers, storage, and networking, creating a long-horizon expansion vector

Caveats: One positive evidence item is clearly misattributed to Brown-Forman and is excluded from the thesis. Part of the India opportunity is strategic-announcement stage with no disclosed economics.

Thesis: The main risk is not a direct adverse event but limited evidence depth: the reviewed sources show only one article and no customer wins, manufacturing milestones, or recurring commercial traction, so execution risk remains hard to assess. The product description is also carried as an undated fact, making recency of operating status less certain

Why now: Why now is the recent financing event on May 28, 2026, which explicitly supports scaling of data-centre infrastructure systems and expansion of teams, creating a long-horizon capacity-building catalyst

Caveats: Only one article in coverage; low corroboration. Product-operational fact is marked undated, so recency of that specific operating description is uncertain.

Thesis: Primary risk is evidentiary rather than operationally adverse: much of the direct-to-chip product proof comes from external company-page context and context-only ownership/financing references rather than broad independent operating disclosures.

Why now: The key timing evidence is Accelsius' Series B at approximately $665M post-money reported on April 23, 2026 as fully funding commercial scale-up, followed by reporting in April-May 2026 describing general availability of NeuCool IR150 and direct-to-chip/two-phase positioning reported April 23, 2026, April 20, 2026, May 25, 2026).

Caveats: Several product-performance points come from external company pages or weak supporting context, not direct dated events. Relation rows are context-only and cannot be used for counterparty propagation.

Thesis: Risk is relatively moderate in these reviewed sources; the main concern is that some evidence mixes Dow Inc. with Dow Jones or broad macro context, and the best liquid-cooling product proof partly comes from supporting source context rather than a fully quantified recent sales disclosure.

Why now: Dow's data-center-cooling relevance is timely because recent May 2026 reporting says it launched a data-center cooling program, while the screen also includes direct external product context on DOWFROST LC for direct-to-chip liquid cooling and recent industrial contracts supporting longer-horizon capacity and feedstock positioning. recency uncertain versus recent evidence

Caveats: Some evidence in the reviewed sources refers to Dow Jones rather than Dow Inc. and was not used for the operating thesis. Supporting source evidence for DOWFROST LC includes a November 5, 2025 publication date, so it is relevance-rich but not a fresh operating catalyst by itself. No direct adverse operating evidence was available in these reviewed sources, but that should not be overread as absence of business risk.

Thesis: Documented adverse evidence is absent, but the main risks are that much of the positive evidence comes from undated or promotional articles, and the reviewed sources does not provide customer wins, financial conversion, or contract-level proof.

Why now: Recent evidence includes H3C's AI/server messaging around its May 2026 summit and a June 2026 Wi-Fi 8 product launch, alongside inclusion in a June 2026 AI server growth report; timing indicates an active product and expansion phase as of May 22, 2026 to June 4, 2026 and later expo context on June 24, 2026/June 25, 2026 reporting.

Caveats: Several positive evidence are marked undated, so recency-sensitive claims should be treated cautiously. Multiple summit-related articles are similar reports around the same theme and not independent confirmation. No direct customer contract, backlog, revenue, or margin evidence is provided.

Thesis: Risk is mainly commercialization uncertainty: the evidence is dominated by company-style launch articles with limited direct proof of orders, financing, or customer adoption.

Why now: Why now is the June 25, 2026 ISC 2026 launch cycle, which adds newer evidence beyond April product claims and suggests expanding platform breadth across infrastructure and management software

Caveats: Evidence is mostly medium/low-credibility distribution of launch materials rather than independent reporting. Some time fields are coarse or undated, so exact sequencing between launches is not fully precise.

Thesis: Risk is mostly evidence-concentration risk: the thesis is compelling but relies heavily on a single article, with no direct negative evidence and limited independent confirmation.

Why now: The why-now is concrete and recent: on April 15, 2026, the company reported follow-up data-center orders and indicated H1 2026 order volume would exceed all of 2025, alongside a planned 17,000 m² plant. That timing fits a 1 year+ capacity and revenue-conversion story. See

Caveats: Almost all material evidence comes from one article. No explicit margin or profitability read-through is provided.

Thesis: Risk is mainly commercialization uncertainty rather than direct adverse evidence: claims are product-led and partner-led, but the reviewed sources provide no revenue, backlog, financing, or manufacturing-scale proof for a private company.

Why now: The most recent cluster of evidence is late May to early June 2026: Lenovo collaboration on May 28, 2026 and multiple June 4, 2026 articles on AirJet Mini in Intel Wildcat Lake and LiquidJet Nexus for Nvidia AI accelerators. That timing suggests active go-to-market momentum now for a 1 year+ commercialization window.

Caveats: Several performance claims come from product/demo articles and should be treated as early validation rather than proven scaled adoption.

Thesis: Risk is low but not absent. The main limitation is concentration: all material evidence comes from versions of the same company announcement, giving little independent confirmation on order visibility, margin impact, or customer conversion. There is no direct adverse event in the reviewed sources, so risk remains mainly execution/visibility risk rather than a documented problem

Why now: Why now is strong because the expansion announcement was crawled June 11-12, 2026 and explicitly covers 2026-2027 capex and capacity buildout, making it highly relevant for the selected 1 year+ forecast horizon

Caveats: All major evidence traces to the same expansion announcement family and should not be treated as independent confirmation. Subsidiary-level visibility is limited.

Thesis: The main risk is early-stage execution: the reviewed sources show market entry and product showcase evidence, but limited proof yet of large customer contracts, backlog, or recurring economics from the new AI cooling segment.

Why now: The catalyst timing is clear and recent: Walrus publicized its Computex 2026 AI server cooling entry in late May 2026, and the reviewed sources link that move with record Q1 2026 consolidated revenue of NT$483 million

Caveats: No direct contract-win or backlog evidence in the reviewed sources. Multiple articles appear to restate the same launch event. Opportunity is more strategic-entry than proven-scale revenue conversion so far.

Thesis: Risk is present but moderate in the reviewed sources: heavy insider selling, elevated valuation-style commentary in some articles, and acquisition/integration burden from the CommScope CCS deal are the main offsets rather than core business deterioration.

Why now: The reviewed sources show a concentrated sequence of late-April to June 2026 evidence: Q1 record results and Q2 guidance, plus prior April 2026 commentary that Amphenol launched liquid-cooling connectors for AI data centers, EV charging, and storage, making this both a current operating story and a durable 1 year+ infrastructure thesis

Caveats: Most of the available evidence is earnings/institutional-flow heavy rather than liquid-cooling-specific, though direct connector relevance is present. Many supportive rows repeat the same Q1 beat and guidance event.

Thesis: There is no direct adverse company-specific evidence in the reviewed sources. The main limitation is narrow coverage: beyond the April 29 investment-expansion article, the rest is broad industry context rather than proof of bookings, capacity, or customer deployment.

Why now: The key dated catalyst is Carrier Ventures' expanded investment on April 29, 2026, which directly points to scaling activity rather than merely technical relevance. That timing is recent within the recent evidence and plausibly matters over a 1 year+ horizon.

Caveats: Coverage confidence is limited because the company has only a small article universe reviewed-source. No direct evidence of customer wins, revenue, or deployment scale.

Thesis: There is no direct adverse evidence on STL itself in the reviewed sources. Main caveat is structural: as an acquired private company, standalone independent operating trajectory are no longer the relevant lens, and no deal terms are disclosed.

Why now: The key dated event is Vertiv's acquisition announcement cluster on April 27, 2026 through April 29, 2026, which directly changed STL's business state from independent specialist to acquired strategic asset.

Caveats: Acquired-company status means the thesis is more a strategic validation thesis than a standalone upside thesis. No deal terms are disclosed. Post-acquisition integration limits transparency.

Thesis: The reviewed sources show little direct adverse company-specific evidence; main risk is that much of the demand signal comes from partnerships/ecosystem context rather than disclosed contract economics, limiting certainty on conversion and magnitude.

Why now: As of June 4, 2026 reporting, Wiwynn formalized a strategic collaboration with Shinwa Controls after three years of joint engineering, signaling product readiness for next-generation data-center cooling solutions targeted at global CSPs. Additional late-May to mid-June evidence ties Wiwynn into AMD's Taiwan AI buildout and Computex optical/liquid-cooling demonstrations

Caveats: Several supportive items are partnership/ecosystem articles rather than disclosed customer wins or commercial commitments. Some supporting context is undated or contextual and should not be treated as independent confirmation.

Thesis: No direct adverse company-specific evidence appears in the reviewed sources. Main risk is thin evidence breadth: the case relies heavily on one technical-validation article plus weaker supporting article context, with limited proof of customer adoption or financial outcomes.

Why now: The direct positive catalyst is recent: technical validation article dated by May 28, 2026, followed by company and product reporting from June 17, 2026 that the CDU was launched and available for order. That sequence supports a 1 year+ product commercialization setup.

Caveats: Coverage confidence is low because article universe is small. Launch context is article-level and weaker than direct event evidence.

Thesis: Main risk is execution and deal uncertainty rather than business deterioration: the transaction is still subject to regulatory clearances, and some deal-term reporting is inconsistent on value.

Why now: Why now is immediate and durable because the acquisition was announced on June 10, 2026, which supersedes the earlier standalone startup state and creates a new business-state setup for the next year

Caveats: The transaction is not yet closed and is subject to regulatory clearances. One article reports an $800M price, conflicting with multiple up-to-$400M reports, so exact consideration is not fully settled in the reviewed sources Much of the operating-performance evidence is promotional or research-oriented rather than customer-conversion evidence.

Thesis: The reviewed sources show limited company-specific adverse evidence; risk mainly comes from execution complexity in very large AI deployments and reliance on broader AI capex cycles rather than any disclosed operational setback.

Why now: The evidence is both recent and sequenced: Q1 2026 beat and guidance reaffirmation were disclosed April 30, 2026; over $290M of AI infrastructure delivery for TeraWulf was announced May 26, 2026; Uniflair XCA chiller launches were announced June 2-3, 2026. That creates a strong 'already converting demand into deployed systems' setup for 1 year+.

Caveats: Some demand-tailwind evidence is thematic and not company-specific, so the main weighting should stay on company events. Several partnership references with Nvidia are repeated across similar summit coverage and are not independent confirmation.

Thesis: Source-specific downside is limited; the main risk is that the strongest direct evidence is concentrated in a small number of customer/project announcements and some newer product context is supporting article context rather than fully structured company event evidence.

Why now: Recent evidence includes the May 2026 TeraWulf/Lake Mariner deployment tied to a 750 MW campus and later June 2026 reporting showing Motivair's expanded CDU range and Schneider's end-to-end AI liquid-cooling positioning, indicating momentum is current rather than stale.

Caveats: Some recency-sensitive product-portfolio evidence is supporting article context rather than direct event evidence. Project concentration is visible in the reviewed sources; much of the strongest evidence ties to Lake Mariner/TeraWulf.