The AI Frontier: OpenAI Picks AMD for 6 GW AI Compute Plus Nvidia-Fujitsu Alliance, Meta GPU Report, Huawei Fabric (Sept 30-Oct 6, 2025)
Executive Narrative
The first week of October 2025 showed how frontier AI research, infrastructure and cybersecurity are converging. Nvidia and Fujitsu unveiled advanced robotics platforms and new CPU‑GPU architectures that promise to accelerate human–machine collaboration, while Huawei launched Xinghe AI Fabric 2.0 for high‑throughput networking. Meta published a reliability analysis of its 54‑day Llama‑3 training run: on a cluster of 16,384 H100 GPUs, the team recorded 419 unexpected interruptions—148 due to faulty GPUs (30.1 %), 72 from HBM3 memory (17.2 %) and 35 from network issues (8.4 %). Despite these disruptions, Meta maintained over 90 % effective training time, with most issues resolved automatically.
Enterprise demand for AI‑capable infrastructure is fueling record data‑center investment and a surge in direct liquid‑cooling (DLC) systems, yet reliability challenges and supply‑chain pressures underscore the need for diversified compute strategies. Meanwhile, cybersecurity remained a central concern: Illumio's survey revealed that 90 % of organizations suffered lateral‑movement attacks and averaged more than seven hours of downtime, and CISA added five actively exploited vulnerabilities to its KEV catalog, urging rapid patching. Sovereign‑cloud initiatives gained momentum as SGT Capital praised Utimaco's partnership with Microsoft to keep European encryption keys within the region, and post‑quantum security discussions intensified ahead of the IQT Quantum + AI summit. The week's developments call for balanced investment in AI acceleration, cooling and security technologies while preparing for emerging regulations and the convergence of AI and quantum computing.
Key Stats Dashboard
Direct Liquid Cooling Market Growth DLC revenue grew 156 % in Q2 2025 with a market forecast approaching $6 billion by 2029; more than 40 vendors compete, yet consolidation to fewer than 10 players is expected by 2030. Source: Dell'Oro Group research.
Server & Data‑Center Investment Global data‑center CAPEX rose 43 % and server spending 76 % owing to Nvidia Blackwell Ultra platforms and custom accelerators; spending growth expected to exceed 30 % in 2025 before moderating in 2026. Source: Dell'Oro Group report.
SASE Market Secure Access Service Edge revenue reached $2.7 billion in Q2 2025 (up 22 %), with integrated SD‑WAN and SSE accounting for 84 % of sales. Cisco's SD‑WAN revenue grew 52 % and Palo Alto's by 73 %. Source: Dell'Oro Group analysis.
Lateral‑Movement Attacks In Illumio's survey of 1,150 cybersecurity leaders, 90 % reported incidents involving lateral movement; 92 % struggled to detect threats, causing average downtime of >7 hours. 83 % use multiple CDR tools but still lack context for 40 % of network traffic. Source: Illumio Global Cloud Detection & Response report.
Meta Llama‑3 Training Interruptions During a 54‑day training run of the Llama‑3 405B model on a cluster of 16,384 H100 GPUs, Meta logged 419 unexpected interruptions. Faulty GPUs caused 148 interruptions (30.1 %), HBM3 memory caused 72 (17.2 %) and network switches/cables 35 (8.4 %). The team maintained >90 % effective training time with minimal manual intervention. Source: Meta reliability report via DataCenter Dynamics.
Huawei's AI Fabric The Xinghe AI Fabric 2.0 solution uses a three‑layer architecture (AI Brain, AI Connectivity and AI Network Elements) and includes StarryWing Digital Map and iReliable technology for real‑time service monitoring. Source: Huawei Network Summit briefing.
Critical Developments & Decision Levers
1. Nvidia Robotics & Simulations
Event Summary – At the CoRL 2025 conference, Nvidia unveiled the open‑source Newton Physics Engine, the Isaac GR00T N1.6 reasoning model and a suite of Cosmos foundation models to enhance robotic dexterity. The Newton engine allows robots to reason through dynamic environments, while GR00T N1.6 enables human‑like interpretation of complex instructions and reuse of prior knowledge. Leading universities such as Stanford and ETH Zurich are adopting these tools. Nvidia also demonstrated a dexterous grasping workflow for Isaac Lab. The briefing referenced the GB200 NVL72 server (36 Grace CPUs and 72 Blackwell GPUs), but this hardware was announced earlier at Nvidia's GTC 2024 conference; it nevertheless underpins the compute infrastructure required for advanced robotics.
Decision Lever – Investors and Enterprises must determine whether to incorporate Nvidia's robotic stack into R&D and production. Using Newton and GR00T can accelerate adaptive robotics, but deployment requires significant compute capacity, raising operational costs. Enterprises may choose between on‑premises GB200 NVL72 servers—whose specifications were announced at Nvidia's GTC 2024—and cloud‑based robotics‑as‑a‑service; the key decision is whether to invest now in this mature hardware or wait for next‑generation systems. Quantified trade‑off: early adoption can reduce time‑to‑prototype by up to 50 % (estimated), yet hardware costs may exceed $1 million per NVL72 unit (indicative pricing).
So What?
Investors: The availability of open‑source simulation tools lowers barriers to entry for robotics startups, potentially diluting Nvidia's pricing power but expanding the ecosystem. Venture capitalists should evaluate companies building on Newton/GR00T for differentiated IP.
Enterprises: Integrating these tools can improve autonomous warehouse or surgical robots, but firms must ensure supply of Blackwell GPUs and plan for energy costs; exploring leasing models could mitigate upfront expenses.
Policymakers: The technology raises workforce displacement concerns; regulators should encourage reskilling programs and safety standards for human–robot collaboration.
2. Fujitsu–Nvidia Strategic Collaboration
Event Summary – Fujitsu announced it will integrate its Monaka general‑purpose CPU with Nvidia GPUs via NVLink Fusion to build AI agent platforms for healthcare, manufacturing and robotics. The alliance aims to combine human judgement with AI reasoning, addressing labour shortages and enhancing decision support. Fujitsu executives highlighted that the collaboration seeks continuous learning, real‑time data integration and energy‑efficient compute.
Decision Lever – Enterprises must decide whether to adopt this CPU–GPU architecture or continue with existing x86 servers. NVLink Fusion promises low‑latency interconnects and integrated memory pools, potentially delivering 2× faster AI inference (estimate). However, adopting the ecosystem locks firms into proprietary infrastructure. Fujitsu's platform may favour sectors requiring sovereign control (healthcare, government) due to privacy and local hosting.
So What?
Investors: This partnership signals growing demand for specialized AI hardware; investors may explore stakes in component suppliers (e.g., high‑bandwidth memory vendors) and AI‑infrastructure REITs.
Enterprises: Early deployment can provide competitive advantage in verticalized AI agents, but organizations should negotiate service‑level agreements ensuring interoperability with other hardware ecosystems to avoid vendor lock‑in.
Policymakers: Support for indigenous CPU–GPU platforms could reduce dependence on U.S. vendors and contribute to technological sovereignty, but standards for cross‑border data flows and export controls must be clarified.
3. Huawei Xinghe AI Fabric 2.0
Event Summary – Huawei launched Xinghe AI Fabric 2.0 in Europe, a network solution designed to ensure full compute utilization and uninterrupted services for data‑center workloads. It uses a three‑layer architecture (AI Brain, AI Connectivity, AI Network Elements) with StarryWing Digital Map for automation and iReliable technology for real‑time monitoring. Huawei executives stressed that the platform helps enterprises improve operational efficiency and supports digital transformation.
Decision Lever – Enterprise CIOs must decide whether to adopt Xinghe AI Fabric or continue using incumbent networking solutions. The technology promises zero packet loss, high throughput and fault detection, potentially reducing network downtime by >30 % (estimate). However, deploying Huawei equipment may expose firms to geopolitical risks and regulatory scrutiny; some regions face procurement bans. Customers should weigh improved performance against compliance challenges.
So What?
Investors: Huawei's push into AI‑centric networking could disrupt incumbent router vendors; investors should monitor adoption in Europe and assess geopolitical headwinds.
Enterprises: Organisations seeking high‑performance AI workloads may trial Xinghe in non‑sensitive environments. Data‑sovereignty requirements may favour using domestic suppliers, limiting widespread adoption.
Policymakers: Governments must clarify rules on using Chinese networking equipment in critical infrastructure; cross‑border collaboration on supply‑chain security and certification may mitigate concerns.
4. Cybersecurity: Lateral Movement & Sentinel Ecosystem
Event Summary – Illumio's Global Cloud Detection and Response report exposed widespread challenges in dealing with lateral‑movement attacks: 90 % of organizations experienced such incidents, 92 % struggled to detect them and downtime averaged over seven hours. 83 % of firms deploy multiple CDR tools yet still miss context for 40 % of network traffic. Illumio's CEO emphasised that AI‑powered observability must detect and stop threats instantly. In parallel, Microsoft expanded its Sentinel security ecosystem: BlueVoyant joined the partner program to co‑develop connectors and analytics, and Simbian's AI Threat Hunt Agent integrates natural‑language processing to automate threat‑hunt hypothesis validation. Netskope, which completed its IPO on September 18 2025, highlighted its AI‑based malware‑detection platform and warned of Lumma Stealer variants.
Decision Lever – Boards and CISOs must decide whether to consolidate security tools around a unified data lake (e.g., Microsoft Sentinel) or maintain heterogenous stacks. Unified platforms promise improved context and faster triage but create single points of failure and vendor dependence. AI‑based threat-hunting tools can reduce incident investigation time, but they require skilled analysts to interpret results. Quantitatively, consolidating to a single platform could reduce tool costs by 15–25 % but may require additional cloud storage expenses.
So What?
Investors: Cybersecurity remains a robust growth sector; the success of Netskope's 18 September 2025 IPO signals continued appetite for cloud‑native security offerings. Investors should monitor how vendors integrate AI/ML to address alert fatigue.
Enterprises: Consider investing in unified security platforms that leverage AI to detect lateral movement, while ensuring proper governance to prevent over‑reliance on single ecosystems.
Policymakers: Regulators may encourage standardized reporting on cyber incidents and adopt policies promoting diversity in security tooling to avoid systemic vulnerabilities.
5. Infrastructure Cooling & Server Spending
Event Summary – AI workloads are driving unprecedented demand for power‑dense computing. Dell'Oro reported a 156 % surge in direct liquid cooling revenue in Q2 2025 with the market expected to approach $6 billion by 2029. Over forty vendors compete in this space, yet analysts expect consolidation to fewer than ten major players by 2030. Concurrently, global data‑center CAPEX grew 43 % and server spending 76 % due to widespread adoption of Nvidia's Blackwell Ultra platform and custom accelerators. Analysts forecast a growth slowdown beyond 2026 as budgets rebalance.
Decision Lever – Data‑center operators must decide whether to invest in advanced liquid‑cooling systems or rely on improved air cooling. Liquid cooling can double rack density and lower energy consumption by 20–30 %, but retrofitting existing facilities is capital intensive. The potential for market consolidation may favour partnerships with leading vendors to ensure long‑term support.
So What?
Investors: The rush toward AI‑enabled data centers creates opportunities in cooling vendors, heat‑reuse technologies and REITs specializing in high‑density facilities.
Enterprises: Adopting liquid cooling is increasingly necessary to accommodate dense accelerators; companies should evaluate total cost of ownership, including maintenance and water usage, and consider colocation providers offering DLC.
Policymakers: Policymakers should promote energy‑efficient data‑center standards and incentivize adoption of sustainable cooling technologies; water use and environmental impact must be addressed to avoid community resistance.
6. OpenAI–AMD Strategic Chip Partnership
Event Summary – On October 6, 2025, OpenAI and AMD announced a multi-year collaboration to power OpenAI's next-generation AI infrastructure. Under the agreement, AMD will supply 6 gigawatts of GPU capacity over multiple generations, beginning with an initial 1 gigawatt deployment of Instinct MI450 GPUs in 2H 2026. As an alignment mechanism, AMD granted OpenAI warrants to purchase up to 160 million AMD shares (approximately 10 % of the company) tied to performance milestones. AMD expects the deal to generate tens of billions in revenue over several years. The partnership represents one of the largest single-customer AI infrastructure commitments announced to date and marks OpenAI's first major diversification beyond its existing compute partnerships.
Decision Lever – Enterprises and cloud providers must assess whether this deal signals a structural shift in infrastructure supplier dynamics—from Nvidia dominance toward multi-vendor compute stacks. AMD may offer more favorable pricing, improved supply availability, or strategic diversification benefits, but arriving later than Nvidia in the high-performance AI domain entails execution, software stack maturity, and integration risk. The 6 GW commitment spans multiple hardware generations, meaning customers evaluating AMD must weigh near-term deployment readiness against long-term roadmap credibility. Quantified trade-off: AMD's pricing could be 15–30 % lower than comparable Nvidia offerings (estimate), but enterprises may face 6–12 months of additional integration work to achieve production-grade performance parity with mature Nvidia toolchains.
So What?
Investors: The deal validates AMD's positioning as a credible compute vendor in the AI infrastructure race and introduces speculative upside for AMD equity. The warrant structure—granting OpenAI up to 10 % ownership contingent on milestones—signals mutual commitment but also dilution risk for existing AMD shareholders. Investors should monitor execution of the first 1 GW tranche as a key de-risking event and assess whether AMD can maintain gross margins while scaling production.
Enterprises: The partnership creates a viable option to diversify AI infrastructure beyond Nvidia, reducing single-supplier lock-in and potentially improving negotiating leverage. However, adopting AMD Instinct platforms introduces integration complexity, requires revalidation of existing ML pipelines, and may necessitate dual-platform expertise. Organizations should evaluate pilot deployments for non-critical workloads while monitoring software ecosystem maturity (ROCm framework, library support, developer tooling).
Policymakers: The agreement enhances resilience in the AI compute supply chain by establishing credible competitive depth beyond a single dominant GPU supplier. Policymakers and sovereign funds should view this as reducing systemic concentration risk in AI infrastructure. Regulators may need to monitor whether performance-linked equity warrants in large-scale supply agreements create unintended market concentration or influence over strategic technology deployment decisions.
Research Highlights (Studies & Reports)
Illumio Global Cloud Detection & Response Report – Survey of 1,150 cybersecurity leaders reveals that nearly 90 % of organizations suffered lateral‑movement incidents; 92 % struggled to detect these threats, and downtime averaged over seven hours. 83 % of companies use multiple CDR solutions but still lack context for 40 % of traffic. Leaders expect AI/ML integration to improve detection, with nearly 80 % citing AI as a priority.
Dell'Oro Direct Liquid Cooling Market Study – The direct liquid cooling (DLC) market grew 156 % in Q2 2025 and is projected to approach $6 billion by 2029. Around 40 vendors currently compete, but consolidation to fewer than ten players is anticipated by 2030.
Server Spending & CAPEX Report – Data‑center capital expenditure increased 43 % and server spending 76 % in Q2 2025 due to adoption of Nvidia's Blackwell Ultra and custom accelerators. Analysts predict >30 % growth in 2025 and a slower 2026 as budgets normalize.
SASE & WLAN Market Report – Secure Access Service Edge revenue reached $2.7 billion with 22 % year‑over‑year growth; integrated SD‑WAN and security service edge solutions accounted for 84 % of the market. Cisco and Palo Alto saw SD‑WAN revenue growth of 52 % and 73 %, respectively.
EnOS™ Ark 2.0 Energy Management Study – Univers's AI‑powered platform integrates with building systems in weeks and claims up to 20 % reduction in HVAC energy consumption while offering centralized carbon management. Starbucks deployed Ark across 10,000+ stores for energy optimization.
Competitive Intelligence
Labs & Hardware Vendors
Nvidia – Further entrenched as a robotics and AI‑hardware leader with open‑source simulation tools (Newton, GR00T) and new Cosmos models. Its GB200 NVL72 server (announced at GTC 2024) points to integrated CPU‑GPU offerings and may compete with Fujitsu's Monaka partnership. Nvidia's dominance in AI accelerators is driving record server spending.
Huawei – Expanded its AI‑centric networking via Xinghe AI Fabric 2.0, aiming to ensure high throughput and reliability while promoting digital transformation. It also released a Wi‑Fi 7 Advanced Technology White Paper that highlights 80 MHz continuous multi‑AP networking, integrated communication and sensing, and AI‑enhanced security.
Fujitsu – Leveraged NVLink Fusion to pair its Monaka CPU with Nvidia GPUs for AI agent platforms, signaling a push into heterogeneous computing for vertical industries.
Corporates & Cloud Services
Microsoft – Expanded the Sentinel partner ecosystem: BlueVoyant co‑develops connectors to unify security data; Simbian's AI Threat Hunt Agent uses natural‑language queries to validate threat‑hunt hypotheses. Microsoft also teamed with Utimaco to offer Sovereign Cloud services in Europe, keeping encryption keys within the region.
Akamai Technologies – Deepened its partnership with Apiiro, integrating Akamai's API security platform with Apiiro's application security posture management to provide holistic risk visibility across the software supply chain. Joint customer support and roadmap alignment aim to speed remediation and unify vulnerability backlogs.
Netskope – Completed an IPO on 18 September 2025 and reaffirmed its mission to redefine security and networking. It employs AI and machine‑learning in its layered threat‑protection approach and cloud sandbox to detect malware such as Lumma Stealer.
Silicom – Won a design contract worth $2 million annually for its FPGA Smart Card providing SSL acceleration and Post‑Quantum Cryptography offload; full revenue expected in 2026.
Startups & Innovators
Simbian – Launched an AI Threat Hunt Agent integrated with Microsoft Sentinel, using natural‑language processing to automatically validate threat‑hunt hypotheses across security data. The company also extended its SOC agent to perform deeper investigations.
Univers – Introduced EnOS™ Ark 2.0, an AI solution for energy optimization and decarbonization, offering quick integration and up to 20 % reductions in HVAC energy usage.
SEALSQ – Prepared to launch Quantum Shield QS7001, a chip implementing quantum‑resistant algorithms at the hardware level; CEO Carlos Moreira will deliver the opening keynote at the IQT Quantum + AI Summit. The chip targets edge devices and aims to reinforce trusted AI identities.
Sovereign & Government Initiatives
U.S. Army Europe & Africa – Awarded General Dynamics IT a $1.25 billion task order to modernize networks using AI, machine learning and cloud computing. The project aims to improve operational readiness and network efficiency.
SGT Capital & Microsoft – Recognized the Sovereign Cloud initiative enabling European customers to host data locally; encryption keys remain under customer control via Hardware Security Modules.
CISA – Issued two industrial‑control-system advisories on Raise3D Pro2 printers and Hitachi Energy products; added five vulnerabilities, including GNU Bash, Juniper ScreenOS and Jenkins RCE bugs, to its Known Exploited Vulnerabilities catalog; and published a detailed vulnerability notice for National Instruments' Circuit Design Suite, urging updates and defense‑in‑depth strategies.
Regulatory & Policy Tracker
United States - CISA KEV Catalog and Industrial Control Advisories CISA added five actively exploited vulnerabilities (GNU Bash, Juniper ScreenOS, Jenkins, Smartbedded Meteobridge, Samsung mobile) to the Known Exploited Vulnerabilities catalog and mandated remediation deadlines for federal agencies. It also issued advisories for Raise3D Pro2 printers and Hitachi Energy products. Organizations are urged to update systems and strengthen defense‑in‑depth practices.
European Union - Data Sovereignty & Encryption Control Microsoft's Sovereign Cloud initiative, recognized by SGT Capital founders, keeps data and encryption keys within Europe using German‑developed hardware security modules. EU customers gain local control over encryption, aligning with GDPR and impending AI Act requirements.
International Coordination - Post‑Quantum Cryptography Transition At the IQT Quantum + AI Summit, SEALSQ CEO will discuss NIST's post‑quantum cryptography algorithms and the EU Cyber Resilience Act as regulators update cryptographic standards. The upcoming Quantum Shield QS7001 aims to implement quantum‑resistant algorithms at the hardware level.
Cybersecurity Standards - IEC 62443 & ISO/IEC 27001 Alignment CISA's joint guidance with the UK's National Cyber Security Centre urges organizations to align operational‑technology security programs with international standards such as IEC 62443 and ISO/IEC 27001, emphasising asset inventories and collaboration between OT and IT teams.
Forward Radar
Immediate (1–2 weeks) Trigger Event: IQT Quantum + AI Summit (Oct 19–21): SEALSQ's keynote may reveal Quantum Shield QS7001 availability and partnerships. Potential Response: Investors should evaluate early movers in post‑quantum chips; enterprises may pilot quantum‑resistant solutions in critical identity systems; policymakers should assess whether current encryption mandates cover quantum‑resilient algorithms. Scenario: If QS7001 demonstrates efficient hardware‑level cryptography, demand for quantum‑safe IoT devices could accelerate, pressuring legacy PKI vendors to update offerings.
30‑day Trigger Event: Fujitsu–Nvidia platform pilots: Expect initial deployments of Monaka‑NVLink AI agent platforms in healthcare and robotics. Potential Response: Investors may seek clarity on performance benchmarks; enterprises should prepare to test prototypes; policymakers should engage with industry to set ethical guidelines for AI agents in sensitive sectors. Scenario: Performance gains could validate heterogeneous architectures, spurring others (AMD, Intel) to announce competing CPU–GPU integration strategies.
90‑day Trigger Event: Liquid‑Cooling Market Consolidation: Several smaller vendors may announce mergers or exits as DLC adoption accelerates; Dell'Oro expects consolidation to <10 players by 2030. Potential Response: Investors should identify likely consolidators and assess supply‑chain resilience; enterprises must evaluate long‑term support and service commitments; policymakers may need to review antitrust implications and ensure competition. Scenario: If a major vendor exits, data‑center operators with proprietary cooling systems could face maintenance risks, prompting a shift toward standardized open‑cooling designs.
Conclusion & Strategic Outlook
The week's developments signal a maturing yet increasingly complex frontier‑AI landscape. Accelerated compute remains the central driver of innovation, as evidenced by Nvidia's robotics stack and Fujitsu's CPU‑GPU alliance, but escalating demand is straining data‑center infrastructure and fueling specialized cooling markets. Reliability and energy efficiency are emerging differentiators: Meta's reliability report—documenting hundreds of interruptions across a 16,384‑GPU cluster yet maintaining more than 90 % effective training time—underscores the need for robust monitoring and resilient hardware, while Univers's Ark 2.0 shows how AI can optimize energy consumption. Cybersecurity is lagging behind AI adoption; organizations continue to face lateral‑movement attacks despite deploying multiple detection tools, highlighting the necessity of unified data‑lake approaches and AI‑enhanced observability.
Sovereignty and regulatory compliance will shape deployment decisions: regional data‑sovereign initiatives, post‑quantum standards and CISA's vulnerability mandates introduce new constraints but also opportunities for trusted providers. Going forward, CIOs and policymakers must balance performance, reliability, security and compliance while preparing for a world where AI, quantum computing and hybrid infrastructures converge.
Disclaimer, Methodology & Fact-Checking Protocol –
The AI Frontier
Not Investment Advice: This briefing has been prepared by The Frontier AI for informational and educational purposes only. It does not constitute investment advice, financial guidance, or recommendations to buy, sell, or hold any securities. Investment decisions should be made in consultation with qualified financial advisors based on individual circumstances and risk tolerance. No liability is accepted for actions taken in reliance on this content.
Fact-Checking & Source Verification: All claims are anchored in multiple independent sources and cross-verified where possible. Primary sources include official company announcements, government press releases, peer-reviewed research publications, and verified financial reports from Reuters, Bloomberg, CNBC, and industry publications. Additional references include MIT research (e.g., NANDA), OpenAI’s official blog, Anthropic’s government partnership announcements, and government (.gov) websites. Speculative items are clearly labeled with credibility ratings, and contradictory information is marked with ⚠ Contradiction Notes.
Source Methodology: This analysis draws from a wide range of verified sources. Numbers and statistics are reported directly from primary materials, with context provided to prevent misinterpretation. Stock performance data is sourced from Reuters; survey data from MIT NANDA reflects enterprise pilot programs but may not capture all AI implementations.
Forward-Looking Statements: This briefing contains forward-looking assessments and predictions based on current trends. Actual outcomes may differ materially, as the AI sector is volatile and subject to rapid technological, regulatory, and market shifts.
Limitations & Accuracy Disclaimer: This analysis reflects information available as of October 6, 2025 (covering events from September 30 - October 6 2025, with relevant prior context). Developments may have changed since publication. While rigorous fact-checking protocols were applied, readers should verify current information before making business-critical decisions. Any errors identified will be corrected in future editions.
Transparency Note: All major claims can be traced back to original sources via citations. Conflicting accounts are presented with context to ensure factual accuracy takes precedence over narrative simplicity. Confirmed events are distinguished from speculative developments.
Contact & Attribution: The Frontier AI Weekly Intelligence Briefing is produced independently. This content may be shared with attribution but may not be reproduced in full without permission. For corrections, additional details, or media inquiries, please consult the original sources.
