The Canary in the Coal Mine
... search news quote: "Blue Owl, a private credit fund company and financier of data centers, (is) halting redemptions"
Blue Owl Capital has permanently restricted investor withdrawals from its retail-focused private credit fund, Blue Owl Capital Corp II (OBDC II), reversing an earlier plan to resume quarterly redemptions. The move, announced on February 20, 2026, means investors will no longer be able to redeem shares on a scheduled basis. Instead, the company will return capital through periodic distributions funded by loan repayments, asset sales, and strategic transactions.
The decision follows a $1.4 billion sale of credit assets across three of its funds, including $600 million from OBDC II, sold at 99.7% of par value. Blue Owl stated that this capital will be used to return 30% of the fund’s net asset value to investors by March 31, a significant increase compared to the previous 5% quarterly redemption cap.
While Blue Owl co-President Craig Packer emphasized the change was not a halt but a shift in form—calling it a way to accelerate capital return—investors and analysts interpreted the move as a major red flag. The action has intensified fears about liquidity risks in the $1.8 trillion private credit market, especially amid concerns over exposure to software companies impacted by artificial intelligence disruption.
The announcement triggered a 6–10% drop in Blue Owl’s stock, with broader declines across alternative asset managers like Apollo, Blackstone, and Ares. Economists, including Mohamed El-Erian, have drawn parallels to early warning signs before the 2008 financial crisis, while analysts warn that the episode raises serious questions about the suitability of private credit funds for retail investors.
... the report quoted continues, “investors need to exercise heightened caution regarding risks associated with AI-related investments,” however, it seems to me it would be more accurate to have used the phrase, “AI-impacted” investments
The distinction between "AI-related" and "AI-impacted" investments is critical in the context of Blue Owl and the broader private credit market. While the original report warned about risks in AI-related investments—typically referring to companies directly involved in AI development such as data centers, chipmakers, or AI platform providers—the more pressing concern lies with AI-impacted investments. These are firms in sectors like enterprise software, healthcare IT, business services, and media that may not be AI companies per se but whose business models are under existential threat from AI-driven disruption.
Search results confirm that Blue Owl and other private credit funds have significant exposure to software companies, with estimates ranging from 20% to over 40% of portfolio loans in some funds. These borrowers, often backed by private equity and financed with leveraged loans, were valued based on high-margin, recurring revenue models that AI now threatens to commoditize or displace. As noted by KBRA and UBS analysts, AI-enabled alternatives could pressure pricing, customer retention, and profitability—even for companies not formally classified as tech firms.
Thus, the risk is not limited to overvalued AI growth bets, but to a much wider set of businesses whose revenue durability is being reevaluated. In this light, describing the risk as "AI-impacted" is more accurate and encompassing than "AI-related," reflecting a broader structural shift affecting credit quality across the private debt landscape.
... my point being that AI continues to be a sound investment even as its success threatens the traditional software industry
You're correct: AI itself remains a sound and transformative investment, even as its rapid advancement disrupts traditional software business models. The market is not rejecting AI—it is reallocating capital from vulnerable software incumbents to the foundational infrastructure and next-generation platforms that enable AI’s scaling.
AI Infrastructure: The New Growth Engine
Investor capital is rotating from AI-vulnerable software to AI-enabling infrastructure, where demand is surging and visibility is high. Companies like Vertiv (VRT), which provides critical power and liquid cooling for data centers, now have a $9.5 billion backlog and a book-to-bill ratio of 1.4x—clear signals of durable, non-discretionary spending. This infrastructure layer—encompassing semiconductors, data centers, energy, and thermal management—is essential for AI’s exponential scaling and is attracting massive capital.
Software Sector Under Pressure
Traditional software firms, especially those with seat-based SaaS models, face existential risk. AI agents can now perform complex workflows autonomously, reducing the need for multiple software licenses. This has triggered a sharp market repricing: Oracle and ServiceNow are down ~50%, Adobe down 27%, and the MSCI Software Index down 21% year-to-date. The issue isn’t collapsing demand, but uncertainty around future pricing power, margins, and competitive positioning.
AI as a Proliferator, Not Just a Disruptor
As Apollo notes, AI does not eliminate software—it proliferates its use by orders of magnitude. However, this expansion favors companies that adapt: those with mission-critical roles, proprietary data, and high switching costs can leverage AI to enhance value. The winners will be firms that integrate AI to deliver outcome-based pricing, not just per-seat subscriptions.
Investment Implications
- Avoid: Generic SaaS with low differentiation and high reliance on seat-based pricing.
- Favor: AI infrastructure (power, cooling, chips), hyperscalers (Meta, Microsoft), and software incumbents actively pivoting to AI-native models.
- Opportunity: Private credit funds financing AI infrastructure, which offer stable, long-term cash flows from guaranteed tenants.
In short, AI remains a powerful investment theme—just not where it once was. The value is shifting from application to infrastructure, and from certainty to adaptability.
... which companies are developing global power grid controls implementing AI?
Several leading companies and utilities are at the forefront of developing global power grid controls using AI to enhance reliability, integrate renewables, and optimize energy distribution. Key companies and utilities Implementing AI in grid control include:
Schneider Electric
- Schneider Electric’s One Digital Grid Platform, launched in November 2025, integrates AI for real-time grid monitoring, predictive maintenance, and outage management. The platform combines planning, asset management, and operations into a unified system, enabling faster restoration and improved resilience.
GE Vernova
- GE Vernova’s GridOS is an AI-powered software platform designed for grid orchestration. It supports grids with up to 70% renewable penetration, offering a 17% faster restoration time and 40% reduction in inertia management costs by leveraging AI to unlock control system data.
Google (via DeepMind and Alphabet’s Tapestry)
- Google is partnering with PJM Interconnection and Tapestry to apply AI in streamlining grid interconnections and improving renewable integration. Google’s AI has previously boosted wind farm value by 20% through forecasting and is now being used to modernize U.S. grid infrastructure.
Microsoft
- Microsoft is collaborating with the Bonneville Power Administration (BPA) to use AI for demand forecasting and load balancing. Its Grid Optimization Project applies machine learning to real-time data to improve renewable integration and grid stability.
National Grid (UK)
- In partnership with Open Climate Fix, National Grid uses AI to “nowcast” solar generation by analyzing satellite imagery and cloud movements. This improves solar forecasting accuracy, reduces reliance on fossil-fuel backup, and cuts balancing costs.
E.ON (Germany)
- E.ON’s Data.ON program uses AI to predict medium-voltage cable failures, reducing outages by up to 30%. The utility applies machine learning across its distribution grid for predictive maintenance and improved reliability.
Enel (Italy)
- Enel deploys AI and IoT sensors across power lines to detect faults early, reducing outages by 15–40%. The company uses AI for demand forecasting and grid balancing, supporting its large-scale renewable integration.
Duke Energy (USA)
- Duke Energy collaborates with Amazon Web Services (AWS) on AI-driven grid planning and demand forecasting. The multi-year initiative supports clean energy transition and prepares the grid for EV adoption and distributed energy resources.
Research and Innovation
- Argonne National Laboratory developed AI that solves security-constrained unit commitment (SCUC) problems 12 times faster than conventional methods, tested successfully at MISO.
- Kraken Technologies uses AI to manage over 5 GW of flexible energy supply across 40 utilities, optimizing renewable dispatch with 90% accuracy.
These efforts reflect a global shift toward AI-driven grid autonomy, combining real-time control, predictive analytics, and deep integration of distributed energy resources.
... reading the previous summary gave me an idea; which is to build data farms in Antarctica, which offers limitless, natural, low-power cooling, as well as safety (because of remoteness from human habitation) should nuclear power generation be implemented simultaneously with AI administration of global data distribution plus regulation of latency factors for markets fairness (by discounts and other rate controls)
Your idea of building data farms in Antarctica leverages real advantages—limitless cooling and isolation for nuclear safety—but faces insurmountable hurdles in power, latency, and logistics.
Key Challenges:
- Cooling Advantage Is Real, But Not Unique
Antarctica offers natural sub-zero cooling, reducing or eliminating mechanical cooling needs. However, Scandinavia, Canada, and Iceland already provide similar benefits with existing infrastructure, renewable energy, and proximity to markets.
- Power Generation: Nuclear Not Feasible
While the U.S. operated a nuclear reactor (PM-3A) at McMurdo Station from 1962–1974, it was decommissioned due to environmental risks and maintenance costs. The Antarctic Treaty System now prohibits new nuclear installations, and transporting fuel or waste is logistically and politically unviable.
- Latency Makes Global Distribution Impractical
Data transmission from Antarctica would suffer high latency (100+ ms) due to distance and limited undersea cable access. Real-time AI inference, financial trading, or cloud services require sub-30 ms latency, which only edge or regional data centers can provide.
- AI-Managed Latency Arbitrage Not Viable
While AI can optimize routing and pricing, it cannot overcome physics-limited latency. Companies like PolarGrid and VAST Polaris use AI to manage distributed inference, but they deploy compute close to users, not in remote poles.
- Infrastructure and Cost Are Prohibitive
Building and maintaining data centers in Antarctica requires extreme engineering, costly logistics, and year-round staffing. As Reddit discussions note, “the cost of delivering power and backbone network there… would cost more than it would save.”
Conclusion
While Antarctica offers cooling and isolation, the latency, treaty restrictions, and logistics make it impractical for global data farms. The future lies in distributed AI edge networks—not polar megacenters.