US GDP grew 3.4% in the third quarter of 2026, with rising investment in information-processing equipment playing a part. Private equity firms like Blue Owl, Ares, and Blackstone are buying data centers to collect long-term rent from tech companies. Blue Owl alone operates or is building over 130 data centers with nearly 14 gigawatts of capacity, locking in steady income for decades.
Powering all these data centers is a serious challenge, and tech companies are turning to nuclear energy. Amazon took 690 megawatts from Constellation's Calvert Cliffs plant in Maryland, while Google signed a deal adding 890 megawatts of nuclear capacity in the US. Constellation is spending over $4.3 billion to upgrade 11 reactors across Illinois, Pennsylvania, and New Jersey. No new reactors are being built; companies are paying to restart or upgrade existing ones.
The resource demands of AI go beyond electricity. The International Energy Agency estimates global data center power use could reach 945 terawatt-hours by 2030, more than double 2024 levels. Data centers used about 222 billion liters of water for cooling in 2025. In Finland, 330 hectares of forest were cleared for a Google data center, and police are investigating possible environmental violations. Norway and several US states are considering limits on new data center construction.
Nscale, a company backed by Nvidia, is building a large data center campus in West Virginia as part of its $35 billion IPO plans. The project aims to create 1,000 jobs and generate $1 billion in annual revenue but faces serious obstacles. Construction delays have occurred because the site has difficult geology that required deeper drilling and added costs. Local residents have opposed the project due to environmental concerns, and Nscale also struggles with financing risks as it tries to raise funds while competing with other data center operators.
On the cybersecurity front, Israeli startups are dominating Fortune's 2027 Cyber 60 list. Israeli-founded companies made up 55% of growth-stage picks and 60% of early-stage startups. Cato Networks has over 4,800 customers and annual recurring revenue above $415 million. Island raised $400 million in September 2026 at a $6.4 billion valuation. Noma Security, founded in 2023, raised $132 million to protect enterprise AI applications.
Security experts warn that North Korean hackers are using advanced AI and post-quantum encryption in their attacks. The Lazarus group targeted South Korean companies since July 2025 using fake AI-generated resumes. The attackers used a tool called LPEClient that communicates via Dropbox and encrypts data with GOST, a Russian cipher. They also employed ML-KEM-1024, an encryption method designed to resist future quantum computer threats, allowing them to steal data now and decrypt it later.
A ServiceNow study found that 90% of global organizations experienced an agentic AI security incident in the past year. Only 13% of businesses have a defensible security posture for AI agents, and just 44% of security teams build safeguards before an agent is deployed. Many organizations run hundreds or thousands of AI agents without full visibility or automatic protection. CISOs are now focused on turning AI policies into real governance and closing visibility gaps around AI agent activity.
Researchers have developed a new system called JevAct to verify the actions of large language model agents before they execute them. The framework checks if an agent follows rules and completes tasks correctly without needing to change the agent's original code. Testing on eleven different AI agents showed that JevAct improves reliability, especially on telecom and banking tasks.
Kioxia Holdings reported a huge jump in operating profit to 1.99 trillion yen for the quarter ending September 30, driven by high demand for AI data storage. Revenue reached 2.48 trillion yen as customers focused on generative AI paid higher prices. This is the company's first ever dividend, set at 90 yen per share. Kioxia also spent 800.3 billion yen buying back its own shares and now holds 1.4 trillion yen in cash.
MHH reported that its Data and AI revenue grew by 4.7% while bookings increased by 29% in the third quarter. However, the company's total revenue fell by 13.4% compared to the same period last year. The business is shifting its investment focus toward Applied AI, Modern Data, and AI governance. These results show that while overall sales dropped, the strategic push into artificial intelligence is gaining momentum.
Key Takeaways
- US GDP grew 3.4% in Q3 2026, partly driven by rising investment in information-processing equipment
- Blue Owl operates or is building over 130 data centers with nearly 14 gigawatts of capacity
- Amazon took 690 megawatts from Constellation's Calvert Cliffs plant; Google signed a deal adding 890 megawatts of nuclear capacity
- Constellation is spending over $4.3 billion to upgrade 11 reactors in Illinois, Pennsylvania, and New Jersey
- Global data center power use could reach 945 terawatt-hours by 2030, more than double 2024 levels
- Israeli-founded companies made up 55% of growth-stage picks on Fortune's 2027 Cyber 60 list
- Cato Networks has annual recurring revenue above $415 million; Island raised $400 million at a $6.4 billion valuation
- North Korean Lazarus group uses fake AI-generated resumes, LPEClient tool, and post-quantum encryption targeting South Korean companies since July 2025
- 90% of global organizations experienced an agentic AI security incident in the past year; only 13% have a defensible security posture
- Kioxia reported operating profit of 1.99 trillion yen and declared its first dividend of 90 yen per share
AI Investment Drives US Growth as Wall Street Rents Data Centers
US GDP grew 3.4% in the third quarter of 2026, partly due to rising investment in information-processing equipment. Companies like Blue Owl, Ares, and Blackstone are buying data centers to collect long-term rent from tech firms. Blue Owl operates or is building over 130 data centers with nearly 14 gigawatts of capacity. These deals give investors steady income for decades.
AI's Real Bottleneck Is Electricity and Who Will Pay
AI data centers need huge amounts of power, and tech companies are turning to nuclear energy to meet it. Amazon took 690 megawatts from Constellation's Calvert Cliffs plant in Maryland. Google signed a deal adding 890 megawatts of nuclear capacity in the US. Constellation is spending over $4.3 billion to upgrade 11 reactors in Illinois, Pennsylvania, and New Jersey. No new reactors are being built; companies are paying to restart or upgrade existing ones.
AI Is Consuming Resources and the World Is Paying the Price
AI requires massive amounts of electricity, water, and land. The International Energy Agency estimates global data center power use could reach 945 terawatt-hours by 2030, more than double 2024 levels. Data centers used about 222 billion liters of water for cooling in 2025. In Finland, 330 hectares of forest were cleared for a Google data center, and police are investigating possible environmental violations. Norway and several US states are considering limits on new data center construction.
Israeli Startups Dominate Fortune's 2027 Cyber 60 List
Israeli-founded companies made up 55% of growth-stage picks and 60% of early-stage startups on Fortune's 2027 Cyber 60 list. Cato Networks, co-founded by Shlomo Kramer, has over 4,800 customers and annual recurring revenue above $415 million. Island, co-founded by Dan Amiga, raised $400 million in September 2026 at a $6.4 billion valuation. Noma Security, founded in 2023, raised $132 million to protect enterprise AI applications. Act Security and Astelia focus on cloud access and vulnerability management.
The CISO's AI Priority List: What Changes and What Does Not
AI agents are changing how organizations use data and applications, but cybersecurity basics still matter. CISOs must turn AI policies into real governance by understanding what AI tools are running and what data they access. Closing visibility gaps is critical so teams can see an AI agent's full activity from start to finish. Patching and lessons learned remain essential because AI does not remove old risks. Strong fundamentals must continue even as new AI threats emerge.
North Korean hackers use AI and post-quantum encryption
Security experts warn that North Korean hackers are using advanced AI and post-quantum encryption in their attacks. A group called Lazarus targeted South Korean companies since July 2025 using fake AI-generated resumes to trick employees. The attackers used a tool called LPEClient that communicates with operators via Dropbox and encrypts data with GOST, a Russian cipher. They also employed ML-KEM-1024, a strong encryption method designed to resist future quantum computer threats. This strategy allows them to steal data now and decrypt it later when quantum computers become powerful enough.
Nscale faces challenges with its $35bn West Virginia data center
Nscale, a company backed by Nvidia, is building a large data center campus in West Virginia as part of its $35bn IPO plans. The project aims to create 1,000 jobs and generate $1bn in annual revenue but faces several serious obstacles. Construction delays have occurred because the site has difficult geology that required deeper drilling and added costs. Local residents have also opposed the project due to environmental concerns and have demanded the company address their worries. Additionally, Nscale struggles with financing risks as it tries to raise the necessary funds while competing with other data center operators.
Kioxia profit soars on AI demand and declares first dividend
Kioxia Holdings reported a huge jump in operating profit to 1.99 trillion yen for the quarter ending September 30, driven by high demand for AI data storage. Revenue reached 2.48 trillion yen as customers focused on generative AI paid higher prices for their products. This is the company's first ever dividend, set at 90 yen per share for the fiscal year ending March 2027. The company also spent 800.3 billion yen buying back its own shares and now holds 1.4 trillion yen in cash. Kioxia plans to bring American depositary receipts to market next year to help fund future production spending.
MHH sees AI growth and bookings rise despite lower revenue
MHH reported that its Data and AI revenue grew by 4.7% while bookings increased by 29% in the third quarter. However, the company's total revenue fell by 13.4% compared to the same period last year. The business is shifting its investment focus toward Applied AI, Modern Data, and AI governance to support these growth areas. Management did not provide specific financial guidance for the future. These results show that while overall sales dropped, the strategic push into artificial intelligence is gaining momentum.
New tool JevAct improves safety for AI agents using tools
Researchers created a new system called JevAct to verify the actions of large language model agents before they execute them. This framework checks if an agent follows rules and completes tasks correctly without needing to change the agent's original code. It uses a decision model to judge actions based on facts and conversation context rather than generating new text. Testing on eleven different AI agents showed that JevAct improves reliability, especially on telecom and banking tasks. The tool helps prevent agents from making mistakes or taking actions that violate safety policies.
Most Enterprises Lack Proper Controls for Agentic AI
A ServiceNow study found that 90% of global organizations experienced an agentic AI security incident in the past year. Only 13% of businesses have a defensible security posture for AI agents. Just 44% of security teams build safeguards before an agent is deployed. Many organizations run hundreds or thousands of AI agents without full visibility or automatic protection.
Beverly Hills School Board Reviews New AI Curriculum
The Beverly Hills Unified School District Board of Education heard about an AI curriculum on Oct. 27. The curriculum is being taught to Beverly Hills High School students in grades nine and ten. AP Computer Science students are also included in the program.
New Study Combines Group-Lasso With Siamese Networks
Researchers developed a new method that brings Group-Lasso into a Siamese network for feature selection. The approach finds compact feature subsets while preserving data representation and classification performance. The method was tested on 12 benchmark datasets with feature dimensions from 6 to 7070. Results showed competitive or better performance compared to existing methods.
Sources
- AI Investment Fuels US Growth as Wall Street Locks In Data Center Rents
- AI’s real bottleneck is electricity. Who will pay for it?
- Artificial intelligence is consuming resources: what price is the world paying?
- Israeli-founded startups claim nearly half of Fortune's 2027 Cyber 60 list
- The CISO’s AI Priority List: What Changes and What Doesn’t
- Security experts warn North Korean hackers aren't just using AI in their new campaigns, but also 'post-quantum' encryption
- Inside the make-or-break AI mega-project behind Nscale’s $35bn IPO
- Kioxia profit jumps on AI data storage demand as it declares its first dividend
- MHH: Data and AI growth and bookings improved as strategic investment accelerated despite lower total revenue
- Think Before You Act: Decision Gate for LLM Agents
- Ungoverned Agentic AI Exposure is Compounding Inside Enterprises
- School Board Hears Presentation on AI Curriculum
- Integrated Feature Selection for classification and Representation Learning with Group-Lasso Siamese Networks
Comments
Please log in to post a comment.