Good morning, it’s Thursday. What state is famous for tangy barbecue, stunning coastlines, and… nuclear energy? South Carolina, of course! The Palmetto State is reviving a long-stalled nuclear project to fuel the AI era. Meanwhile, we’re diving into the $500 billion Stargate Project—because, seriously, that’s a lot of zeros.
Plus, today’s the grand finale of our 3-part series on Test-Time Compute—don’t miss it!
🗞️ YOUR DAILY ROLLUP
Top Stories of the Day
☢️ South Carolina Restarts Nuclear Plant
South Carolina's Santee Cooper is reviving a long-stalled nuclear project at V.C. Summer Nuclear Station to meet rising AI-driven energy demands. Partnering with Centerview Partners, the state utility seeks proposals for a tech giant-backed consortium to complete the reactors. Despite past setbacks, including a $9 billion loss, the move aligns with trends as AI leaders embrace nuclear energy for its clean, consistent power.
🔍 Perplexity Launches Sonar API for Real-Time Search
Perplexity's new Sonar API brings generative AI-powered search to apps, offering trusted, real-time results. With Sonar Pro, developers access deeper answers and enhanced citation handling. Priced at $5 per 1,000 searches, it’s a budget-friendly option for businesses like Zoom, which uses it for live, citation-backed insights. Backed by $500M, Perplexity aims to rival Google and OpenAI in the AI search landscape as a cost-effective alternative.
🏋️ China's DeepSeek Challenges US AI Dominance
On January 20, 2025, Chinese AI startup DeepSeek unveiled its latest model, R1, demonstrating reasoning capabilities comparable to OpenAI's o1. Developed using pure reinforcement learning without supervised data, R1 excels in mathematics, coding, and logical inference tasks. Notably, DeepSeek has open-sourced R1 under an MIT license, allowing free commercial use and posing a competitive challenge to U.S. AI firms that monetize similar technologies.
🧊 Instant 3D Modeling Breakthrough
Tencent's Hunyuan3D 2.0 transforms text or images into high-res 3D models in just 10 seconds. Open-source and lightweight, it uses multi-view diffusion to deliver unmatched accuracy and detail without heavy computing needs. Ideal for gaming, e-commerce, and entertainment, this tool automates technical tasks, empowering creators to focus on innovation. Available on Hugging Face and GitHub, it's a leap forward in accessible 3D content creation.
💰 Google Bets Another $1B on Anthropic's AI Future
Google has upped its investment in Anthropic by $1 billion, pushing its total backing to $3 billion. The funding fuels Anthropic's 2025 goals, including advanced AI models, new features for chatbot Claude, and the “Virtual Collaborator” system for executing workflows and coding. The investment comes as Anthropic pursues up to $2 billion in additional funding at a $60 billion valuation, doubling down on its position in the competitive AI race.
⚡ ENERGY CRUNCH
AI’s Energy Appetite: How Data Centers Are Reshaping the Global Power Landscape
The Recap: The rapid expansion of data centers, fueled by the demand for AI-powered services, is overloading power grids, raising energy costs, and delaying clean energy transitions. With U.S. data centers consuming 4% of national electricity in 2023 — a figure projected to reach 9% by 2030 — researchers at MIT are exploring sustainable solutions to this growing energy crisis.
A single data center can use as much electricity as 50,000 homes, and the demand for constant, uninterrupted power is a major challenge.
Grid infrastructure is struggling to keep up, with capacity limitations and delays in connecting new renewable energy sources adding to the problem.
Tech companies like Microsoft and Google are turning to nuclear power, including small modular reactors, to meet clean energy goals while supplying their data centers.
Researchers are exploring carbon-aware computing, which shifts computing tasks to areas or times where cleaner energy is available, though technical hurdles persist.
Data centers increase energy costs for residential consumers, with few local job benefits to offset their impact on communities.
MIT is developing energy-efficient chips, cooling systems, and algorithms to lower data center power consumption and improve operational efficiency.
Companies are exploring innovative energy solutions, from fusion and geothermal power to reactivating dormant nuclear plants, to ensure reliable energy for AI needs.
Forward Future Takeaways:
The growing power demands of AI and data centers are reshaping energy priorities, placing unprecedented strain on grids and delaying progress toward decarbonization. As companies and researchers explore innovative energy solutions like modular nuclear reactors and advanced grid management, there is a clear need for systemic change to balance technological growth with environmental and societal impacts. → Read the full article here.
👾 FORWARD FUTURE ORIGINAL
Conclusion and Future Outlook
“Automating the generation of improved model outputs by using additional inference-time computation also provides a path towards a general self-improvement algorithm that can function with reduced human supervision.”
Google, Arxiv-Paper
The future could be described as “sky is the limit”. The reasoning models using TTC are improving dramatically in a very short time. Within three months, OpenAI managed to saturate benchmarks that were thought to take years to saturate. The big breakthrough came recently with the solving of the ARC-AGI challenge.
The ARC-AGI Challenge is a public competition with over $1 million in prize money, aimed at developing a solution for the ARC-AGI benchmark and making it available as open source. The competition is hosted by Mike Knoop, co-founder of Zapier, and François Chollet, creator of ARC-AGI and Keras. The ARC-AGI (Abstraction and Reasoning Corpus for Artificial General Intelligence) benchmark is a unique test that measures an AI system's ability to efficiently learn new skills and apply them to unknown problems.
Unlike many other benchmarks that evaluate specific skills, ARC-AGI tests a system's general intelligence and adaptability. OpenAI recently presented the new model o3, which has achieved remarkable progress in this area. The o3 model achieved an impressive 87.5% accuracy on the ARC-AGI Semi-Private Evaluation Set using extensive computing resources and longer thinking time. This value exceeds the typical human average of 85% and represents a significant milestone in the development of AI systems. → Continue reading here.
💫 STARGATE PROJECT
A $500 Billion Leap Toward AI Leadership and Innovation
The Recap: The Stargate Project is a new initiative led by SoftBank and OpenAI to invest $500 billion over four years into building advanced AI infrastructure across the United States. Backed by major tech players like NVIDIA, Oracle, and Microsoft, this effort aims to solidify U.S. dominance in AI, create significant economic growth, and support global innovation in artificial intelligence.
The Stargate Project will immediately deploy $100 billion, with plans to invest a total of $500 billion over four years in AI infrastructure.
The initiative is expected to create hundreds of thousands of American jobs and support U.S. re-industrialization.
Initial equity funders include SoftBank, OpenAI, Oracle, and MGX, with SoftBank taking financial leadership and OpenAI overseeing operations.
Key technology partners are Arm, Microsoft, NVIDIA, Oracle, and OpenAI, all collaborating to build and manage cutting-edge computing systems.
The first phase of the project is underway in Texas, with plans for additional campuses being evaluated across the U.S.
The partnership leverages long-standing collaborations, such as OpenAI’s relationships with NVIDIA (since 2016) and Microsoft (via Azure).
The project aims to advance artificial general intelligence (AGI) and foster its use to "elevate humanity" through creative and strategic applications.
Forward Future Takeaways:
The Stargate Project signals a transformative shift in AI development, combining unprecedented financial investment and collaboration among top-tier tech companies. By fortifying America’s AI infrastructure and re-industrializing key regions, it is poised to catalyze innovation while reinforcing national security. If executed effectively, Stargate could define global AI leadership, setting the stage for breakthroughs in AGI and broader societal benefits. → Read the full article here.
🛰️ NEWS
Looking Forward
💊 AI-Designed Drugs Entering Human Trials: Google DeepMind CEO Demis Hassabis predicts AI-created drugs will hit clinical trials this year, potentially revolutionizing medicine. Pharma companies eye major cost and time savings.
📽️ VIDEO
The Industry Reacts to DeepSeek R1 - "Beginning of a New Era"
DeepSeek R1, a fully open-source AI model, is shaking up the industry with its groundbreaking performance, rivaling proprietary models like GPT-4 at a fraction of the cost. Its open weights and local deployment capabilities mark a new era in AI accessibility, sparking global excitement and redefining innovation in reasoning tools. Get the full scoop in Matt’s latest video! 👇
Reply