This week in computing news: Google’s AI Co-Scientist leverages the Gemini 2.0 model to assist biomedical researchers with data synthesis and hypothesis generation. South Korea is set to build a 3GW AI data center by 2028 (triple the capacity of many current 1GW hyperscale projects), and Microsoft’s Majorana-1 chip introduces a topological qubit architecture using Majorana particles, laying groundwork for scalable quantum computing. Here’s a look at each.

Google’s AI Co-Scientist

Google’s AI Co-Scientist, built on the Gemini 2.0 foundation model, is designed to accelerate scientific discovery. Unlike general-purpose AI, it’s tailored to synthesize vast scientific literature—decades of research—and generate testable hypotheses and experimental plans. In a trial focused on liver fibrosis, it reviewed 20 years of data in under an hour, producing six novel approaches, all showing measurable activity against disease factors when tested by Stanford and Imperial College researchers.

Integration into research labs is a key next step. Google offers a Trusted Tester Program, allowing scientists to join and provide feedback for refinement. The system aims to collaborate with lab teams by offering real-time suggestions, with plans to enhance its reasoning via continuous updates to its training data—potentially incorporating new publications as they emerge. While specifics on lab deployment timelines remain undisclosed, Google’s blog emphasizes iterative improvements based on tester input, signaling a phased rollout to ensure reliability in high-stakes research environments.

This tool doesn’t replace scientists but augments their work, processing data at a scale—hundreds of papers in minutes—that humans can’t match. Its success hinges on lab adoption and data access, with Google likely eyeing partnerships to embed it into academic and industry workflows.

South Korea’s 3GW AI Data Center

South Korea is spearheading a $35 billion project to construct a 3GW AI data center in Jeollanam-do province, set to begin construction in late 2025 and finish by 2028. This dwarfs other major projects, such as South Korea’s own 1GW Gangwon and Jeollanam-do campuses announced in 2023 and 2024, or Microsoft’s multi-gigawatt clusters in the U.S., which typically hover around 1GW per site.

The facility will support next-generation AI training and inference, featuring advanced cooling systems and high-bandwidth fiber for regional and international connectivity. It’s engineered to handle sudden energy load spikes—common in AI workloads—with a projected 10,000 jobs created in energy, equipment, and R&D sectors.

Compared to 1GW projects, the 3GW scale reflects surging AI compute needs—training models like GPT-4 already demands hundreds of megawatts. The data center services market, forecasted to hit $438.3 billion by 2030, underscores this trend. South Korea’s government backs the project with permits and approvals, aiming to decentralize tech hubs from Seoul and cement its role in the AI ecosystem.

Microsoft’s Majorana-1 Chip

Microsoft’s Majorana-1 chip, unveiled on February 19, 2025, marks a quantum computing milestone after 17 years of research. It’s the first to use topological qubits powered by Majorana Zero Modes (MZMs)—quasi-particles theorized by Ettore Majorana in 1937 and now created using a new “topoconductor” material of indium arsenide and aluminum. Fabricated atom-by-atom, this material induces topological superconductivity when cooled near absolute zero and tuned with magnetic fields, hosting MZMs at nanowire ends.

Unlike traditional qubits, which suffer from noise and error rates (e.g., IBM’s superconducting qubits require extensive error correction), topological qubits leverage MZMs’ parity-based information storage—odd or even electron counts—making them inherently stable. Microsoft’s Nature paper confirms MZMs’ existence and a new microwave-based measurement technique distinguishing states with billion-electron precision. This digital control simplifies scaling, targeting one million qubits on a single chip versus competitors’ analog-heavy approaches.

The framework sets a massive advancement trajectory. DARPA selected Microsoft for the final phase of its US2QC program, signaling confidence in scaling prototypes. Potential applications include simulating molecular interactions for drug discovery or optimizing complex systems—tasks needing trillions of operations impractical with current quantum designs. While commercial deployment is years away, the Majorana-1’s stability and scalability shift quantum computing from theoretical to feasible, with Microsoft betting on topoconductors as the “transistor of the quantum age.”