AI dealmaking closes 2025 with a bang

Meta’s purchase of Singapore-based Manus, Nvidia’s $20 billion Groq deal, and SoftBank’s completion of its $22.5 billion investment in OpenAI have turned the final weeks of 2025 into a showcase of how aggressively capital is repositioning around AI infrastructure, agents, and foundation models.

Together, these moves signal that AI dealmaking is re-accelerating into year-end and laying the groundwork for the next leg of the AI bull run in 2026.​

Key late-2025 AI deals

• Meta acquired Manus, a Singapore-based AI agent startup, for more than US$2 billion, adding a production-ready “digital employee” platform, meaningful subscription revenue, and deep agent talent to Meta’s AI stack.​

• Nvidia struck a roughly US$20 billion deal around Groq’s AI inference technology and assets, securing its low-latency LPU architecture and key engineers to reinforce Nvidia’s dominance as AI shifts from training to real-time deployment.​

• SoftBank closed an additional US$22.5 billion investment into OpenAI on December 26, bringing its total commitment under a March pledge to US$40 billion and taking its stake to about 11%, tightly aligning SoftBank with one of the core model providers in the ecosystem.​

What this means for 2026

• Acceleration of AI infrastructure: Capital is concentrating in chips, data centers, and foundation models, with late-December deals pushing AI-related transactions toward near-record levels globally.​

• Validation of AI agents and inference: Meta–Manus and Nvidia–Groq show that agents and ultra-low-latency inference are now strategic battlegrounds, not side bets.​

• Setup for the next AI bull leg: With balance sheets committed, platforms integrated, and strategic stakes locked in, these year-end deals position 2026 for a renewed AI upcycle as deployment, monetisation, and real-world use cases catch up to 2025’s infrastructure build-out.​​