Nvidia–Groq and Meta–Manus deals herald 2026 AI consolidation wave

The Nvidia–Groq and Meta–Manus transactions mark a decisive phase shift in AI consolidation. Nvidia is spending around $20 billion to license Groq’s inference stack and hire away its founder and core engineers—an acqui‑hire that effectively neutralises a promising specialised‑hardware rival while absorbing its low‑latency LPU architecture.([elnacional.cat](https://www.elnacional.cat/oneconomia/es/on-ia/semana-ia-inteligencia-artificial-va-compras_1532545_102.html)) Meta’s acquisition of Manus, a Chinese‑founded autonomous agent platform, similarly folds a frontier agent system directly into Meta’s product pipeline and Meta AI ambitions.([en.wikipedia.org](https://en.wikipedia.org/wiki/Manus_%28AI_agent%29?utm_source=openai))

Strategically, this is big tech locking up two critical pieces for the next phase of the race: cheap, deterministic inference at scale (Groq) and robust, real‑world‑capable agents (Manus). These aren’t just talent plays—they’re bets on where future bottlenecks lie as models become more “thinking” intensive and as agentic systems orchestrate complex workflows. The ElNacional piece also notes that SpaceX, OpenAI and Anthropic are positioning for gigantic IPOs, suggesting public markets are about to be directly exposed to the economics of frontier AI.([elnacional.cat](https://www.elnacional.cat/oneconomia/es/on-ia/semana-ia-inteligencia-artificial-va-compras_1532545_102.html))

All of this points toward further centralisation of key capabilities in a handful of US‑centric platforms, backed by immense war chests. That likely accelerates the frontier but raises serious questions about competition, geopolitical leverage and the bargaining power of smaller labs and open‑source communities.