Two $20B+ deals in one week (and what they tell us about 2026)

While everyone was on holiday, the AI chess board completely changed.

Already the end of the year? Hey 👋 Welcome to your first "end of year" BullCity AI. If you're reading this between Christmas and New Year's, you're exactly the kind of person I write this for. While most people were offline last week, two deals worth over $40 billion combined just redrew the AI landscape.

🔥 The Big Story: Meta and Nvidia Just Bought the Same Thing (Sort Of)

Meta acquired Manus for $2 billion. Nvidia grabbed Groq's assets for $20 billion. Different companies, different price tags, but the same strategic move: buying execution speed. Both deals closed in roughly 10 days. That's not normal M&A. That's panic buying.

Here's what actually happened: Manus went from zero to $100 million in annual recurring revenue in eight months. That's the fastest revenue ramp of any software company in history. They built AI agents that actually work - doing market research, screening job candidates, analyzing portfolios. Not demos. Actual products people pay for.

Groq built chips (LPUs - Language Processing Units) that make AI inference blazingly fast. Their tech is fundamentally different from Nvidia's GPUs: all on-chip memory, no external bandwidth bottleneck. For single-user inference, nothing is faster.

Why both companies got acquired at massive premiums: Neither was for sale. Manus was raising at a $2B valuation when Meta showed up with $2B cash. Groq raised at $6.9B in September; Nvidia paid $20B four months later. When acquirers move this fast and pay this much over valuation, they're not buying companies. They're buying time they think they don't have.

💡 What This Tells Us About 2026

The era of general-purpose AI is over. Specialization is here.

Three predictions for 2026: - More "acqui-hires" structured to avoid antitrust. Both deals were carefully framed as "licensing agreements" and "non-exclusive partnerships." Expect this to be the new playbook. - The agent explosion is real this time. Manus proved the business model works. Expect 50+ AI agent companies to launch in Q1 2026. - Hardware differentiation returns. Inference chips, edge chips, specialized accelerators - the hardware stack is fragmenting again.

📰 What Else Happened

Google's Gemini 2.0 is quietly impressive - While everyone was watching OpenAI, Google shipped Gemini 2.0 with multimodal live API, spatial understanding, and production-ready features.

OpenAI's valuation is now $300B - That's more than Intel, AMD, and Qualcomm combined.

Character AI hit 4 million daily subscribers - They're selling $9.99/month companionship to Gen Z. It's weird, it's working.

🎯 The 2026 Landscape

• Training: Still dominated by Nvidia, but alternatives building fast
• Inference: About to fragment. Different chips will win different workloads.
• Applications: The race to revenue is on. Manus proved you can build a real business fast.
• Business Models: Subscription is winning over API usage for consumer apps.

🎯 Quick Hits

• Anthropic's computer use is now in general availability
• Mistral raised at $6B valuation - European AI is alive
• Apple's on-device AI is better than expected
• Runway's Gen-3 Alpha Turbo - AI video generation crossed a threshold