AI Investment: Beyond the Bubble to 2026 Inference Phase

No AI Bubble, Just R&D

• 💡 Beth Kindig states there is no AI bubble, but rather a phase of Research & Development where companies are investing heavily in creating the technology.
• 🎯 The next crucial phase for investors is the inference and monetization stage, expected to begin in 2026, leading to recurring revenue and profits.

Muted Multiples and Nvidia's Value

• 📈 Ivana Delevska argues that current multiples are muted and on the lower end of historical ranges, even for companies like Nvidia, which are seen as AI proxies.
• 💰 Nvidia is considered very cheap due to its significant earnings growth, contrary to typical bubble behavior where multiples expand ahead of fundamentals.

Google vs. OpenAI and Monetization

• 🚀 Both Google and OpenAI are seen as winners, especially as applications built on their large language models proliferate.
• 💰 OpenAI is noted for its rapid monetization trajectory, reaching $20 billion quickly and predicting hundreds of billions by the end of the decade.

Chip Specialization and Power Constraints

• 🧠 Nvidia is expected to maintain leadership in general-purpose computing, while specialized chips like Google's TPUs may suit specific use cases like advertising.
• ⚡ The AI race is now power-constrained, with companies like Bloom Energy standing out for providing on-site power solutions that don't rely on the grid.
• 💡 Bloom Energy's solid oxide fuel cells can energize GPUs in months, addressing the bottleneck of power availability for AI infrastructure.

Understanding Inference

• 🔍 The training phase of AI is compute-intensive and not meant for monetization.
• ✅ Inference is the phase where applications and large language models run within those applications, enabling monetization and recurring revenue.