My Investing in AI Book Chapter 2: AI Isn't Solved Yet

Key Takeaways

• AI adoption now a permanent plateau across industries
• Hallucinations and black‑box nature limit enterprise use cases
• High inference costs push firms toward smaller, specialized models
• Infrastructure bottlenecks slow large‑scale AI deployment

Pulse Analysis

AI’s current surge resembles a new plateau rather than a fleeting boom. After a series of historic AI winters, today’s large language models are embedded in everything from customer support chatbots to content creation tools, creating a baseline of commercial relevance. This pervasive integration means that, unlike past cycles, a complete lull is unlikely; instead, the market is poised for steady, incremental growth as companies fine‑tune existing models and explore niche applications. For investors, the takeaway is clear: the sector’s valuation will be driven less by speculative AGI timelines and more by tangible productivity gains across verticals.

Technical limitations remain the primary friction points. Hallucinations—fabricated but plausible outputs—continue to erode trust in high‑stakes environments such as legal or medical advice. The black‑box nature of transformer architectures hampers regulatory compliance, especially in finance where explainability is mandatory. Moreover, inference costs have emerged as a hidden expense; running large models can consume megawatts of power and cost upwards of $10 million annually for major providers, prompting enterprises to adopt smaller, task‑specific models that balance performance with budget. Infrastructure gaps, from API throttling to the need for sophisticated routing and security layers, further slow large‑scale rollouts, reinforcing the importance of ecosystem‑level solutions.

Looking ahead, AI will likely evolve through a series of iterative upgrades—AI‑2.0, AI‑3.0—rather than a single breakthrough to artificial general intelligence. Companies that build modular model stacks, leverage emerging architectures like hierarchical temporal memory or state‑space models, and invest in robust data pipelines will capture disproportionate value. Investors should therefore prioritize firms with flexible AI stacks, clear cost‑management strategies, and a track record of translating marginal model improvements into measurable ROI. By focusing on these fundamentals, capital can be allocated to the next wave of AI‑enabled productivity rather than chasing the elusive AGI myth.