Daniel Mercer
Share
Listen On

The situation in global markets has become especially interesting. U.S. tech giants have posted extraordinary gains: Nvidia’s market capitalization has risen roughly 14x since late 2022, reaching $5 trillion, while average returns across tech-sector strategies have topped 150% over the past three years.

AI-focused IPOs have been another standout trend. Over the last two years, more than 50 AI companies have gone public, and many of them doubled or even tripled in value early on (Astera Labs: from $36 to $165.5; Reddit: from $34 to $194.5). That naturally raises the question: are we looking at the next bubble?

In my view, this is not a classic speculative bubble so much as the early construction of a new economic foundation. The key difference in today’s AI wave is the low barrier to entry and the broad applicability of the technology. But the real “battle” isn’t over yet another chat interface—it’s about controlling the infrastructure: data centers, compute capacity, cloud platforms, and the underlying LLM stack.

That is why capital spending in this space is so massive. The gold-rush analogy fits well: the biggest winners are often not the prospectors, but the “shovel sellers”—the builders of data centers, chipmakers, and infrastructure software providers (Oracle, Nvidia, AMD).

For investors, the most practical approach right now is a barbell portfolio: 80% allocated to the top 7 tech incumbents for stability, and 20% to higher-upside names—pre-IPO or newly public companies positioned to benefit from the next leg of AI adoption (Anthropic, Glenn, Scale). This structure helps balance risk with exposure to the fastest-growing segment of the modern economy.

One more point matters: even though AI already accounts for roughly 50% of global direct investment flows, its full impact is only beginning to translate into public-market repricing. Over the next 12–24 months, we will likely see an even deeper reshaping of capital markets driven by AI—and this is the window to clarify an investment strategy rather than chase momentum later.

AI Research Contributor
Daniel Mercer is an AI research contributor specializing in large language models, benchmarking, and multimodal systems. He writes about model capabilities, limitations, and real-world performance across leading AI assistants and platforms.

Recent Podcasts

AI as a Role Model for Generation Alpha: Promise, Risks, and the Future of Childhood

Artificial intelligence is becoming the main role model for Generation Alpha. 2026 may mark a...

AI as a Toy: Why Humanity Always Misuses New Technology First

Artificial intelligence could, in theory, help solve all of humanity’s problems. Stop wars, cure...

Why the AI Boom Is Not a Bubble but the Foundation of a New Economy

The situation in global markets has become especially interesting. U.S. tech giants have posted...