Industry

AI winners may shift as bottlenecks move beyond chipmakers

The AI investment boom is likely to continue, but the companies that benefit most may change as the system’s bottlenecks shift.

AI winners may shift as bottlenecks move beyond chipmakers

Over the past two years markets focused on whether the AI revolution is real and whether Nvidia’s sharp share‑price rise is sustainable. But the history of technology revolutions suggests the biggest investment mistake is not underestimating AI’s success, but misunderstanding where that success will produce value. AI will likely transform the economy, yet it is far from certain that today’s best‑known companies will capture the greatest gains.

This analysis is based on observations by Nemesi Péter, quantitative strategy analyst at OTP Alapkezelő, and on recent market developments.

Why market reactions have become more muted

Nvidia’s latest earnings report did not deliver a major surprise: over the past twenty quarters the company beat the consensus on earnings per share nineteen times. Despite this consistency, the stock’s response has become more restrained because investors are less preoccupied with whether Nvidia can again beat expectations and more concerned with how long the investment cycle that drives growth will last.

That raises a different question: are investors asking the right question about AI? Are they looking for winners in the parts of the ecosystem that will create the most value over time?

The AI investment wave is far from over, but its shape may change

Major tech firms — Amazon, Google, Meta and Microsoft — are spending ever larger absolute amounts on data centers, chips and supporting infrastructure. Their combined AI‑infrastructure expenditures are expected to be nearly three times 2024 levels by 2026, and current forecasts foresee a new record for combined spending in 2027.

However, the first signs of a slowdown may show up not as falling absolute spending but as a deceleration in growth rates. Projections indicate that around 2028 we should not expect further acceleration, even if total investment stays above $800 billion. Historically, the transformation of an investment cycle — not just its end — often creates the biggest investment opportunities.

Nvidia was the first major AI winner

At the start of the boom the key bottleneck was compute capacity; providers of that resource naturally rose to prominence. Nvidia filled that role, building an ecosystem that currently has few real alternatives. The company’s management expects more than 12 percent quarter‑on‑quarter revenue growth in the next quarter, with adjusted gross margins around 75 percent, indicating exceptional pricing power.

Nvidia’s trailing P/E is about 35. That looks high, but it is below its ten‑year average; forward metrics nonetheless imply continued earnings growth. The main risk is therefore not strictly valuation but the tendency of investors to search for future winners among those that have already delivered top performance.

Bottlenecks migrate — memory moved into the spotlight

As models grew larger it became clear that memory and memory bandwidth, not pure compute, were the binding constraints. Insufficient high‑bandwidth memory created persistent shortages, and memory suppliers — notably SK Hynix and Micron — found themselves in pivotal roles, often with capacity booked years in advance. Memory came to represent a larger share of chip value, and profitability and pricing power shifted toward memory makers. That helps explain why over the past year Micron and SK Hynix shares rose more sharply than Nvidia’s.

Where might the next bottleneck appear?

Historical analogies (railroads, the internet, mobile) show that the greatest value often accrued to those solving the next practical constraint in the system. For AI that next constraint may not be semiconductors alone. Physical limits are increasingly relevant: power supply and cooling capacity, further advances in high‑performance memory, and robust data‑center infrastructure are all critical.

Consequently, in a few years investors may be talking more about companies building power grids, data‑center infrastructure providers, or leaders in specialized cooling techniques than about chipmakers alone.

Conclusion: the main investment risk is looking in the wrong place

The AI investment cycle remains strong and the revolution is far from over. The principal investment risk today is not that the AI boom fails, but that it succeeds while investors search for winners in the wrong places. In past technology revolutions, the largest gains went to those who recognized the true system bottlenecks in time — and accepted that those bottlenecks move over time.


Tags: earnings, artificial intelligence, technology revolution, Nvidia, memory, data center, Micron, AI, chip manufacturing, semiconductors