AI: Bubble about to burst or entering maturity?

AI - Ảnh 1. — The AI craze may have peaked

Artificial intelligence (AI) is considered the "new industrial revolution" with the expectation of creating a trillion-dollar boost. However, after a period of speculative boom and record revenue reports, the market is sending clear signs of cooling down.

From the decline of major tech stocks, the drop in GPU costs, to a series of studies showing that 95% of AI projects have not yet brought financial benefits, all are making observers ask: Is AI past the "bubble peak"?

AI: From the hope of a "new industrial revolution" to the harsh reality

In just two years, AI has been elevated to the level of a "new industrial revolution", attracting a series of huge investments from technology corporations, venture capital funds and even governments .

Nvidia continues to post record revenue thanks to booming demand for AI chips, Microsoft and Google spend tens of billions of dollars to expand data centers, while Meta bets its future on AI as a key growth engine.

However, this euphoria is showing a downside. The stock market reacts negatively even when companies “beat expectations,” suggesting that investors’ expectations have been pushed too high.

Nvidia is a case in point: revenue soared, but shares fell after earnings because the market assumed the outlook was less than stellar.

"AI-based" stocks such as AMD and Super Micro also adjusted sharply when profits were not attractive enough, signaling that investors were becoming more cautious.

Huge infrastructure investments and the question of real efficiency

Behind the optimism are still huge spending figures. Microsoft plans to spend more than $80 billion on AI infrastructure in fiscal year 2025, a figure equivalent to the defense spending of many countries.

Google and Meta are not far behind, constantly announcing plans to build new generation data centers. This proves that long-term faith in AI has not diminished, but also raises a difficult question: will such huge amounts of money bring commensurate profits?

In fact, the cost of operating and building AI infrastructure is increasingly weighing on the profit margins of the "big guys". That is why many technology companies are simultaneously cutting staff and restructuring to offset investment costs.

A paradox exists: AI is seen as a cost-optimization tool for businesses, but the development of AI itself is causing corporations to spend more than ever.

AI - Ảnh 2. — AI is gradually "cooling down" in the market

GPU costs cool, supply is less stressed

One of the clearest signs that the AI craze is cooling is the price of AI chips. Just a year ago, renting an Nvidia H100 chip in the cloud cost as much as $8 an hour, making it difficult for many startups to sustain their experimental models. Now, that price has dropped to around $2.80 to $3.50 an hour.

The reason comes from the fact that hardware supply has begun to stabilize, competition between suppliers has increased, and businesses are forced to optimize GPU usage instead of "spending money" like before. This is both a positive signal for the maturity of the market and a reflection that "demand" is no longer as hot as the peak period of 2023-2024.

95% of GenAI projects fail and the investment paradox

A shocking study from MIT found that 95% of Generative AI pilot projects do not yield clear financial benefits. Many businesses deploy AI without specific KPIs, resulting in results that cannot be measured or integrated into production and business processes. This does not mean that AI is useless, but it exposes a huge gap between expectations and reality.

At the same time, the capital market also reflects an alarming imbalance. In the first quarter of 2025 alone, the AI sector attracted $104 billion in venture capital, but only $36 billion in divestments. This means that money is still pouring in like a waterfall, while the exit (IPO, M&A) is unclear. This situation is very similar to previous technology bubbles, where investment capital far exceeded short-term value creation.

AI - Ảnh 3. — The AI market is saturated

While the US and Europe focus on technological competition, China faces the risk of “building more than it needs.” A series of AI data centers are deployed according to policy directions, even without specific tenants.

Alibaba has warned of an “infrastructure bubble” as supply exceeds demand. This “planned” development model allows China to catch up quickly in technology, but can also create costly overcapacity if actual demand does not grow fast enough.

All of these signs suggest that the AI craze is cooling. But this is not a bubble that is about to burst, like the dotcom bubble of 2000. The big difference is that there is a real need for AI, the global digital infrastructure needs upgrading, and businesses are still looking for ways to use the technology to increase productivity.

It’s all about speed and efficiency. The market is entering a “filtering” phase, where only companies with sustainable business models and clear ROI will survive. “Flashy” startups without foundations will gradually disappear, while AI will continue to evolve, but at a more realistic pace.

AI is no longer a “fever” that defies all, but is gradually becoming a mature industry where every expenditure is closely scrutinized. Investors, businesses and governments have moved beyond the dream stage to enter the more difficult problem: how to turn this technology into real value.

Back to topic

PHAN HAI DANG

Source: https://tuoitre.vn/ai-bong-bong-sap-vo-hay-buoc-vao-giai-doan-truong-thanh-20250903103854784.htm