AI Supercycle 2026: High Valuations, Low Revenues, and Massive Capex - Where Is the Return?

In 1990, Apple spun out its internal project for a “Personal Intelligent Communicator” into a separate company called General Magic. It was staffed by some of the best talent of the time - people who had worked on the original Macintosh. The company developed a device with a touchscreen, intelligent agents, and a closed communication network, which was essentially the first true concept for a smartphone. The devices launched in 1994 but failed commercially. The project was simply too far ahead of its time - the internet was not yet in mass use, people were not accustomed to constant communication on the go, and the infrastructure and consumer habits were not ready. General Magic remains in history as an example of a major technological invention that appeared too early and failed to take hold, even though many of its ideas were later realized in the iPhone and modern smartphones.

Today, as we witness extremely large investments in artificial intelligence, record valuations of companies, and ambitious promises to transform the entire way we work and live, the question arises whether we are once again witnessing a similar situation. Is it possible that part of the current AI supercycle in 2026 is too early, just like General Magic was in the 1990s? In this article, we examine the real state of the AI industry in 2026 - the massive capital expenditures, still relatively low revenues compared to valuations, the role of Chinese competition, the political moves of leading companies, and the risk of disappointment if visible returns do not appear quickly enough.

In the first half of 2026, global venture capital hit a new record

In the first half of 2026, global venture capital reached a new record - over $510 billion. The majority of these funds flowed into artificial intelligence. OpenAI closed a historic $122 billion round at a valuation of $852 billion, while Anthropic is moving toward levels close to $965 billion. Capital is concentrating in an extremely narrow group of leading labs and related infrastructure players with unprecedented intensity.

At the same time, the revenues of leading companies remain dozens of times smaller than their market valuations, while capital expenditures on chips, data centers, and energy continue to grow rapidly. This is the central paradox of the current moment.

Why is this happening?

We are in the middle of a genuine technological platform shift. Foundation models are changing how value is created across almost every industry - on a scale and depth comparable to the emergence of the internet between 1995 and 2005 or smartphones after 2007. Investors and strategic corporations (Microsoft, Amazon, Google, NVIDIA, SoftBank) are betting that these models will become foundational infrastructure for the next two to three decades.

This is precisely why capital is flowing in massively and early. In a regime of highly uneven return distribution - and artificial intelligence is a classic example of such a regime - a significant portion of profits concentrates in a very small number of positions. This explains why over half a trillion dollars is flowing to such a limited number of companies.

High Valuations, Low Revenues, and Massive Capex

This is where the most important question lies.

OpenAI has an annual recurring revenue run-rate in the range of $20-25 billion as of mid-2026. At a valuation of $852 billion, this implies a multiple in the range of 35-40x revenue. For comparison, mature technology companies typically trade in the 5-15x range. At the same time, the company and its partners are taking on infrastructure commitments worth hundreds of billions of dollars, and in some scenarios, significantly over one trillion dollars cumulatively by 2030. Capital expenditures on chips, electricity, and data centers are massive, while gross margins remain under pressure due to high costs of generating responses.

A similar picture is observed among the other leading players. Revenues are growing quickly, but they are still far from the levels that could justify current valuations based on traditional return metrics.

Table 1: Key Financial Metrics of Leading AI Companies (July 2026)

Where Is the Return on This Capex?

This is the question investors should be asking with the greatest seriousness.

In the short term (2026-2028), the main source of value comes from increasing the productivity of existing employees, rather than from mass job replacement. Artificial intelligence is already being used effectively in programming, data analysis, customer service, legal analysis, marketing, and research and development. Companies that implement it purposefully report higher employee productivity. For now, the process is more about augmenting human labor than pure replacement.

In the medium term, returns may come from new categories of value - autonomous AI agents capable of executing complex multi-step tasks, and physical artificial intelligence (robotics and autonomous systems). If these areas develop successfully, they can generate entirely new markets and higher margins.

In the long term, if artificial intelligence truly becomes foundational technology on the scale of electricity or the internet, returns will manifest through overall economic growth. However, this remains a hypothesis, not a guaranteed outcome.

The Chinese Factor - Price Reality

The picture is further complicated by Chinese competition.

Leading U.S. models (GPT-4o and the o-series, Claude Sonnet and Opus) charge tokens in the range of $2.50-5+ per million input tokens, with significantly higher prices for output tokens in the most advanced models.

Chinese models (DeepSeek V3/V4, Qwen3 series, Kimi, and others) offer comparable or very close performance on many tasks at prices that are 5 to 20 times lower - often in the range of $0.10-0.50 per million tokens for strong models. In some tasks, the difference reaches 30-50 times in favor of Chinese solutions.

This creates serious price pressure. If enterprise clients begin using Chinese models for more routine tasks at scale, monetization for leading U.S. labs becomes significantly more difficult.

Table 2: Token Price Comparison - U.S. vs Chinese Models (July 2026)

OpenAI and 5% for the U.S. Government

In early July 2026, information emerged that Sam Altman and OpenAI had discussed the possibility of giving approximately 5% of the company to the U.S. government - an amount that, at the current valuation, would be worth roughly $42-43 billion.

This is not merely a gesture of goodwill. The more likely motives are several: political insurance against heavy regulation, an attempt to share benefits with society to reduce political pressure, and a pragmatic move ahead of a potential IPO. Whether this is “voluntary extortion” or a reasonable strategic step remains debatable - most likely a combination of both.

The Cycle Trap: The Primary Risk for Investors

The artificial intelligence cycle in 2026 is real. The technology is advancing rapidly and has the potential to transform productivity across multiple sectors. At the same time, the economics behind it are still immature: revenues are growing but remain small relative to valuations and capital expenditures; returns are uncertain and spread over time; Chinese competition exerts serious price pressure. In this context, OpenAI’s proposal to give 5% to the government appears as a pragmatic political and strategic move - a combination of insurance and an attempt to legitimize the value being created.

Concentrated positions in infrastructure with real scarcity, in autonomous AI agents and physical artificial intelligence with proven defensibility, and in companies that demonstrate a clear path to better capital efficiency - these are the areas where returns have the highest chance of materializing. Everything else represents an expensive experiment with an unclear outcome.

At the same time, another less discussed process is unfolding - a growing divergence between the public messaging of leading CEOs and visible results in the real economy. Over the past two to three years, several key figures in the industry have repeatedly described a future in which people “will not need to work,” in which labor will lose meaning because the abundance generated by artificial intelligence will create such wealth that some form of universal basic income or radical redistribution will be necessary. These messages were actively picked up and amplified by media and social networks, creating widespread euphoria among parts of the public and investors - expectations that “everyone will get rich from stocks,” that life will become significantly easier, and that the transformation will happen quickly and visibly.

At the same time, companies have consumed unprecedented amounts of tokens and computing power for training and generating responses, but the broader economy still does not show a dramatic acceleration in productivity that matches these promises. Results remain concentrated in a narrow circle of companies and specific tasks, while for large parts of business and employees, artificial intelligence still functions primarily as a tool for partial acceleration rather than fundamental change.

When this gap between rhetoric and reality becomes more visible, euphoria can turn into disappointment relatively quickly. Investors begin looking for evidence of real returns, not just potential. The problem is that opportunities for capital migration are structurally limited. Technology companies with the largest market capitalizations, and especially AI leaders, occupy a dominant share in major indices and exchange-traded funds. Passive investing has directed trillions of dollars precisely into these positions. If disappointment over the lack of visible results deepens, any serious rotation out of these assets becomes extremely difficult - large-scale capital withdrawal would have a significant impact on the very indices and ETFs in which millions of individual and institutional investors are invested. This creates a kind of “liquidity trap”: large positions that seem easy to hold in conditions of euphoria but become very difficult to exit without substantial price impact.

In such a scenario, it is possible to see either prolonged stagnation of valuations despite revenue growth, or a sharper correction if the market decides that expectations were overly optimistic. Alternative destinations for capital - value stocks, international markets, commodities, certain segments of private credit, or real assets - exist, but they are significantly smaller in capacity and cannot easily absorb the trillions of dollars currently concentrated in technology giants through passive investment flows. This makes the entire market more vulnerable to a potential shift in the dominant narrative around artificial intelligence than it appears at first glance.

When the First Become Last

General Magic showed that even the strongest technological vision can fail if the market and infrastructure are not ready. Today, there is a real risk that the AI supercycle could find itself in a similar situation - with massive investments and high expectations, but still unclear and delayed returns. The question is not whether artificial intelligence will change the world, but whether the current wave of investments and promises will prove to be too early, just as General Magic was more than thirty years ago.

Frequently Asked Questions (FAQ)

1. Why do companies continue to invest so heavily in infrastructure when their revenues are still relatively low?

Leading labs and their strategic partners are betting on a long-term scenario in which artificial intelligence will become foundational infrastructure. They accept that high capital expenditures today will be justified by significantly higher revenues and margins in the next 5-7 years, if new applications develop and higher efficiency in capacity utilization is achieved.

2. What exactly does OpenAI’s proposal to give 5% of the company to the U.S. government represent?

It is a conceptual idea according to which the U.S. government could receive approximately 5% stake in OpenAI through a special public fund. The goal is to share part of the economic benefits with society and reduce political pressure.

3. Do Chinese models really pose a serious threat to U.S. leaders?

Yes, especially in terms of price. Chinese models offer highly competitive performance at significantly lower token prices. This creates strong price pressure on the enterprise segment.

4. When can real returns on massive AI investments be expected?

In the short term (until 2028), returns will come primarily through increased productivity of existing employees. More significant returns are expected in the medium term through the development of autonomous AI agents and physical artificial intelligence.

5. What could happen if visible productivity in the economy does not accelerate significantly over the next 2-3 years?

In that case, disappointment among investors and the broader public could emerge. This could lead to more selective investing, lower valuations for companies without clear returns, and a potential correction in the segment, especially given the high concentration of technology companies in indices.

Important notice

This article is for information and education only. It is not financial advice, an investment recommendation or professional advice of any kind. The views expressed are the author’s personal opinions and are not recommendations to buy, sell or hold securities, assets or investments. Investing involves substantial risk, including the possible loss of all capital. Past performance is not a guarantee of future results. Each reader should conduct independent research, consult a qualified financial adviser and weigh financial circumstances, risk tolerance and investment objectives before taking any decision. The author and borsite.net accept no liability for any loss arising from the use of the information in this article.