Key Variables Determining the AI Bull Market

Article by Zhao Ying

Source: Wall Street Journal

Oil prices have risen above $100 per barrel, the Strait of Hormuz has not yet resumed normal operations, inflation and interest rate pressures are resurfacing, and expectations for Fed rate cuts have become more fragile. According to traditional macro frameworks, this is not an ideal environment for high-valuation tech stocks. Yet, U.S. equities have reached new highs, and AI-related sectors continue to attract strong capital inflows.

On May 25, Song Xuetao, macro analyst at Guojin Securities, noted in a research report: "The current AI market is in a phase of rational exuberance—bubbles have emerged but are not out of control." The key point is not the "bubble," but the "rational" exuberance: Agentic AI is transitioning from an auxiliary tool to an autonomous execution tool, enabling the market to clearly see, for the first time, the business闭环 from "burning cash" to "generating profits."

On the rational side, the proliferation of agent applications has driven increased token consumption, demand for reasoning compute power, and rapid growth in ARR for leading vendors. On the speculative side, valuations have already priced in growth expectations for 2027–2028. As of May 20, the forward P/E ratio of the seven major U.S. tech stocks is approximately 35x, compared to about 25x for the remaining 493 companies in the S&P 500. This premium implies not just typical growth stock logic, but an AI adoption rate five to eight times faster than previous technological revolutions.

What truly determines whether the AI bull market can continue is not a single quarter’s performance or any one breakout application, but three key variables: in the short term, liquidity shocks—particularly from oil prices, inflation, interest rates, and the unwinding of yen carry trades; in the medium term, industry execution—whether the pace of AI adoption can match current valuations; and in the long term, more fundamental constraints such as energy, power grids, employment, social resistance, and breakthroughs in hardware technology.

In the previous AI trading cycle, the market's biggest concern was that giants were spending too quickly—investing heavily in data centers, GPUs, and cloud infrastructure—yet the path to revenue generation remained unclear. The shift with Agentic AI is that it is no longer merely a Copilot-style assistant, but is evolving into an Autopilot-style autonomous execution tool.

This has resulted in two outcomes.

First, token consumption is accelerating again. The first wave of demand after GPT’s emergence came from improved model capabilities; the second wave, following the adoption of agents, stems from a surge in inference computing power. Autonomous task execution means longer contexts, more complex steps, and more frequent model calls—inferring is no longer a byproduct of training but has become the primary arena for sustained computational consumption.

Second, revenue expectations have been raised. Following the widespread adoption of representative agent applications such as OpenClaw and Claude Cowork, model providers have seen rapid growth in their annual recurring revenue (ARR). Mid-year projections cited in the material show that Anthropic’s full-year ARR forecast has been increased from $9 billion at the start of the year to $44 billion, doubling approximately every six weeks. If this trend continues, ARR could exceed $300 billion next year.

This explains why the market no longer simply penalizes capex. As long as revenue growth is fast enough, capital expenditures shift from a burden to a moat. NVIDIA, Broadcom, and the hardware chain—including optical modules and storage—have thus regained support.

This round of AI assets rising against oil prices is not due to the disappearance of macro risks, but because several forces have temporarily outweighed the risks.

First, demand is spreading across the industrial chain. During the inference phase, not only GPUs but also CPUs, optical modules, and storage are being pulled into the high-growth narrative. 800G/1.6T optical modules are in high demand, and demand for high-end storage is rising. Light Counting forecasts that 800G transceiver shipments will more than double by 2026, while 1.6T port shipments will grow from a small base in 2025 to tens of millions, with 1.6T chipset sales exceeding $2 billion in 2026 and maintaining high growth rates over the next three years.

Second, the earnings of tech giants have been exceptionally strong. In the first quarter, the S&P 500’s EPS growth rate reached approximately 27.1%, the highest since Q4 2021, with Meta, Alphabet, and Amazon accounting for 70% of the index’s earnings growth. As long as these heavyweight companies continue to generate profits, the downward pressure from oil price shocks on the index will be delayed.

Third, U.S. growth is becoming increasingly dependent on AI infrastructure. Over the past several quarters, investment in AI infrastructure has contributed more than half of U.S. GDP growth. While non-farm payrolls, retail sales, and other aggregate data remain decent, and despite emerging divisions in employment structure, markets are unlikely to shift immediately to stagflation trading until there is clear evidence of a broad weakening in overall economic activity.

Another more direct factor: large tech companies are less sensitive to oil prices than industries such as aviation, express delivery, rail, chemicals, automotive, and tourism. They are more concerned about electricity prices than oil prices. When traditional real economy sectors are pressured by rising oil prices, capital tends to flock more readily to AI assets, blending "safe-haven" trades with growth trades.

The danger of AI market trends lies not in the lack of industrial support, but in the market pricing too quickly.

The seven major U.S. tech stocks trade at a forward P/E ratio of 35x, while the remaining 493 companies in the S&P 500 trade at 25x. This valuation gap implies a highly seamless future: over the next 3 to 5 years, AI infrastructure will continue expanding, with sustained high demand for computing power, cloud services, data centers, and semiconductors; AI will progressively penetrate scenarios such as advertising, search, cloud services, office software, code generation, financial risk management, customer service, investment research, and content creation; and both revenue contributions and efficiency gains will materialize simultaneously.

But technological revolutions rarely proceed so smoothly. It took about 40 years for electricity to move from invention to widespread adoption in production lines, and about 25 years for computers. Now, the market’s pricing of AI’s diffusion implies it must be 5 to 8 times faster than these general-purpose technologies.

It’s not impossible, but the margin for error is thin. If AI commercialization lags behind capital expenditures, inference demand fails to keep up with training demand, or depreciation and electricity costs begin to erode profit margins, valuations will react first. Being on the right industry trajectory doesn’t mean stock prices can advance indefinitely in advance.

The real short-term pressure comes from liquidity.

If the Strait of Hormuz remains closed for an extended period, and oil prices stay above $100 or continue rising, inflation will spread from energy prices to services, transportation, and raw materials. In April, the U.S. PPI rose year-over-year to 9.8%, the highest level since October 2022. Once inflation becomes entrenched, the Fed’s policy path will be forced to be rewritten.

The swap market has priced in 0.8 rate hikes by the Fed this year, and more than two rate hikes by the ECB and the Bank of England. Meanwhile, concerns over policy independence due to Fed leadership changes and increasing internal divisions within the FOMC are undermining market confidence in future easing.

Japan is also a gray rhino. Japan has long been a global funding source for leveraged trades, but yen depreciation and inflationary pressures have forced the Bank of Japan to signal a tightening stance, with 30-year Japanese bond yields rising above 4%. If Japan’s financing costs continue to rise and trigger unwinding of global carry trades, highly valued AI assets will find it difficult to remain unaffected.

A rehearsal occurred on May 15: the 10-year U.S. Treasury yield breached 4.5%, the 30-year yield surpassed 5%, high-conviction momentum trades cooled, the Philadelphia Semiconductor Index dropped about 4% in a single day, and the Nasdaq fell approximately 1.5%. This is not evidence of a trend reversal, but it demonstrates that crowded trades are highly sensitive to interest rates.

The most critical short-term comparison is simple: Can the upward revision rate of ARR (Annual Recurring Revenue) outpace the rise in interest rates? If not, capital may first retreat to the more certain hardware segment; if liquidity continues to deteriorate and AI revenue expectations fail to keep rising, valuation pressures will become significantly more pronounced.

The mid-term test is industry realization. General-purpose technological revolutions typically do not follow a straight upward path, but rather follow a pattern of “accelerate, decelerate, then accelerate again.” First comes a wave of capital, then organizational alignment, and finally the release of productivity. The early internet also experienced investment frenzies, expanded capital expenditures, and asset bubbles—true productivity improvements emerged gradually only years later.

The challenge with AI pricing lies in the fact that it almost requires businesses to rapidly adapt their organizational structures, workers to quickly retrain, business models to be swiftly validated, and no strong societal resistance to emerge. Such speed is uncommon in human history.

Long-term constraints are stricter.

First is energy and infrastructure. AI data centers require massive amounts of electricity and cooling water—grid expansion, transformers, and energy storage are not hypothetical variables on a PowerPoint slide, but real bottlenecks. If AI infrastructure continues to drive up electricity costs across society, regulatory and social pushback will intensify.

Second is employment and consumption. In the short term, AI can boost corporate efficiency and reduce demand for roles such as engineers and customer service staff; however, if technological unemployment outpaces the creation of new jobs, household purchasing power will be weakened. Ultimately, efficiency gains on the B-side still depend on consumer spending on the C-side—if non-AI sectors enter a downturn, AI alone will struggle to remain dominant in the long term.

Third is social acceptance. At the beginning of the year, there was a nationwide surge in China to install Openclaw, but public resistance in the U.S. to data centers driving up electricity prices and technological unemployment is growing. This will impact the pace of AI adoption.

Fourth, a hardware technology breakthrough. If an engineering milestone akin to the "DeepSeek moment" occurs, significantly boosting computing power, storage, and transmission efficiency, today’s most constrained hardware components could suddenly become oversupplied. The high-growth logic of the hardware supply chain is not invincible.

The long-term outlook for the AI industry remains optimistic. Ignoring social tensions arising from technological unemployment and the restructuring of production relations, AI indeed has the potential to enhance total factor productivity and help the economy overcome stagflation pressures. Even if financial markets undergo deleveraging along the way, the remaining data centers, low-cost technologies, and validated application scenarios could serve as the foundation for the next wave of industrial expansion.

But stock pricing is not the industry vision itself. What this AI bull market most needs to validate is whether the market’s current bets on ARR, ROI, and the speed of technological adoption can still be realized amid rising oil prices, inflation, higher interest rates, and tighter social constraints. Getting the direction right explains why a bull market exists; the pace of execution determines whether the bubble will spiral out of control.