What Are the Key Variables Determining the AI Bull Market?
TechFlow Selected TechFlow Selected
What Are the Key Variables Determining the AI Bull Market?
In the short term, liquidity shocks; in the medium term, industry implementation progress; in the long term, harder constraints such as energy, power grids, employment, social resistance, and disruptive hardware technology advancements.
Navigating Web3 tides with focused insightsBy: Zhao Ying
Source: WallStreetCN
Oil prices stand above $100 per barrel; the Strait of Hormuz remains closed; inflation and interest-rate pressures are resurfacing; and market expectations for Federal Reserve rate cuts have grown more fragile. Under traditional macroeconomic frameworks, this is not an ideal environment for high-valuation tech stocks—yet U.S. equities hit new highs, and the AI chain continues attracting capital.
Song Xue Tao, macro analyst at Guojin Securities, noted in a research report dated May 25: “The current AI rally is in a phase of ‘rational euphoria’—a bubble has emerged, but it remains under control.” The key phrase here is not “bubble,” but “rational” euphoria: Agentic AI is evolving from an assistive tool into an autonomous execution tool, allowing markets—for the first time—to see more clearly AI’s commercial closed loop, shifting from “burning cash” to “generating revenue.”
The “rational” side lies in the surge in token consumption, inference compute demand, and rapid growth in annual recurring revenue (ARR) among leading vendors driven by Agent application adoption. The “euphoric” side lies in valuations having already priced in growth expectations for 2027–2028. As of May 20, forward P/E ratios for the U.S. “Magnificent Seven” stood at ~35x, versus ~25x for the remaining 493 companies in the S&P 500. This valuation premium implies not ordinary growth-stock logic—but rather that AI’s penetration speed must reach five to eight times that of prior technological revolutions.
What truly determines whether the AI bull market can sustain itself is not a single quarter’s earnings nor any one breakout application, but three variables: In the short term, liquidity shocks—especially oil prices, inflation, interest rates, and unwinding of yen-carry trades; in the medium term, industry delivery—whether AI’s penetration speed can match current valuations; and in the long term, harder constraints—including energy supply, grid capacity, employment dynamics, social resistance, and hardware technology breakthroughs.
From “Co-Pilot” to “Auto-Pilot”: Markets Begin Rewarding Capex
In the previous AI investment cycle, markets worried most about hyperscalers spending too quickly—massive investments in data centers, GPUs, and cloud infrastructure lacked clear revenue-recapture paths. Agentic AI changes this: it is no longer just a Copilot-style assistive tool, but evolving toward an Autopilot-style autonomous execution tool.
This shift yields two outcomes.
First, token consumption is accelerating again. The first wave of demand post-GPT stemmed from improved model capabilities; the second wave—driven by Agent deployment—is fueled by an explosion in inference compute demand. Autonomous task execution entails longer contexts, more complex steps, and more frequent model calls—making inference no longer a marginal afterthought following training, but instead the primary, sustained consumer of compute resources.
Second, revenue expectations are being revised upward. Following the diffusion of representative Agent applications such as Openclaw and Claude Cowork, model vendors’ annual recurring revenue (ARR) has surged rapidly. Mid-year estimates cited in this article show Anthropic’s full-year ARR forecast has been raised from $9 billion at the start of the year to $44 billion—doubling roughly every six weeks. If this trend continues, its ARR could exceed $30 billion next year.
This explains why markets no longer simply penalize capex. As long as revenue growth remains sufficiently strong, capital expenditure transforms from a cost burden into a moat. NVIDIA, Broadcom, and hardware chains—including optical modules and memory—thus regain support.
Why Are AI Assets Rising Amid Oil Prices Above $100?
This AI rally against rising oil prices does not reflect vanishing macro risks—but rather several forces temporarily outweighing them.
First, demand diffusion across the supply chain. The inference phase requires not only GPUs but also CPUs, optical modules, and memory—all pulled into a high-visibility, high-demand narrative. 800G/1.6T optical modules are in tight supply; demand for high-end memory is rising. Light Counting forecasts that shipments of 800G transceivers will more than double by 2026, while 1.6T port shipments will grow from a small base in 2025 to tens of millions in 2026. Chipset sales for 1.6T are projected to exceed $2 billion in 2026—and maintain high growth over the next three years.
Second, mega-cap tech earnings remain exceptionally strong. Q1 S&P 500 EPS growth came in at ~27.1%, the highest since Q4 2021. Meta, Alphabet, and Amazon alone contributed 70% of the index’s earnings growth. So long as these heavyweight constituents continue delivering profits, the downward pressure from oil prices on the broader index gets deferred.
Third, U.S. growth has become increasingly dependent on AI infrastructure investment. Over recent quarters, AI infrastructure investment has accounted for over half of U.S. GDP growth. Aggregate indicators—including nonfarm payrolls and retail sales—remain broadly stable. Though labor-market structure is already diverging, the absence of a clear broad-based weakening makes it difficult for markets to pivot immediately to stagflation trades.
A fourth, more direct factor: Large tech firms are far less sensitive to oil prices than industries like aviation, logistics, rail, chemicals, autos, and tourism. They fear electricity prices—not oil prices. When traditional real-economy sectors feel squeezed by oil, capital tends to flock to AI assets—blending “safe-haven” and growth trades.
Valuations Have Already Pre-Consumed the Good Years of 2027–2028
The danger in the AI rally lies not in lack of industrial fundamentals—but in excessively rapid market pricing.
The Magnificent Seven trade at ~35x forward P/E, while the remaining 493 S&P 500 firms trade at ~25x. This valuation gap implicitly assumes an unusually smooth future: AI infrastructure expansion continuing over the next 3–5 years; sustained high demand for compute, cloud, data centers, and semiconductors; continuous AI penetration across advertising, search, cloud services, office software, code generation, financial risk management, customer service, investment research, and content creation; and simultaneous realization of both revenue contribution and efficiency gains.
Yet technological revolutions rarely unfold so smoothly. Electricity took ~40 years from invention to widespread adoption in assembly lines; computers took ~25 years. Today’s market pricing demands AI’s diffusion speed be five to eight times faster than those general-purpose technologies.
That is not impossible—but margin for error is razor-thin. Should AI application commercialization lag behind capex, inference demand fail to catch up with training demand, or depreciation and power costs begin eroding margins, valuations will react first. Correct directionality in the underlying industry does not guarantee stock prices can advance indefinitely ahead of fundamentals.
Short-Term Biggest Risk: Rates Rising Faster Than ARR
The true near-term pressure comes from liquidity.
If the Strait of Hormuz remains closed long-term, keeping oil prices above $100—or even higher—inflation may spread from energy into services, transportation, and raw materials. April’s U.S. PPI YoY rose to 9.8%, the highest since October 2022. Once inflation becomes entrenched, the Fed’s policy path would be forced to reset.
Swap markets now price in ~0.8 Fed hikes this year, with the ECB and Bank of England pricing in over two hikes each. Meanwhile, concerns over the Fed’s policy independence amid leadership transition—and growing internal FOMC dissent—are further undermining market confidence in future easing.
Japan is another gray rhino. Long the world’s funding pool for leveraged trades, Japan now faces yen depreciation and inflationary pressures prompting the Bank of Japan to signal tightening—causing 30-year JGB yields to rise above 4%. If Japanese funding costs keep climbing, triggering global unwinds of carry trades, high-valuation AI assets will struggle to remain unscathed.
A preview occurred on May 15: the 10-year U.S. Treasury yield breached 4.5%; the 30-year yield topped 5%. Highly crowded momentum trades cooled, the Philadelphia Semiconductor Index fell ~4% intraday, and the Nasdaq declined ~1.5%. This was not evidence of a trend reversal—but it did confirm how acutely crowded trades respond to rate moves.
The crucial near-term comparison is simple: Can ARR upgrades outpace rising rates? If not, capital may rotate first into hardware segments offering greater near-term certainty; if liquidity deteriorates further while AI revenue expectations stall, valuation pressure will intensify markedly.
Harder Medium- to Long-Term Questions: Organization, Power, Jobs, and Hardware Roadmaps
Medium-term tests center on industrial delivery. General-purpose technological revolutions rarely follow straight-line trajectories—but rather “accelerate, decelerate, then accelerate again.” First comes a wave of capital investment, then organizational adaptation, and finally productivity release. The early internet era similarly saw investment surges, capex expansion, and asset bubbles—while meaningful productivity gains emerged only years later.
The difficulty in pricing AI today lies in its implicit assumption that enterprises must rapidly reconfigure organizations, workers must quickly retrain, business models must swiftly achieve product-market fit, and society must avoid strong resistance. Such velocity is historically rare.
Longer-term constraints are even harder.
First is energy and infrastructure. AI data centers consume vast amounts of electricity and cooling water. Grid upgrades, transformers, and energy storage are not PowerPoint slides—they’re real bottlenecks. If AI infrastructure continues pushing up nationwide electricity costs, regulatory scrutiny and public backlash will escalate.
Second is employment and consumption. AI boosts enterprise efficiency in the short term—reducing demand for engineers, customer-service staff, and other roles—but if technological unemployment outpaces new job creation, household purchasing power weakens. B2B efficiency gains ultimately rely on C2C spending power for monetization; should non-AI sectors slide into recession, AI cannot sustain its outperformance indefinitely.
Third is societal acceptance. China saw a national rush to install Openclaw earlier this year—but in the U.S., public resistance is mounting against rising electricity bills caused by data centers and fears of technological unemployment. This will constrain AI’s penetration speed.
Fourth is hardware technology breakthroughs. A “DeepSeek moment”—an engineering leap dramatically improving compute, storage, or transmission efficiency—could suddenly render today’s most constrained hardware segments oversupplied. The high-visibility logic underpinning the hardware chain is not immune to disruption.
AI’s long-term industrial outlook remains optimistic. Excluding social tensions arising from technological unemployment and restructuring of production relations, AI genuinely holds potential to lift total-factor productivity and help economies escape stagflation. Even if financial markets undergo mid-cycle deleveraging, the legacy infrastructure—data centers, low-cost technologies, and validated use cases—may form the foundation for the next industrial expansion.
But stock pricing is not synonymous with industrial vision. What this AI bull market most urgently needs to validate is whether the ARR, ROI, and technology-penetration speeds currently priced in can still materialize amid hardening constraints—from oil prices and inflation to interest rates and social pushback. Directional correctness explains why a bull market exists—but execution speed determines whether the bubble spins out of control.
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
