On July 24, 2026 (UTC), US equities underwent a market-shaking correction. The Nasdaq Composite plunged 553.21 points to close at 25,137.69, a drop of 2.15%. The S&P 500 fell 1.21%, while the Dow declined 0.97%. The Tech Seven Giants Index dropped 4.8% in a single day, wiping out a combined $797 billion in market value.
On the individual stock level, Tesla (TSLA) closed at $319.69, down 14.52%—its largest single-day decline since March 11, 2025. The company’s Q2 net profit missed expectations and its gross margin deteriorated further. Google (GOOG) closed at $318.34, down 6.89%, with its total market cap falling below $4 trillion. Amazon (AMZN) ended at $233.66, down 4.57%; Meta (META) at $606.10, down 3.36%; Microsoft (MSFT) at $381.58, down 2.24%; Apple (AAPL) at $321.66, down 1.30%; Nvidia (NVDA) at $208.76, down 1.56%; AMD at $539.69, down 2.29%; and Broadcom (AVGO) at $392.47, down 1.09%.
This downturn wasn’t an isolated event. It unfolded against the backdrop of global tech giants collectively ramping up their AI capital expenditures—Alphabet raised its 2026 capex guidance from $180–$190 billion to $195–$205 billion; Microsoft expects $190 billion in 2026 capex; Amazon projects $200 billion. The combined capex ceiling for the top four cloud providers in 2026 has surged to $725 billion, up 77% from 2025.
On one side, we see unprecedented capital investment; on the other, sharp share price corrections. The market is grappling with a central question: Are we still in the early phase of the AI investment cycle, or has it already been overextended?
Layer One: Compute Infrastructure—AI’s "Engine Room"
At the foundation of the AI value chain lies the chips and network equipment that deliver computational power. The prosperity of this layer directly determines the upper limit of the entire AI industry.
Nvidia remains the undisputed leader in this space. For the first quarter of fiscal 2027 (ending April 26, 2026), Nvidia posted quarterly revenue of $81.615 billion, up 85% year-over-year; data center revenue reached $75.2 billion, up 92%. Data center business now accounts for over 90% of total revenue. Nvidia holds roughly 80% market share in AI accelerator sales. The mass production of Blackwell architecture is driving continued performance growth, and anticipation for its next-generation products is further boosting valuation expectations.
AMD is accelerating its catch-up. On July 24, 2026 (UTC), AMD unveiled the Instinct MI400 GPU series and sixth-generation EPYC "Venice" server CPUs at the Advancing AI 2026 conference. The MI455X, built on the CDNA 5 architecture, boasts 320 billion transistors and up to 432GB of HBM4 memory. The company also announced the Helios rack-scale AI platform has entered full production, with deliveries starting at the end of Q3 2026. Each rack integrates 72 MI455X GPUs. CEO Lisa Su stated at the conference that the AI accelerator market will reach $1.4 trillion by 2030. Although AMD shares fell 2.29% that day due to broader market weakness, its rapid product iteration is posing a substantial challenge to Nvidia’s market share.
Broadcom is charting a differentiated path. As a leader in ASIC (application-specific integrated circuit) technology, Broadcom is benefiting from the trend of hyperscale cloud providers developing their own chips. In Q2 fiscal 2026, Broadcom’s total revenue reached $22.19 billion, up 48% year-over-year—a new record. The company expects AI semiconductor revenue to hit $56 billion in fiscal 2026, up about 180% from fiscal 2025. Q3 AI semiconductor revenue is projected to accelerate to $16 billion, up over 200% year-over-year. Notably, Broadcom’s Q3 AI revenue guidance was slightly below some analysts’ expectations, triggering after-hours price volatility—highlighting that expectations for the AI chip sector are already highly saturated, and any marginal miss can prompt valuation adjustments.
From an industry perspective, Deloitte estimates the global AI chip market will approach $500 billion in 2026. JPMorgan projects total global AI chip shipments at about 16.3 million units in 2026, with 6.8 million ASIC chips and 9.5 million general-purpose GPUs. The World Semiconductor Trade Statistics (WSTS) organization forecasts the global semiconductor market will grow 89.9% year-over-year in 2026, reaching $1.5112 trillion.
The demand logic for compute infrastructure is clear—AI model training and inference are consuming computational power at an exponential rate. But valuation challenges are equally real: when the market has already priced in 2–3 years of future growth, any downward revision in performance guidance can trigger sharp volatility.
Layer Two: Cloud Platforms—AI’s "Pipeline System" for Commercialization
If chips are the "engine" of AI, cloud platforms are the "pipeline system" that delivers AI capabilities. Without cloud infrastructure, AI models cannot be deployed at scale, and AI services cannot reach end users.
Alphabet (Google) is the most noteworthy case in this layer. The company’s capital expenditures for the first half of 2026 totaled $80.598 billion, up 103% year-over-year. Management has clearly stated that AI demand is driving growth in search, cloud, and enterprise services. However, massive capital spending is also creating financial pressure—Alphabet’s free cash flow turned negative for the first time. As of the end of June, future spending commitments reached $811 billion, up nearly $500 billion from three months prior, covering chips, data centers, power, inventory, and content licensing. This is the deeper reason behind Google’s over 7% share price drop on July 24: the market is now questioning when such large-scale capex will translate into meaningful profit.
Microsoft is following a similar path. The company expects 2026 capex to reach $190 billion, with about $25 billion directly attributable to price increases for AI core components like GPUs and CPUs. CFO Amy Hood stated that, given rising demand signals, increased product usage, and platform-driven efficiency gains, the company remains confident in its investment returns. Azure cloud services and the Copilot AI assistant are Microsoft’s dual engines for AI commercialization, and the company’s expanded partnership with France’s Mistral AI (providing billions to build GPU data centers in Europe) underscores its global ambitions.
Amazon is also ramping up. The company expects 2026 capex to hit $200 billion, up from $131 billion in 2025, with most funds allocated to data centers, chips, and supporting hardware. The AWS CEO has made it clear that most of 2026’s capex will be monetized in 2027–2028, and a substantial portion is already backed by customer commitments. To support this investment cycle, Amazon launched a bond issuance plan of at least $25 billion.
The central dilemma for cloud platforms is this: capex expansion is certain, but the pace and efficiency of monetization are uncertain. Goldman Sachs projects global AI capex around computing, data centers, and power will reach $7.6 trillion from 2026 to 2031, with annual investment rising from $765 billion in 2026 to $1.64 trillion in 2031. Such massive capital consumption demands that cloud providers achieve high enough commercialization conversion rates in AI services.
Layer Three: Application Ecosystem—AI’s "Ultimate Value Realization"
The value of AI infrastructure and cloud platforms must ultimately be realized through end-user applications. Companies in this layer represent the "last mile" of AI’s journey from technology to business.
Tesla is the most dramatic example. On July 24, Tesla plunged 14.52% as its Q2 net profit missed expectations and gross margin declined further. Tesla’s AI narrative—full self-driving (FSD), Optimus robot, Dojo supercomputer—has long been a key driver of its high valuation. But when the core automotive business’s profitability falters, the market’s patience for the AI story quickly tightens. Tesla has now retraced 34% from its peak.
Meta’s AI strategy centers on optimizing recommendation algorithms and AI ad tools. The company is also investing heavily in AI infrastructure, as one of the four leading cloud providers. Meta’s 3.36% drop on July 24 reflects market concerns over both the ceiling for social media ad growth and the ROI of AI investments.
Apple is taking a more cautious approach to AI. Apple is pursuing an "on-device AI" strategy—integrating AI capabilities directly into end devices, rather than building massive cloud compute infrastructure. On July 24, Apple fell just 1.30%, the smallest decline among the Tech Seven Giants. This may indicate that, in the current environment, a more restrained capex strategy is gaining favor.
The defining feature of the application ecosystem layer is that AI’s revenue contribution remains in its early stages. Whether it’s Tesla’s FSD, Meta’s AI ads, or Apple’s on-device AI, none have yet become dominant forces in their respective revenue structures. But this "not yet realized" status creates the greatest imagination space—and the greatest uncertainty—for application-layer companies.
Cycle Positioning: What Stage Is AI Investment In?
To answer the opening question: Is the AI investment cycle still in its early phase?
From an industry fundamentals perspective, the answer is yes. The global AI chip market is expected to approach $500 billion in 2026, and WSTS’s forecast of a $1.5112 trillion total semiconductor market shows that AI chips’ share is rising rapidly but is far from saturated. Hyperscale cloud providers may invest over $6 trillion in AI by 2030. AMD projects the AI accelerator market will reach $1.4 trillion by 2030. All these data points indicate that AI infrastructure construction is still climbing, nowhere near a plateau.
But from a market valuation perspective, the picture is more complex. The broad sell-off on July 24 signals that the market is no longer buying the "AI narrative" unconditionally. Investors are now distinguishing between "who is truly creating AI value" and "who is just telling an AI story." Alphabet’s capex hike triggered a sharp share price drop—not because the market doubts AI’s value, but because it questions the efficiency and return cycle of capital expenditure.
Blackstone expects to complete about $100 billion in investments or commitments in its own data center portfolio by the end of 2026, plus roughly $200 billion in third-party enterprise AI chip capacity investments during the same period. In the short term, global AI infrastructure investment will reach about $300 billion—a scale comparable to the GDP of the world’s top 50 economies. At this magnitude, scrutiny of investment returns becomes far more rigorous.
Another signal worth noting is the actual progress of data center deployments. Data Center Watch reports that, in Q1 2026 alone, the value of US data center projects blocked or delayed reached about $130 billion—nearly matching the total for all of 2025. Capital is flooding in, but implementation is lagging—this mismatch is intensifying market anxiety.
Conclusion
From compute chips to cloud platforms to end-user applications, the three-layer AI value chain is undergoing a shift from "expectation-driven" to "performance-validated." The $797 billion evaporation in Tech Seven Giants’ market cap on July 24 is less a sign of an AI bubble bursting than a stress test for the AI investment cycle.
In the short term, the tension between high capital expenditures and uncertain return cycles will continue to drive volatility in tech stock valuations. But in the medium to long term, AI’s impact on global economic efficiency is just beginning. Goldman Sachs projects annual global AI-related capex will reach $1.64 trillion by 2031—this is still an early-stage story, though the market narrative is shifting from "storytelling" to "accounting."
For investors, the next phase of competition in the AI value chain will shift from "who invests more" to "who invests more efficiently." Technical barriers in compute infrastructure, ecosystem stickiness in cloud platforms, and commercialization capability in applications—these three dimensions will determine which assets lead the next wave of tech stock growth.
FAQ
Q1: What stage is AI infrastructure investment currently in?
Global AI capex is still in a rapid expansion phase. Goldman Sachs projects related capex will reach $7.6 trillion from 2026 to 2031, with annual investment rising from $765 billion to $1.64 trillion. However, the market has shifted from "unconditional optimism" to "scrutinizing return efficiency," and investment logic is moving from expectation-driven to performance-validated.
Q2: What is the core driver for chip stocks in this AI cycle?
The fundamental driver is the computational power needed for AI model training and inference. The global AI chip market is expected to approach $500 billion in 2026. Nvidia holds about 80% of the AI accelerator market share, AMD is catching up with its MI400 series, and Broadcom is benefiting from custom ASIC chip trends among cloud providers.
Q3: Can the massive capex of cloud computing giants deliver corresponding returns?
In the short term, there is a mismatch between capex and profit growth—Alphabet’s free cash flow has turned negative. However, Microsoft and Amazon both report that future compute demand is already locked in by customer commitments. Realizing returns requires a 2–3 year cycle, with the key being the commercialization efficiency of AI services.
Q4: What is the biggest challenge facing AI application companies right now?
AI’s revenue contribution has not yet become dominant. Tesla’s AI narrative was hit as core automotive profitability declined, resulting in a 14.52% plunge. The main challenge for application-layer companies is how to convert AI capabilities into sustainable revenue growth while maintaining profitability in their traditional businesses.
Q5: Does the tech sector’s July 24 sell-off signal an AI bubble burst?
The $797 billion drop in Tech Seven Giants’ market cap on July 24 is more a concentrated response to the tension between high capex and uncertain returns than a reversal in AI industry fundamentals. Global AI capex is still expanding, but the market narrative is shifting from "storytelling" to "accounting," and valuation divergence is becoming the norm.
