On July 24, 2026 (Beijing time), the US tech sector experienced a dramatic shake-up. Tesla (TSLA) closed at $319.69, plunging 14.52% in a single day—the biggest one-day drop since March 10, 2025. Google (GOOG) ended at $318.34, down 6.89%. Both stocks fell far more than the broader market: the Nasdaq dropped 2.15% and the S&P 500 slipped 1.21% that day. Collectively, the "Magnificent Seven" tech giants lost nearly $800 billion in market value in a single session.
The immediate trigger for the sell-off was the near-simultaneous release of quarterly earnings reports from both companies. The numbers themselves weren’t bad—in fact, they were quite strong. Yet the market’s reaction was to "vote with its feet." This contradiction points to an emerging consensus: the "story phase" of AI capital expenditure is ending, and the "validation phase" has begun.
Earnings Beat Expectations—So Why Did the Market Shrug?
Let’s start with Tesla. In Q2 2026, Tesla reported revenue of $28.236 billion, up 26% year-over-year and beating the market’s expectation of $25.71 billion. Automotive revenue reached $20.516 billion, up 23% year-over-year. Quarterly deliveries soared 25% to 480,100 vehicles. By the numbers, this was a robust earnings report.
But the profit side told a different story. Adjusted earnings per share for Q2 came in at just $0.33, well below the $0.51 expected. Operating profit was only $398 million, down 57% year-over-year, and the operating margin shrank from 4.1% a year ago to just 1.4%. More concerning for the market, Tesla’s free cash flow turned negative for the first time in over two years, reaching -$1.092 billion. Capital expenditures for the quarter were $5.789 billion, putting the company on pace for just $17 billion for the year—well below Elon Musk’s previous target of $25 billion.
Now to Google’s parent company, Alphabet. Q2 revenue rose 24% year-over-year to $119.8 billion. Google Cloud revenue hit $24.768 billion, up 82% year-over-year, with operating profit up 212% and margins improving to 35.6%. For the first time, cloud backlog surpassed $500 billion, reaching $514 billion. This was another set of results that blew past expectations.
However, Alphabet’s Q2 capital expenditures soared to $44.924 billion—double the prior year. The company raised its full-year 2026 capital expenditure guidance from the $180–190 billion range set in April to $195–205 billion. As a result, free cash flow turned negative by $5.855 billion in Q2—the first time Alphabet has reported negative free cash flow since going public decades ago.
Both companies’ earnings revealed a common structural dilemma: revenue is growing, profits are getting squeezed, cash is flowing out, and capital expenditures keep climbing. The market’s interpretation: the costs of AI investment are becoming visible much faster than expected, while the returns remain in the "storytelling" phase.
Wedbush’s "15% Theory": Comforting or Convincing?
In response to the sell-off, Dan Ives, Managing Director at Wedbush Securities, offered a clear perspective on CNBC. He argued that capital spending on AI infrastructure is still in its early days—only about 15% complete. He likened today’s hyperscaler capital outlays to the early construction of the Las Vegas Strip: "build first, the returns will follow." Ives believes this correction is a matter of timing, not valuation.
This view directly counters the growing concerns about an "AI bubble." GMO co-founder Jeremy Grantham has compared AI to the railroad and internet bubbles. Noted short-seller Jim Chanos has said the scale and risk of this AI infrastructure buildout far exceed those of the dot-com era. JPMorgan’s strategy team recently warned that today’s AI stocks are behaving much like late-1990s internet stocks.
Whether Ives’s "15% theory" holds up depends on one unavoidable metric: Return on Investment (ROI). If the ROI from AI infrastructure materializes in the medium term, today’s capital spending is "planting seeds." If not, it becomes "sunk cost." This is the core divide in current market pricing.
Seven Giants, Seven AI Narratives
Alphabet: Cloud is the bright spot, but the cash burn is accelerating. Google Cloud’s 82% growth and $514 billion backlog show that AI demand is real and converting to revenue. The problem: capital expenditures (up 100% year-over-year) are growing much faster than cloud revenue (up 82%). This means that each new dollar of cloud revenue requires more capital—the scale effect hasn’t kicked in, and the "return on investment" is actually worsening.
Tesla: The most ambitious AI story, but the payoff is furthest away. Tesla’s AI narrative rests on three pillars: autonomous driving, Robotaxi (driverless ride-hailing), and the Optimus humanoid robot. The earnings report shows Tesla is installing the first-generation Optimus production line at its Fremont, California plant, aiming for production later this year. Musk called 2026 an "extraordinarily capital-intensive year," with spending to exceed $25 billion. However, the auto business remains Tesla’s only meaningful revenue source; new ventures like Robotaxi have yet to contribute. As Ives bluntly put it, "Investor patience is wearing thin."
Microsoft: 2026 capital expenditures are projected at about $190 billion, up 48% year-over-year, with roughly $25 billion attributed to rising costs for AI hardware (GPUs, CPUs, etc.). The company plans to double its overall computing power within two years.
Amazon: 2026 capital spending is expected to approach $200 billion. To support this, Amazon has raised over $82 billion through multi-currency bond offerings.
Meta: 2026 capital expenditure guidance has been raised to $115–135 billion, about double the $72.2 billion spent in 2025.
Taken together, Microsoft, Google, Amazon, and Meta—the four leading cloud providers—are now targeting a combined capital expenditure ceiling of $725 billion for 2026, up 77% from 2025. Morgan Stanley projects that the five major hyperscalers (adding Oracle) will spend $805 billion on capex in 2026, rising to $1.116 trillion in 2027.
How Is the Valuation Logic Changing?
From "asset-light platforms" to "asset-heavy intelligent factories"—this is the most profound shift now underway in tech stock valuation.
For the past two decades, the world’s best platform companies have typically been asset-light, with high ROE (return on equity), strong free cash flow, powerful network effects, and near-zero marginal costs. In the AI era, however, every inference and call requires compute, storage, networking, energy, and data centers. Platform companies are shifting from "asset-light" to "asset-heavy"—and the valuation logic is fundamentally different.
UBS data shows that as AI spending commitments have soared over the past two years, large tech companies’ forecasted Cash Flow Return on Investment (CFROI) has dropped by 200 basis points. At the same time, free cash flow across the tech sector has fallen to a ten-year low. Moody’s data further reveals that Amazon, Meta, Alphabet, Microsoft, and Oracle have signed, but not yet recognized on their balance sheets, long-term data center lease commitments totaling about $662 billion—equal to 113% of their combined adjusted debt. These "hidden liabilities" will become explicit in the coming years.
The market is shifting from "pricing for growth" to "pricing for efficiency." In the past, investors were willing to pay a premium for the AI story. Now, they want proof that AI investments can deliver positive returns. This is the root cause of the post-earnings sell-off in Tesla and Google—not poor performance, but a pace and scale of "cash burn" that exceeds the market’s tolerance for delayed payback.
Where Might the Long-Term Winners Be?
Looking at the current landscape, different segments of the AI value chain will benefit at different times and to varying degrees.
The first layer is AI infrastructure providers—chips, optical modules, servers, storage, and so on. No matter which cloud provider ultimately wins, demand for infrastructure is a given. In the first half of 2026, the Philadelphia Semiconductor Index rose 101.14%. This business model is the most straightforward: the "shovel sellers" don’t need to know who finds gold.
The second layer is cloud providers with proprietary data and application scenarios—Microsoft, Google, Amazon, etc. Their edge is combining AI with existing business ecosystems to monetize across platforms. Google Cloud’s 82% growth and $514 billion backlog provide early validation for this approach. The challenge: can capital expenditure growth remain sustainable, and will margins keep getting squeezed?
The third layer is AI-native applications—still in early exploration, with business models far less mature than the first two layers.
Robert Subbaraman, Head of Global Macro Research at Nomura, notes there’s no sign yet that hyperscale cloud companies are voluntarily slowing investment. Gartner forecasts global AI spending will reach $2.52 trillion in 2026, up 44% year-over-year. But Subbaraman also admits that a key risk to the bubble is "AI ROI falling short of expectations."
Conclusion
The sharp one-day drops in Tesla and Google are not the end of the AI narrative. Rather, they mark a pivotal transition from "proof of concept" to "financial validation." The market is now asking a very basic question: What kind of returns will $725 billion in annual capital expenditures actually deliver?
There won’t be a definitive answer in the short term. Building AI infrastructure is akin to constructing the internet backbone in the 1990s—massive upfront investment, long payback periods, and few ultimate winners. The difference: this round of spending dwarfs the internet bubble era by several orders of magnitude.
For investors, the key question is no longer "Is AI important?"—that’s a given—but "Which AI investments will generate positive returns?" Capital expenditure itself isn’t the problem; the issue is how efficiently it’s deployed. As the market shifts from "paying for the story" to "pricing for efficiency," the revaluation of tech is just beginning.
FAQ
Q: Why did Tesla and Google stocks plunge despite beating earnings expectations?
The market’s focus has shifted from "revenue growth" to "AI return on investment." Both companies beat revenue estimates, but surging capital expenditures turned free cash flow negative—Tesla at -$1.092 billion, Alphabet at -$5.855 billion. Investors are concerned that massive AI spending won’t generate commensurate returns in the near term, prompting them to sell.
Q: What does Wedbush mean by "only 15% of AI investment is complete"?
Wedbush analyst Dan Ives believes AI infrastructure buildout is still in its early stages—akin to 1996, not 1999, in internet terms. He likens today’s hyperscaler capex to "build now, profit later," arguing that the sell-off reflects a timing issue, not a valuation problem.
Q: How large will the Big Four’s AI capital expenditures be in 2026?
Microsoft, Google, Amazon, and Meta are targeting a combined capex ceiling of $725 billion in 2026, up 77% from 2025. Alphabet’s guidance is $195–205 billion, Microsoft around $190 billion, Amazon close to $200 billion, and Meta between $115–135 billion.
Q: How does the current wave of AI investment compare to the dot-com bubble?
Similarities: Both involve disruptive technologies driving massive capital outlays and frothy valuations. Differences: This AI buildout is on a scale of hundreds of billions per year—far above the dot-com era’s $20 billion annual pace. JPMorgan sees strong parallels, while short-seller Chanos argues "this time is worse."
Q: Who are the likely long-term winners in AI capital expenditure?
At present, the first beneficiaries are AI infrastructure providers (chips, optical modules, servers, etc.), with the highest certainty of demand. The second layer is cloud providers with data and application scenarios (Microsoft, Google, Amazon), who have cross-monetization potential but face efficiency pressures. The third layer is AI-native applications, where business models are still nascent. Ultimately, the winners will be those who can generate the highest AI commercial returns with the least capital invested.
