Amazon and Intel See Sharp Stock Declines Amid AI Infrastructure Challenges

The aftershocks from Google's 13% single-day plunge yesterday haven't yet subsided, as both Amazon and Intel plunged sharply in after-hours trading this morning.

Interestingly, the most standout segments in both earnings reports were AI-related businesses: AWS revenue grew nearly 40% year-over-year, and Intel’s data center and AI business surged an impressive 78%. If viewed solely from the demand side, the vitality of AI infrastructure appears to be in excellent shape.

But the market reaction was negative: Amazon's stock plunged about 7% after hours, erasing its entire 5.6% gain for the year; Intel, which had surged over 150% earlier this year and was seen as a benchmark for turnaround success, tumbled more than 22% in after-hours trading.

Is the market no longer applauding the "AI grand narrative" unconditionally? How is Wall Street’s evaluation logic around AI infrastructure shifting?

When reviewing Google's plunge yesterday, we highlighted its core issue: AI capital expenditures are visibly devouring cash flow, while its core money-making business—search advertising—is losing momentum (see further reading: Google Drops 10%: Is the Core Business Collapsing, or Is This Just Another 'Microsoft-Style' Mispricing?).

Just 24 hours later, Amazon delivered an almost identical "medical record."

Looking solely at its cloud computing business, AWS's performance is truly impressive: revenue in Q3 increased by 39% year-over-year to $46 billion, marking six consecutive quarters of accelerating growth. CEO Andy Jassy even assured on the earnings call that AWS will maintain annual growth of over 40% for the next two years.

More importantly, driven by ongoing expansion of compute leasing by major AI companies like OpenAI and Anthropic, AWS is not only growing in scale but also delivering tangible profits, boosting Amazon’s overall operating profit for the third quarter to $17.8 billion (up from $11.4 billion year-over-year), while also improving the operating margin of its cloud business.

In previous years, such a strong cloud business performance would have pushed the stock to new highs, but why is Wall Street not only refusing to buy in but also selling off outright?

First, the threshold of "exceeding expectations" has been raised by Microsoft.

On the same day, Microsoft reported its earnings, with cloud revenue surging 48% and capital expenditures remaining relatively restrained, resulting in a stock price decline of less than 1%. In contrast, AWS’s 39% growth not only failed to outperform the benchmark but was achieved at the cost of extremely heavy capital expenditure.

Second, the two major "cash cows"—e-commerce and advertising—are simultaneously slowing down.

Amazon's online retail revenue in the third quarter grew only 2% to $68.5 billion, falling to the lowest level since early 2023; what alarmed the market even more was the high-margin advertising business, whose growth rate dropped sharply from a stable pace of over 20% over the past year to just 11% this quarter, hitting the lowest level since the data was first reported separately in 2020.

Although management attempted to explain the weakness in retail by citing that Prime Day was moved to the second quarter, the chain reaction of consumers cutting spending and advertisers subsequently tightening budgets is now evident—this is why its guidance for fourth-quarter holiday season revenue and operating profit both fell below market expectations.

Finally, it was the completely out-of-control capital expenditures and free cash flow that crushed market sentiment.

In the third quarter alone, Amazon spent over $77 billion on equipment and property, far exceeding Wall Street expectations; as a result, its free cash flow over the past 12 months reached -$42.4 billion, compared to a positive $14.8 billion during the same period last year, representing a cash burn of nearly $60 billion in just one year!

Even though the books have suffered heavy losses, Jassy still stated on the earnings call, "Capital expenditures of $220 billion in 2026 will not be reduced, and will be significantly increased in 2027."

However, it is worth noting that Amazon’s operating cash flow has not deteriorated; over the past 12 months, operating cash flow still reached approximately $168.3 billion, a year-over-year increase of about 29%. However, the pace of spending on data centers, servers, networking equipment, and other infrastructure has far outstripped the growth of operating cash flow.

In other words, Amazon isn't unable to make money—it's just that the money it makes isn't enough to cover the costs of its infrastructure expansion.

This provides an interesting contrast to yesterday’s Google, which reported a free cash flow negative of $9.1 billion and saw its stock price plummet 13%:

• Google's challenge lies in having multiple entry points such as Search, Gemini, TPU, and Google Cloud, but there is a clear lag between improvements in AI capabilities and their commercial monetization;
• Amazon's dilemma is that while AWS is already growing rapidly and generating substantial profits, it must exhaust all of its current cash flow to meet the demands of future AI needs;

From this perspective, Amazon has also caught the same illness as Google.

When the cash flow generated by core operations can no longer keep up with the capital consumption of AI infrastructure, the massive CapEx sitting on the balance sheet becomes a real financial burden.

From Google to Amazon, the core issue is "throwing too much money at it"; while Intel, which plunged over 20% after hours, is trapped in the opposite extreme—even with enormous demand, it may not be able to deliver on time.

The stock price trajectory of this veteran chip giant this year is truly worth reviewing.

Back in January, we published an article titled "Intel's Moment of Truth: At the ICU Door, How Did Chen Liwu Settle the Legacy and Save Himself?", analyzing the first major move by new CEO Chen Liwu after taking over from Kissinger.

At the time, Intel remained in a rather awkward position. The aggressive expansion plan left behind by Kilian required retrenchment, massive capital expenditures continued to strain cash flow, the 18A process had not yet been fully proven, and the first priority for Victor Peng was not to reclaim industry leadership, but to ensure the century-old chip giant survived.

At that time, the core judgment of the article was clear: Chen Liwu used a combination of "global layoffs of 15% + suspension of factory construction in Europe + Q1 guidance reset" to barely lift Intel's gross margin back to 38%, achieving initial stabilization; the 18A process and Panther Lake are Intel's only ticket back to the table.

This is a financial report that has successfully stopped the bleeding but has not yet recovered; whether it can truly leave the ICU depends not on this quarter’s results, but on execution over the next 12–18 months.

Pull up the K-lines from the past nine months: during the first half of this "12–18 month survival validation period," Intel indeed experienced a highly enthusiastic honeymoon phase.

With cost-cutting through layoffs, the influx of external strategic funding from NVIDIA and the U.S. government, and the completion of a $23 billion equity financing in August, coupled with the gradual ramp-up of mass production for its consumer-grade Panther Lake chips and server products Clearwater Forest based on the 18A process, the market once believed Intel had been taken off life support and exited the ICU, with its stock price surging over 150% year-to-date.

Even looking solely at the Q3 earnings disclosed today, the figures remain impressive: Q3 revenue increased 21% year-over-year to $16.5 billion, adjusted EPS reached 52 cents (above the expected 41 cents), and data center and AI (DCAI) business surged 78% year-over-year to $7.3 billion, with the overall gross margin for the quarter climbing as high as 43%.

But as soon as the Q4 guidance was released, all optimistic expectations were instantly put on hold:

• Revenue shortfall: Q4 revenue is expected to be only $15 billion to $16 billion, significantly below the market average expectation of $17 billion (with optimistic forecasts exceeding $18 billion);
• Profit halved: Adjusted earnings per share are expected to be only 25 cents, far below Wall Street’s forecast of 43 to 45 cents;
• Gross profit margin reverts to previous levels: Gross profit margin is expected to narrow from 43% in Q3 to 38% in Q4;

After swirling around for nearly a year, the stock price more than doubled, but when we crunched the numbers for the fourth quarter, Intel’s gross margin neatly returned to exactly where it was back in January when we wrote that ICU article...

Why, despite AI server CPU demand being so strong that "shortages will persist until 2028," did Intel suddenly falter in its most critical quarter?

The earnings call explained that it is trapped in an extremely awkward internal conflict.

On one hand, the new engine is facing “orders that can’t be fulfilled.” AI has indeed surged demand for general-purpose server CPUs; the CFO expects DCAI department revenue to grow another 15% quarter-over-quarter in Q4, with industry-wide server CPU shipment growth exceeding 30% in both 2026 and 2027. However, advanced organic substrates used for EMIB packaging are experiencing industry-wide shortages, severely capping production capacity. Even Chen Liwu himself admitted in an interview: “I’m somewhat disappointed that we can’t meet our customers’ needs.”

On the other hand, the traditional PC market is being drained and backfired by AI. In the third quarter, Intel’s client business (PC chips), traditionally its cash cow, generated $8.1 billion in revenue, a 5% year-over-year decline, falling short of the expected $8.98 billion. More critically, the company forecasts that the weak PC market, which has accelerated its correction over the past two months, will remain sluggish at least until the first half of 2027—global PC shipments are expected to drop 12% in 2026 and a further 10% in 2027.

The culprit is none other than the booming AI servers themselves.

The surge in demand for AI servers is wildly consuming global DRAM memory chip capacity, causing severe shortages and sustained price increases for memory from Samsung, SK Hynix, and Micron, which in turn has driven up the prices of laptops and desktops, directly crushing end-user demand. "My prediction is that by 2027, this will become a serious issue for everyone—not just us, but the entire market will be impacted."

Meanwhile, to advance its foundry expansion, Intel has designated 2026 as a “strategic investment year,” raising its capital expenditure forecast from $18 billion in July to $20 billion, with a further significant increase planned for 2027. Management also expects no positive cash flow in 2026, with the target for breakeven cash flow now delayed to 2027.

Its traditional PC business has been drained by rising memory prices, while its new AI server engine is bottlenecked by supply chain issues; it must continue burning through $20 billion on its balance sheet with no positive cash flow—this is the deeper logic behind the 20% post-market plunge.

If Amazon and Intel have exposed the AI infrastructure's "cash flow anxiety" and "delivery bottlenecks," respectively, then a series of recent events in the Middle East have brought another previously severely underestimated risk to the forefront:

Is a data center really just a commercial facility?

Carefully examine the immediate trigger for Intel's Q4 guidance collapse—beyond weak PC demand, the biggest sudden shock was the explosion at Intel’s wafer fab in Kiryat Gat, Israel, in October this year.

This factory primarily uses a 10-nanometer process to produce data center chips. The accident not only resulted in fatalities but also caused the facility to shut down directly for approximately two weeks. Management is still individually assessing the usability of the damaged wafers. According to estimates, this single explosion and subsequent shutdown will directly eliminate about $300 million in server chip revenue for Intel's fourth quarter, reducing the company's overall gross margin by 200 basis points, with impacts extending into the first quarter of next year.

Looking back at Part 4 of the article from January this year, we wrote: "NVIDIA's $5 billion investment in Intel to support its domestic production capacity is essentially purchasing a long-term 'geopolitical insurance policy' to reduce the industry's reliance on a single overseas source."

The bitterly ironic result is that the geopolitical boomerang arrived faster than anyone anticipated: just as Intel’s most profitable data center business urgently needed production capacity, its core factory located in the heart of the Middle East’s storm erupted in an explosion, causing a shutdown.

Following this logic, Amazon's sharp decline tonight is also inevitably shadowed by this underlying factor.

When Amazon spends $77 billion in a single quarter on equipment and property, and expands its AWS computing infrastructure globally, its business model has already shifted from a lightweight internet platform to a heavy-asset infrastructure heavily reliant on physical security.

Keep in mind that the Middle East has been a key growth market for Amazon AWS and Google Cloud in recent years, with numerous core cloud server nodes deployed across regions from Tel Aviv in Israel to multiple Gulf countries.

If you review the historical messages from the past period, you’ll find that as the situation in the Middle East has repeatedly escalated, Amazon and Google’s cloud server nodes and data centers in the region have been listed multiple times on public threat lists, with threats to target them physically.

• In March this year, Iran targeted two AWS facilities in the UAE, and a drone attack occurred near an AWS facility in Bahrain, causing infrastructure damage;
• By October, security threats surrounding data centers, AI facilities, and undersea communication cables in the Gulf region began to escalate again;
• Iran has also publicly warned that if the war reignites, relevant facilities could become targets;

To some extent, the smart money that swiftly sold off Amazon after the earnings call and expressed concern over the $220 billion capital expenditure may already have priced in this geopolitical risk—given that Intel’s wafer fab in Israel has been shut down due to explosions, resulting in a $300 million quarterly loss, it may only be a matter of time before its cloud server clusters, which are similarly concentrated in Middle Eastern nodes, are affected.

In the age of technology, attack methods can sometimes be surprisingly simple.

In times of peace, everyone competes over large model parameters, algorithm optimization, and millisecond-level cloud service latency; but when geopolitical conflict turns violent, no sophisticated hacking skills are needed—a single physical explosion at a wafer factory or cloud node data center can instantly cut off the most advanced AI computing power, bringing hundreds of billions of dollars in capital investment to a halt.

From Google yesterday to Amazon and Intel today, the successive declines of Silicon Valley giants are essentially forcing a much-needed cooldown on the AI narrative that has been overheated over the past two years.

In the past, the market believed that “big spending leads to miracles”—as long as companies dared to invest heavily in CapEx and buy chips to build factories, their valuations would keep rising. But now:

• Cash flow has become the only hard currency: whether it’s Amazon’s -$42.4 billion in free cash flow or Intel’s projected cash flow turnaround not until 2027, both underscore one thing—in a cycle where core businesses (e-commerce, advertising, PCs) are slowing, any expansion that consumes free cash flow will be harshly punished by the market;
• The "physical vulnerability" of hardware and infrastructure is being reassessed: whether it's the ripple effect of DRAM memory shortages on the PC supply chain, or the supply disruptions caused by factory explosions in the Middle East and physical threats to cloud nodes, all serve as reminders to investors that the deeper AI goes into the fundamentals, the more it is constrained by the physical laws of traditional manufacturing and geopolitics;

In January, we said that Intel pulling the oxygen tube from ICU and attempting to breathe on its own would require 12–18 months of execution to validate. Now that we’re halfway through, it’s clear this capital-intensive自救 effort is far from being cause for celebration.

For Amazon and Google, which also carry heavy infrastructure burdens, the real test has only just begun.