Silicon Valley's Big Five Report Strong Earnings, But Market Reactions Diverge Sharply

This week, the five largest tech giants in Silicon Valley submitted their results in quick succession.

Google (Alphabet), Microsoft, Meta, Amazon, and Apple—these five giants, with a combined market capitalization exceeding $14 trillion, have a combined quarterly net profit nearing $150 billion.

From a financial performance standpoint, this was a collective triumph—all five companies exceeded Wall Street’s expectations, with none experiencing a business collapse.

But the stock price response reveals a nearly brutal disconnect.

After earnings reports, Google Alphabet's stock surged 10%, Apple rose 4%; meanwhile, Amazon's stock remained nearly flat, Microsoft fell 4%, and Meta plunged 7%.

A gap of 17 percentage points opened up between the highest and lowest. Ironically, all five companies were doing the same thing—making money—and all earned more than analysts had expected.

Over the past two years, Silicon Valley earnings seasons have followed an unspoken rule: as long as you claim to be all in on AI and are aggressively spending on chips and building data centers, the market will reward you. This is a premium based on the future potential of AGI.

But in the spring of 2026, this rule completely broke down.

To understand this divide, you must consider Google and Meta together.

Google's parent company, Alphabet, reported first-quarter revenue of $109.9 billion, a 22% year-over-year increase. Net profit reached $62.6 billion, a 81% year-over-year increase on paper, driven by approximately $36.9 billion in unrealized gains from non-publicly traded equity investments. Excluding this non-operational factor, core operating profit grew by around 18%. The stock rose 10% in after-hours trading.

Meta reported first-quarter revenue of $56.3 billion, a 33% year-over-year increase—11 percentage points faster than Google's growth. Shares fell 7% after hours.

The one with faster growth received a vote of opposition from capital.

What made Google win? It won because every AI investment points to a revenue stream that is already being realized.

Google Cloud's quarterly revenue surpassed $20 billion, a 63% year-over-year increase. Profit margin rose from 17.8% a year ago to 32.9%, not only growing faster but also becoming more profitable over time.

On the search side, query volume hit a new all-time high, and advertising revenue increased by 19% to $60.4 billion. Two years ago, everyone was saying AI would kill search engines, but the opposite turned out to be true—AI has made search more useful, increased usage, and encouraged advertisers to spend more. The total number of paid subscription users is 350 million. The cloud business backlog doubled from approximately $240 billion last quarter to $462 billion.

Wall Street sees a closed loop: spending to build AI infrastructure leads to rising cloud revenue, rising search revenue, and expanding profit margins. Money goes out, and the return on that investment is visible.

Meta’s underlying numbers are actually strong. The Advantage+ advertising system reduced customer acquisition costs by an average of 14%, and return on ad spend increased by more than 30% in some categories. With $56.3 billion in quarterly revenue and a 41% profit margin, these figures are top-tier by any company’s standards.

But the market is focused on something else. Meta raised its full-year 2026 capital expenditure guidance to $125 billion to $145 billion, an increase of $10 billion from its previous guidance last quarter.

Where is the corresponding increase in AI revenue after this money is spent? While advertising efficiency has improved, can that improvement sustain an annual investment of over $140 billion? The conference call did not provide an answer that satisfied the market.

The rules of the game have changed. Previously, if you were willing to spend money, I would increase your valuation. Now, there is only one question: How much of the money you spent has come back?

If Meta lost on the path to returns, then Microsoft lost on the ending of the story.

Revenue for the third quarter of FY2026 (ended March 31) reached $82.9 billion, an 18% year-over-year increase, significantly exceeding expectations. Azure grew by 40%. AI business annualized revenue exceeded $37 billion, up 123% year-over-year.

After hours, it dropped by approximately 4%.

The reason lies not in these numbers, but the day before the earnings report. On April 28, OpenAI’s GPT-5.5 was officially launched on Amazon AWS’s Bedrock platform.

Over the past two years, Microsoft’s AI narrative was built on the premise that it was OpenAI’s exclusive cloud partner, with the world’s most powerful AI models running only on Azure. This drove significant customer migration to Azure and supported Microsoft’s valuation premium in the AI era.

This premise has suddenly changed. OpenAI has signed a $38 billion, seven-year agreement with AWS for compute resources. GPT-5.5 can now be accessed on both Azure and AWS.

Microsoft's CFO mentioned on the earnings call: "We are no longer paying royalties to OpenAI."

On the surface, it appears to be stopping the bleeding, but Wall Street heard it as the largest AI partner moving away.

When compared horizontally, the difference becomes even clearer: Google Cloud’s growth rate of 63%, AWS’s 28% which is still accelerating, and Azure’s 40% placing it in the middle among the three. Azure itself isn’t bad, but in this context, it’s the one without any surprises.

Microsoft’s ability to generate profits is not in question. The problem is that the story supporting its valuation has begun to crack—its income statement can’t absorb the losses from that narrative.

Meanwhile, Amazon (AWS) has revealed another harsh truth of the AI era: the more you earn, the less you have left.

Amazon's cloud business, while accelerating, generated $37.6 billion in AWS revenue, a 28% year-over-year increase—the fastest growth in 15 quarters. The company's overall operating profit was $23.9 billion.

However, free cash flow over the past 12 months was only $1.2 billion, down from $25.9 billion a year ago—a 95% decline.

Management is candid: Free cash flow will be under pressure until production capacity begins to generate revenue and revenue growth outpaces capital expenditures.

This is asset inversion— to secure the $360 billion in computing power orders, Amazon must burn $500 million every day just to secure electricity and build data centers.

This asset-intensive battle has pushed Amazon into a high-level equilibrium, with the market choosing to remain flat, waiting for it to prove itself.

In this frenzied arms race, Apple has emerged as the most composed player.

The Q2 fiscal year 2026 results (ended March 28) reported revenue of $111.2 billion, a 17% year-over-year increase and a record high for the March quarter. Strong demand for the iPhone 17 series drove a 28% year-over-year revenue growth in Greater China. Service revenue reached $31 billion, a historical high.

It rose about 4% after hours. Strong earnings and a rising stock price seem only natural.

But if you think one step further, you’ll notice the difference: Apple announced a $100 billion stock buyback this quarter while also increasing its dividend.

In the same week that Google spent $190 billion, Microsoft spent $190 billion, Meta spent $14.5 billion, and Amazon spent $200 billion on AI infrastructure, Apple allocated $100 billion to repurchase its own shares.

Apple is barely involved in this AI arms race—it doesn’t build data centers, doesn’t train large models, and relies on partners and on-device small models for its AI features.

Over the past year, Wall Street has claimed Apple is falling behind in AI. But this week’s stock price reaction told a different story: when everyone is pulling out, the one holding onto resources feels most reassuring.

Apple can't remain entirely unaffected. TSMC’s 3nm capacity has been squeezed by AI chip demand, leading to tight supply of the A19 chip and insufficient iPhone production. Rising global memory prices have increased hardware costs, and Apple expects its gross margin for the next quarter to decline from 49.3% to between 47.5% and 48.5%.

Cook's exact words on the call were: "After the third fiscal quarter, the impact of memory costs on the business will continue to increase."

This is the interconnected cost of globalization—even if you’re not directly involved in the conflict, you still pay for the expensive ammunition.

Looking at five earnings reports stacked together reveals something larger than any single company's quarterly figures.

In 2026, the combined full-year capital expenditure guidance for Google, Microsoft, Meta, and Amazon approaches $700 billion.

Estimating based on the upper end of each company's guidance range: Google is approximately $185 billion, Microsoft $190 billion, Meta approximately $135 billion, and Amazon over $200 billion. Two years ago, the combined total was approximately $245 billion, nearly tripling since then.

$700 billion exceeds Israel's entire annual GDP. Four companies spend more on AI infrastructure in a year than most countries create in total wealth annually.

But this week's stock price shows that spending alone is no longer a good sign.

Over the past two years, the market has rewarded belief—you're investing in AI, you believe in the future, so I'll give you a valuation premium.

This week, focus on your verification score—it’s not how much you invested that matters, but what your invested funds have become.

Google rose because its $190 billion in capex is already translating into growth in cloud revenue, advertising revenue, and profit margins—measurable, visible, and quantifiable.

Meta fell because the path to returns on its $145 billion investment remains unclear. Microsoft fell because the narrative supporting its AI valuation has begun to crack.

Amazon stood still, as the market waits for it to prove that earning less while spending more is only temporary. Apple rose, because amid everyone spending heavily, it still had the capacity to return money to shareholders—appearing the most like a company operating normally.

AI investing has moved from the belief phase into the validation phase. This threshold was officially crossed this week.

This week's five earnings reports are about more than just five companies' quarterly results.

At the beginning of 2024, the world asked whether AI was a bubble. In 2025, the question became whether AI can generate profits.

By spring 2026, the question had shifted another layer: AI can certainly make money, but who is making that money?

In the same week, OpenAI released GPT-5.5, tripling its API pricing. DeepSeek launched V4, reducing all prices to one-tenth of the original. Anthropic’s Claude Mythos was internally assessed as too powerful and remains unreleased.

Three AI-native companies, three entirely different paths.

But one thing is common: all three rely on someone else’s infrastructure. OpenAI runs on Azure and AWS, Anthropic runs on AWS, and DeepSeek depends on chips from NVIDIA and Huawei. Model companies drive innovation; infrastructure companies foot the bill.

The logic of the gold rush 150 years ago is playing out again: those who dig up gold may not make money, but those selling water and jeans definitely will. Today, water is computing power contracts, and jeans are data centers.

The difference lies in the cost. This round of mining equipment is far more expensive than it was 150 years ago—$700 billion per year, and accelerating.

Morgan Stanley estimates that U.S. data centers alone will face a power shortfall of approximately 55 GW between 2026 and 2028. In the first quarter of 2026, global large model weekly token consumption surged from 6.4 trillion at the start of the year to 22.7 trillion, a 250% increase in a single quarter.

Meta’s planned data centers aim for a combined capacity of 6 GW, while OpenAI’s Star Gate project targets 10 GW over four years. When these companies place orders, the unit has already shifted from dollars to gigawatts.

The barriers to AI competitions are shifting. Algorithms can be open-sourced, and chips can be purchased, but power capacity, data center location, cooling systems, and long-term power supply contracts cannot be downloaded or copied—building one takes at least two to three years.

For the past two decades, Silicon Valley changed the world through code. But in the AI era, the barrier to competition is shifting from code to concrete and copper wires.

The tech industry is undergoing a fundamental shift: from light to heavy, from soft to hard, from writing to building.

Whoever builds the infrastructure first will control pricing for the next decade. This was true for railroads 150 years ago, and it’s true for computing power today.

All five companies made profits and exceeded expectations, but the market only rewarded one and a half: Google and Apple. One provided evidence that spending yields returns, while the other demonstrated that it can thrive without spending.

The other three haven’t done anything wrong. Meta’s ads are strong, Microsoft’s Azure is growing, and Amazon’s AWS is accelerating. But “still growing” today, which two years ago would have meant a limit-up, now only means not falling.

This is probably the most memorable thing of this earnings season. When everyone is growing, growth is no longer scarce. What becomes scarce is something else: evidence.

Prove that every dollar you spend is returning to your pocket in a predictable and verifiable way.

This article is from the WeChat public account "Beyond the Layout," authored by Hua Hua.