Big Tech just reported earnings and the numbers are genuinely staggering. Amazon, Microsoft, Meta, and Alphabet collectively spent more than $130 billion on capital expenditures in Q1 alone, almost all of it on AI infrastructure. Their full-year forecasts now project up to $725 billion combined. Microsoft is committing $190 billion in 2026. Meta raised its number to $145 billion and the stock dropped 7 percent in after-hours trading. Amazon is holding at $200 billion. Google raised its forecast too. This is not the AI hype cycle anymore. This is the AI industrial buildout. Trillions of dollars are being committed across the largest companies on earth based on a bet that AI capability translates into AI revenue at a pace that justifies the spending. The question every honest investor and every honest engineer is asking tonight is the same one. Is this real, or is this the biggest bubble in tech history forming in real time? Tonight we look at both sides.

Here is what each company actually committed.

Microsoft: $190 billion in 2026. Microsoft's first 2026 capex guidance came in at $190 billion, matching Alphabet. Most of it goes to data centers, GPUs, and AI infrastructure. CEO Satya Nadella pointed to rising adoption of Microsoft 365 Copilot but acknowledged enterprise adoption remains uneven. The market read it as Microsoft betting heavily on AI even after loosening parts of its OpenAI partnership.

Meta: Raised to $125-145 billion. Meta posted 33 percent revenue growth this quarter, the fastest since 2021. Stock dropped almost 7 percent anyway. The reason is the new capex forecast. $145 billion at the high end is roughly double what Meta spent last year. Investors flagged the "asymmetric risk" of pouring that much money into superintelligence research and Reality Labs while ad revenue is the only thing actually paying for it.

Amazon: $200 billion held steady. AWS posted strong AI growth and the stock ended green. Amazon is the calmest of the four because AWS is already monetizing AI through cloud revenue. Analysts are watching whether AWS growth keeps pace with the spending.

Alphabet: Raised forecast. Google's capex jumped meaningfully tied to TPU buildout and Gemini infrastructure. They announced a $40 billion commitment to Anthropic on top of all this last Friday. Google is essentially making two simultaneous AI bets, one on Gemini and one on Claude.

The total: Up to $725 billion in 2026. Combined across the four companies. For context that is more than the entire US defense budget. More than Apple's annual revenue. More than the GDP of most countries on earth. All of it spent in twelve months. All of it bet on a single thesis: AI capability becomes AI revenue.

The case that this is not a bubble is actually strong.

ChatGPT crossed 900 million weekly users last month. Anthropic's annualized revenue topped $30 billion. AWS, Azure, and Google Cloud all posted double-digit cloud growth driven primarily by AI workloads. Microsoft 365 Copilot, GitHub Copilot, and enterprise Claude deployments are all generating real, recurring, billable revenue at scale. This is not 1999 with no users and no revenue. This is users in the billions and revenue in the tens of billions.

The infrastructure is also being built for products that already exist and already work. Coding agents finish real software engineering tasks. Cybersecurity AI finds real vulnerabilities. Customer service AI handles real support tickets. Productivity AI writes real documents. The use cases are not theoretical. They are deployed.

And the strategic logic is clear. Compute is the bottleneck for everything. Whoever has the most compute trains the best models. Whoever trains the best models wins the most enterprise contracts. Whoever wins the most enterprise contracts captures the recurring revenue. Spending $725 billion now to lock in compute supply for the next decade is rational if you believe AI demand keeps growing at current rates.

The case that this is a bubble is also strong.

Meta's 7 percent stock drop after raising capex is the market sending a message. Investors are starting to question whether the spending is rational. Reality Labs has lost tens of billions over multiple years with minimal product traction. The same Meta is now putting $145 billion into AI infrastructure based on a similar bet that the technology will eventually generate proportional returns. That is a lot of investor faith in a company that has spent heavily before without delivering.

The Harvard Business Review analysis from this week landed at an awkward time. 59 percent of organizations have AI in production but only 16 percent report any real measurable value from it. That gap matters. If three out of four companies cannot prove ROI from their AI deployments, the demand assumptions underlying $725 billion in infrastructure spend get harder to defend.

Capex this large also creates fragility. If growth slows even slightly, the depreciation on that infrastructure becomes catastrophic to earnings. Data centers have 20-year useful lives but the GPUs inside them depreciate in 3-5 years. Every Q1 of capex needs Q2, Q3, and Q4 of revenue growth to justify it. Any slowdown breaks the math.

And the historical pattern is uncomfortable. Telecom in 1999 spent unprecedented capital on fiber networks based on assumptions about internet traffic that turned out to be roughly right but on a much longer timeline than investors could absorb. Most of those companies went bankrupt anyway. Being right about the technology and right about the timing are different things.

The defining feature of this moment is that everyone is forced to bet on the same outcome at the same time.

Microsoft cannot afford to spend less than Google. Google cannot afford to spend less than Amazon. Amazon cannot afford to spend less than Meta. Meta cannot afford to spend less than Microsoft. The competitive dynamic compels every Big Tech CEO to raise their capex even if they are unsure about the demand picture. Falling behind on AI infrastructure is treated as an extinction-level threat. So everyone keeps raising. The number grows quarter over quarter not because demand has been proven but because the cost of not spending feels worse than the cost of spending.

This is what makes the question interesting. It is not "will AI work" because AI is clearly working. It is "will AI work at the pace and scale required to justify $725 billion this year and likely north of $1 trillion next year." The honest answer is that the bet is being placed by the smartest companies in the world based on the best information they have. They might be right. They might be wrong. They are absolutely not stopping.

For users this means more capable AI faster than anyone planned. For developers this means cheaper compute and better tools as cloud providers fight for market share. For investors this means higher volatility as quarterly earnings get measured against capex commitments that compound. For everyone watching the AI industry this means we are now in the part of the cycle where the financial decisions get bigger than the technical ones.

Big Tech just committed up to $725 billion to AI infrastructure in 2026. Microsoft at $190 billion. Amazon at $200 billion. Meta at up to $145 billion with the stock dropping 7 percent on the news. Google raising its forecast and adding a $40 billion bet on Anthropic on top. The bull case is real. ChatGPT has 900 million weekly users. Anthropic is at $30 billion in revenue. Cloud growth is real. The bear case is also real. Most enterprises cannot prove ROI on AI yet. Capex this size creates earnings fragility. History does not love spending this concentrated. Whether this is the largest infrastructure buildout in tech history or the biggest bubble forming in real time depends on whether AI revenue grows fast enough to absorb $725 billion in twelve months and likely double that the year after. Nobody knows yet. But the bet is being placed right now and the answer comes in the next four quarters. Tonight we asked the question honestly. The verdict is up to the market.

Stay building. 🤖