Microsoft earns 17 cents in revenue for every $1 it spends on AI. Four companies - Microsoft, Alphabet, Meta, Amazon - will collectively spend $645 billion on AI in 2026. And tomorrow, April 29, they have to stand before markets and answer: what is all that money actually doing?
This is the most consequential earnings week in tech this year.
$645 Billion: The Largest Bet in Technology History
To put this number in context: $645 billion exceeds Sweden’s entire GDP. And this is the spending of just four companies within a single year.
The 2026 breakdown (LiveNewsChat, 2026):
- Amazon: $200 billion - the largest single spender
- Alphabet (Google): $175-185 billion
- Microsoft: ~$146 billion
- Meta: $115-135 billion
This represents a 74% increase from the $381 billion these four companies spent in 2025 (BlockNow, 2026). To fund this pace, all four have issued over $400 billion in long-term debt - nearly triple last year’s $165 billion. Even the world’s most profitable companies are leveraging up to stay in this race.
As Tomasz Tunguz of Theory Ventures put it: “You’ve had these cash-generating machines. And now, all of a sudden, they need that cash.”
The ROI Question: Markets Are Running Out of Patience
The central question entering this earnings cycle is whether spending is translating into proportional revenue.
Microsoft’s numbers tell a clear story: AI-related revenue is tracking toward roughly $25 billion for the fiscal year - translating to ~17 cents of revenue per dollar of AI capital spending (LiveNewsChat, 2026). For a long-horizon infrastructure bet, that ratio isn’t alarming on its own. But markets want to see acceleration, not just commitment.
Alphabet faces a more structurally uncomfortable tension: AI Overviews in Google Search are actively cannibalizing traditional ad clicks. Google is simultaneously spending hundreds of billions to build AI while that same AI erodes its core business model. That’s a difficult narrative to package for investors.
Meta’s $115-135 billion spend lacks a direct B2B AI revenue line. Most of the expected return is priced into advertiser performance improvements - a long-dated bet with no clean quarterly metric to point to.
Gil Luria of DA Davidson framed the strategic logic: “The four companies see the race to provide AI compute as the next winner-take-all market. But there are and will be bottlenecks.”
Borrowing to Build: When Big Tech Has to Leverage Up
One underreported detail: to fund this capex surge, Big Tech is issuing debt at historically unusual scale. Over $400 billion in 2026 (BlockNow, 2026), with approximately 90% of operating cash flow being recycled back into infrastructure rather than buybacks or dividends.
This has two meaningful implications.
First, this is no longer “spending excess cash flow.” This is a strategic bet carrying real balance sheet risk - if AI adoption doesn’t accelerate fast enough, these companies will face difficult capital allocation conversations.
Second, when the entire industry simultaneously bids for the same infrastructure - GPUs, data centers, energy contracts - supply tightens and prices rise. This is why each company’s capex guidance keeps climbing every quarter: no one wants to be the one who blinked in a potential winner-take-most market.
What the Compute Race Means Beyond Silicon Valley
While Big Tech burns through $645 billion to build infrastructure, something important is quietly being engineered: AI compute will become a commodity.
Electricity was once the exclusive advantage of large industrial operators. Then it became something you plug into a wall. AI compute is following the same trajectory - just faster.
When Microsoft, Google, and Amazon pour $645 billion into infrastructure, the output isn’t just corporate earnings. The output is progressively cheaper, more powerful, more accessible AI for everyone downstream. Businesses in Vietnam, Southeast Asia, and every market outside the hyperscaler budget range don’t need to “win” the compute race. They need to be ready for a post-abundance world: when AI becomes as cheap as bandwidth, competitive advantage shifts entirely to execution - how fast you experiment, adopt, and integrate.
Worth watching in parallel: OpenAI - not in the Mag-7 - is projecting $2.5 billion in ad revenue in 2026 and $100 billion by 2030 (eMarketer, 2026). If those numbers materialize, the entire structure of digital advertising that most marketers currently operate within will need fundamental reassessment.
April 29 will reveal whether markets still have patience for the AI infrastructure story, or whether they’re beginning to demand harder proof. Either way, the compute race doesn’t pause for earnings calls - and its implications are arriving faster than most realize.
NateCue's Take
The real question isn't "can Microsoft justify $146B." The question is: when $645B of compute capacity floods the market, who actually benefits? History has a consistent answer: the winners in infrastructure races aren't the ones who build the rails - they're the ones who figure out what to ship on them. The Big Tech compute race is engineering something important for everyone else: cheaper, stronger, more accessible AI. For marketers and businesses in Vietnam - and anywhere outside Silicon Valley's budget range - stop waiting to see who "wins" the compute race. Competitive advantage in the next 2-3 years won't come from access to AI. It'll come from the speed at which you learn to use it. That window is open now.