Why are NVIDIA’s robots sitting like this? The answer is alarming. (from Behind the Markets)
Alphabet plans to spend up to $185 billion on AI in 2026.
It’s a structural reset.
The spending isn't slowing down. It's accelerating.
Alphabet's earnings call demolished what remained of the "capex peak" narrative. The company isn't pulling back on infrastructure investment, it's doubling down. And it's not alone.
Tech companies will deploy an estimated $562 billion in capex during 2026, rising to $637 billion in 2027.
For context, the same companies spent $256 billion in 2024. We're witnessing a 119% increase in two years.
This isn't a temporary spike. It's a permanent shift in how these companies allocate capital.
The Numbers That Matter
The median capex-to-sales ratio for Big Tech has vaulted from approximately 10% pre-2023 to above 20% today. That's not an adjustment. It's a fundamental re-rating of what it costs to compete in AI infrastructure.
The asset-light software model that defined the 2010s? Paused indefinitely.
Big Tech now resembles industrial infrastructure utilities more than pure software platforms.
The capital intensity profile has converged with sectors historically considered "heavy," telecommunications, energy distribution, data centers.
But here's the thing: This spending is generating returns.
Alphabet's Massive Signal
Alphabet's recent guidance wasn't just about maintaining competitiveness.
Google Cloud ended 2025 at an annual run rate of over $70 billion, representing a wide breadth of customers, driven by demand for AI products.
2026 capex investments are anticipated to be in the range of $175 to $185 billion.
The company is pushing infrastructure spend to support Gemini's deployment, Search modernization, and YouTube's AI integration. Management made it clear: under-spending is an existential risk. Falling behind in model capability or inference speed means losing search queries, ad impressions, and cloud contracts permanently.
Meta pre-announced similar increases for 2025/26.
Microsoft and Amazon have signaled comparable trajectories, though with more geographic dispersion (data center buildouts in emerging markets, sovereign cloud requirements).
The Street's reaction? Mixed.
Near-term margin compression is real. Depreciation costs are rising as assets come online. Operating leverage deteriorates when capex growth outpaces revenue growth by 3-4x.
But investors also understand the alternative.
Companies that underspend now face a compounding disadvantage. Model training costs are falling, but the scale required to compete is rising faster.
You either build the infrastructure or you rent it, and renting means ceding margin and control.
Which Big Tech capex strategy concerns you most in 2026?
Where's the Payoff?
This is where 2026 diverges sharply from the dot-com era.
Return on Invested Capital (ROIC) for the Big Tech has held above 25% through Q3 2025. Its measured performance on deployed capital.
Compare that to the late 1990s, when telecom and internet infrastructure spending generated near-zero returns for most participants. Fiber got laid. Capacity went unused. Balance sheets imploded.
Current AI capex is funding infrastructure that's already showing utilization. Cloud revenue growth remains in the high teens. Enterprise AI adoption is accelerating. Model API calls are scaling exponentially.
The earnings floor supports this. Big Tech consensus projects 22% EPS growth in 2026/27 versus low-teens growth for the broader S&P 500. Free cash flow for tech companies is expected to remain above $400 billion annually, meaning the spending is self-funded.
As long as ROIC stays above 20% and FCF remains robust, the capital deployment is mathematically rational. These companies aren't burning cash.
They're reinvesting operating profits into assets that generate measurable returns.
2026 Playbook
The trade is shifting.
2023-2025 was about "Who buys the GPUs?" Infrastructure acquisition defined competitive positioning. Nvidia's chips became the scarcity resource.
2026 is about "Who monetizes AI?"
Unit economics matter now. Companies need to demonstrate that inference costs are falling faster than pricing pressure from competition. Nvidia's upcoming Rubin platform targets a 40% reduction in cost-per-token vs current Hopper architecture. That's the enabler for enterprise adoption at scale.
But there's an adoption gap.
Infrastructure deployment is running 12-18 months ahead of enterprise revenue realization. Data centers come online in Q1 2026. Enterprise software migrations happen in Q3-Q4 2026.
The lag creates near-term margin pressure and investor anxiety.
Separate the Hyperscalers from the Enablers from the Adopters.
Hyperscalers (Alphabet, Meta, Microsoft, Amazon): Funding the infrastructure build. Directly exposed to capex risk but also positioned for winner-take-most dynamics in cloud AI services.
Semiconductors (Nvidia, AMD, Broadcom, memory manufacturers): Selling the picks and shovels. Revenue visibility is strong but margin pressure emerges as supply normalizes.
Adopters (Enterprise SaaS, vertical AI applications): Monetizing AI capabilities without infrastructure burden. Lower capex but execution risk on product-market fit.
Watch the Q2 and Q3 2026 earnings cycles closely. That's when "Usage vs. Capex" ratios become the dominant narrative. Companies demonstrating pricing power and enterprise traction will re-rate higher.
Those showing infrastructure drag without revenue acceleration will face pressure.
What's the biggest risk to Big Tech's $562B capex thesis?
Bottom Line
2026 marks the year capital intensity becomes normalized in Big Tech.
The $562 billion annual capex run rate isn't a bubble. It's the new baseline. Companies that fail to maintain this pace will lose competitive positioning permanently.
AI infrastructure advantage compounds, early leaders gain data flywheel effects, ecosystem lock-in, and cost curve advantages.
Analysts suggest 30% YoY capex growth through 2028. If that materializes, we're looking at a $1.3 trillion cumulative infrastructure buildout over four years.
The investment thesis: Remain overweight on names demonstrating:
Deep ecosystem integration (proprietary model deployment, enterprise distribution, cloud platform effects)
Margin protection capability (pricing power, cost reduction roadmaps, operating leverage at scale)
Capital efficiency (ROIC >20%, self-funded capex, FCF generation despite heavy investment)
Capital intensity is the new reality.
The question isn't whether these companies will spend. It's whether they can convert infrastructure into durable competitive advantage and margin expansion.
The earnings data through 2027 will provide the answer.
Subscribe to ETF Alert for real-time market news.
We track the trends that move billions, before they hit mainstream headlines.
What's catching investor attention today: Trump's Fed Pick Sends Gold Crashing: What Investors Need to Know Now
Disclaimer: This is not financial or investment advice. Do your own research and consult a qualified financial advisor before investing.