$175 billion. That’s the lower end of Google’s capital expenditure guidance for 2026, and it represents one of the largest infrastructure bets in corporate history. When numbers reach that scale, investors start asking hard questions about which companies can actually survive the AI arms race.
Doug Clinton, CEO of Intelligent Alpha, has a clear answer. In April 2026, he named Nvidia and Google as the “safest AI bets” available in public markets. Coming from someone who runs an investment firm focused on disruptive technology, that’s not a casual observation.
The Bot Builder’s Perspective
I build bots for a living, which means I spend my days elbow-deep in API calls, model endpoints, and infrastructure costs. When someone calls a stock “safe” in the AI space, my first instinct is to laugh. This sector moves fast enough to give you whiplash. But Clinton’s picks actually make sense when you look at them through the lens of what bot builders need.
Nvidia controls the picks and shovels. Every training run, every inference cluster, every edge deployment eventually traces back to their chips. I’ve watched GPU availability dictate project timelines more times than I can count. When your business model depends on hardware that’s perpetually backordered, you start to understand market position in visceral terms.
Google’s position is different but equally entrenched. They’re not just selling cloud compute—they’re running some of the largest production AI systems on the planet. Search, YouTube recommendations, Gmail’s smart features, Google Photos organization. These aren’t demos. They’re battle-tested systems processing billions of requests daily.
What “Safe” Actually Means
That $175 to $185 billion capital expenditure range for 2026 is the elephant in the room. Sundar Pichai is betting the company’s cash flow on AI infrastructure at a scale that would have seemed absurd five years ago. This is either visionary or reckless, depending on whether the AI economy materializes as predicted.
But here’s what makes it a “safe” bet in Clinton’s framework: Google can afford to be wrong. They have revenue streams that predate the current AI boom. Search advertising still prints money. Cloud services have real customers paying real bills. If AI hype deflates tomorrow, Google doesn’t disappear.
Nvidia has similar cushioning. Data centers need GPUs for more than just AI training. Gaming isn’t going anywhere. Professional visualization, scientific computing, cryptocurrency mining—these markets existed before ChatGPT and they’ll exist after whatever comes next.
The Infrastructure Layer Always Wins
I’ve built bots on top of OpenAI, Anthropic, Cohere, and a dozen smaller providers. The model layer is where all the excitement lives, but it’s also where companies burn through funding at terrifying rates. Training costs are astronomical. Inference costs add up fast at scale. Customer acquisition is expensive when everyone’s fighting for the same enterprise deals.
Meanwhile, Nvidia sells to all of them. Google Cloud hosts workloads for competitors. The infrastructure layer captures value regardless of which specific AI application wins.
This is why Clinton’s assessment resonates with builders. We see the cost structure from the inside. We know which expenses are compressible and which ones aren’t. You can optimize your prompts, cache aggressively, and batch requests cleverly. But you can’t train a large language model without serious compute, and you can’t get serious compute without going through a very short list of vendors.
Risk Still Exists
“Safest” doesn’t mean “safe.” That capital expenditure number should keep investors awake at night. If AI adoption stalls, if regulation kneecaps the industry, if a new computing paradigm emerges—these companies will have spent hundreds of billions on infrastructure that generates disappointing returns.
But if you’re going to bet on AI, betting on the companies that sell infrastructure to everyone else is probably smarter than picking which application layer startup will become the next Google. Clinton’s thesis isn’t that Nvidia and Google are guaranteed winners. It’s that they have the best risk-adjusted position in a sector full of uncertainty.
From where I sit, writing code that depends on this entire stack, that analysis tracks. The safest place to stand during a gold rush has always been selling shovels.
đź•’ Published: