Google, Amazon, Microsoft and Meta report tomorrow. Apple on Thursday. The market has been waiting for some sign of capex fatigue for years now. It hasn't happened.
And there are no signs of it happening when they report over the next couple of days.
Remember, it was just earlier this year that Google doubled its expected 2026 capex to as much as $185 billion. Meta guided up to $135 billion for the year, a near double of what they spent last year. Amazon announced $200 billion in capex.
And Andy Jassy said they were "monetizing capacity as fast as we can install it."
It's because the "always-on" inferencing phase of AI has arrived. Moreover, it shifted into an even higher gear since early February, as autonomous agents have exploded in numbers — running around the clock, executing tasks, managing workflows, transacting, calling other agents.
It's a demand explosion.
Every agent is consuming compute, power, bandwidth, and storage — continuously.
This is why the hyperscalers have said they can't build fast enough.
And maybe the cleanest signal on how it's all going, is from the storage industry.
In fact, if we look back to early February, we talked about the "Nvidia moment for storage" — the point in time when the world realized AI demand for data storage is endless. SanDisk had just reported $3 billion in revenue and then guided for $4.4 to $4.8 billion the next quarter (a 50% sequential leap over just 90 days).
With that, we heard from Seagate today, one of the data storage kings.
Revenue was up 44% year-over-year to $3.1 billion. Record gross margins. Almost a billion dollars of free cash flow in a single quarter — the highest level Seagate has reported in over a decade.
But here was the bigger takeaway: Seagate told us that the top three global cloud providers have nearly doubled their remaining performance obligations to $1.1 trillion.
That's the future revenue that enterprises and developers have already contractually committed to those cloud providers, primarily for AI infrastructure. And it's that number that's taking down Seagate's storage capacity all the way through calendar 2027.
They are sold out.
Seagate's CEO described what is happening as the "inference inflection." He said compute infrastructure is shifting from periodic training to becoming engines that continuously generate mass capacity data.
Agentic AI is the new demand driver (continuously ingesting inputs, generating reasoning, and storing output). Physical AI will take it to yet another level. AI robots create massive amounts of data that needs to be stored, retained, and reused, making storage demand continuous.
So, this sets the stage for tomorrow. If the top three cloud customers have committed $1.1 trillion in future revenue and are scaling AI capex into the infrastructure that fulfills it, then Google, Amazon and Meta should be guiding higher on capex, not lower.
They should be confirming that AI demand is accelerating, not stabilizing.
This is the thesis we've been building our AI-Innovation Portfolio around since June 2023. Endless demand drives outsized value in the companies that provide the scarce infrastructure resources in this fourth industrial revolution. Join us here, or click below for more details on how we're positioned and why.
