AI Everywhere, But Not All at Once

I started the morning assuming AI was one lane: GPUs & chips lead, the rest of the market follows. In other words: I expected to see continued investment in AI supporting growth - but things are changing.

By the close this morning, my view had been stretched. I keep circling back around the idea of "continued investment in AI supporting growth" is all we see. Sounds obvious but the details complicated things, it's getting a bit blurry out there, but in a good way.

On chips, I am watching this play out as the conversation shifts from a single supplier story to more of a market mosaic. Isn't this how things should be anyway? To avoid monopolies? Shall I go here? Let's not.

But investors are now asking themselves: what's happening with some of the demand from hyperscalers for specific application chips as well as customized silicon chips now? Again, this points me away from standardization & toward a mix of merchant GPUs & bespoke accelerators. The Nvidia narrative feels more balanced through this lens & less spiky. Further it is most likely the competition with Google & their TPU chips is legitimate competition for Nvidia. At the same time, I am also seeing that we are compute limited outside of this market segment regardless; there's more demand than supply, so buyers will continue to pick Nvidia while also looking anywhere they can get compute. Thankfully for many in executive & leadership positions, they can all coexist when the choke point is capacity not demand.

If we zoom down, software makes the theme feel tangible. With MongoDB, there looks like there was a blowout quarter & raised guidance & as we know, investors go where AI is being monetized quickly. I'm also taking in that demand for the database software maker is outstripping supply &this represents an inflection point for AI. Usage growth tied to guidance turns AI from a buzzword into booked revenue, not just a halo from hardware.

Cyber & security sounds less structural & more emerging within the AI bubble. There are investors looking to find ways to connect robots, data centers & HPC into one requirement: robotic infrastructure, AI infrastructure, certainly high performance computing (HPC) infrastructure - all this requires cyber secure environments & walled gardens to operate within. That framing means we should anticipate in 2026 that cybersecurity becomes more essential & core across copmany BOM's for wider AI adoption. If the attack surface expands, the budget follows - essentially.

Industrials will pull in this direction. Maybe I am wrong but more industrial tech companies will work together on AI powered factory robots, which can respond to verbal commands & can work around humans safely (e.g., Siemens & NVIDA in addition to leaders in Asia/China). Verbal commands & human-safe workflows for AI enablement & augmentation will again require new capital plans, safety validation & integration cycles that won't match legacy cloud setups & cadence. It also loops back to power & networking footprints that utilities & data centers must scale up accordingly.

By the end, I no longer see one AI trade; I see a stack with distinct cash flows, risks, & clocks. Custom silicon chips grows alongside already established & "merchantized" CPU/GPUs. Software then monetizes consumption, usage & raises forward guidance. Security becomes even more mandatory. Industrial AI deployments follow their own timelines. The throughline still holds - continued investment in AI supporting growth but the dispersion inside that sentence is the real story. That's what I'm watching next: who books revenue now, who scales later & who must spend simply to stand still.