Microsoft now operates three generations of internally designed AI accelerators, according to a Motley Fool analysis published this week. The milestone underscores how far Microsoft has pushed into custom silicon without loosening Nvidia's grip on the chip-investor pecking order.

The Fool's column, titled "Microsoft Now Has 3 Generations of Its Own AI Chips. Nvidia Is Still the Chip Stock to Buy," frames the development as Redmond progress rather than a competitive threat to Santa Clara.

What is Microsoft actually building?

The three-generation count suggests a sustained design cadence that began before the current AI build-out peaked. The Fool does not enumerate part numbers in its column, and Microsoft has not published a public roadmap for its in-house accelerator line.

Three design cycles imply Microsoft engineers have moved from early samples to a more mature tape-out rhythm. Sustaining such a programme requires dedicated verification teams, packaged-software investment, and chip-fabrication capacity that few companies outside the largest hyperscalers can underwrite. The Fool does not name a foundry partner.

Why does the Fool still favour Nvidia?

The investment thesis rests on scale. Custom silicon at hyperscalers typically targets a narrow workload envelope, while Nvidia sells a general-purpose platform that spans training, fine-tuning and inference across thousands of customer configurations. That structural difference is the crux of the Fool's reasoning for keeping Nvidia atop the chip-stock list.

The software ecosystem seals the argument. Even when hyperscalers ship their own accelerators in volume, Nvidia's installed developer base continues to define the economics of generative-AI training. Shifting an entire fleet off that software is a multi-year effort, and most enterprise customers never attempt it.

What does it change for the supply chain?

Practically, not as much as the headline suggests. Microsoft's custom parts absorb a fraction of its AI compute demand, leaving external suppliers as the primary source for the most demanding training jobs. Nvidia's data-centre revenue continues to grow at high percentages, even as hyperscalers publicly celebrate their own silicon milestones. The two paths coexist more often than they collide.

A useful comparison is the handset-modem era, when phone makers invested in custom cellular silicon while continuing to buy external parts. The compute supply chain rarely becomes an either-or proposition at the system level.

Where does this leave investors?

The Fool's verdict is unchanged: Nvidia remains the chip stock to buy. Microsoft's three-generation tally is a tactical datapoint worth noting, but it does not yet constitute an earnings event for the chip giant or its foundry partners.

Watch Microsoft's next quarterly disclosures for any commentary on AI-chip supply allocation. A meaningful shift toward in-house silicon would compress Nvidia's addressable market inside Redmond, and that would show up in capex disclosures or supply-chain checks before it surfaces in the stock price. Until then, the GPU leader's position looks intact.