Fabless design is evolving alongside manufacturing.

1. The fabless bargain

Specialist foundries let designers avoid direct factory ownership while retaining access to increasingly capable processes. That specialization widened participation, improved focus, and transferred a large class of asset risk to manufacturers. The model remains productive.

The enduring advantage is flexibility: a designer can rely on a supplier’s capacity while focusing its own resources on products, software, and customers. When a qualified layer becomes strategic, allocation, duration, and shared commitment become part of the partnership.

2. Shadow Capital

“Shadow Capital” is a research taxonomy for manufacturing exposure that reported CapEx alone does not capture. It includes contractual capacity, firm purchases, prepayments, strategic inventory, facility guarantees, and manufacturing investments. The framework keeps these categories distinct so their different accounting and strategic roles remain clear.

The useful comparison is not “commitments equal factories.” It is that a low owned-asset line can coexist with long-duration support required to ship the product. Companies can build meaningful access to capacity while their partners retain the physical asset, learning curve, subsidy, or residual option.

3. Contracts define resilience

Capacity agreements become most informative when regulation, product schedules, or demand change. Allocation rights, cancellation terms, product-specific inventory, and excess purchase obligations show how resilience and downside are shared.

4. The expanded fab

A working system depends on leading-edge and mature-node wafers, memory, packaging, substrates, test, optics, power, cooling, and integration. The strategic “fab” is therefore the narrowest qualified layer that can stop the complete product.

That is why customer-backed supplier expansion and public support for packaging matter to the thesis. They show capital following the bottleneck even when the buyer does not become the asset owner.

5. AI infrastructure has many builders

Frontier-model demand now links model labs, cloud platforms, custom-silicon teams, network suppliers, power providers, and construction partners. Anthropic’s work with AWS and Google, OpenAI’s work with Microsoft, Amazon, NVIDIA, and Oracle, and Meta’s silicon partnerships all make the same point: the machine is built through a web of complementary commitments.

Those commitments should be read with care. An equity investment, a cloud-service contract, a planned gigawatt of capacity, a construction obligation, and a manufacturing purchase commitment can each move the physical system forward while remaining economically different records.

6. Ownership, control, and productive alignment

Ownership can preserve residual value, learning, allocation, and public-policy standing when the process is competitive and the asset is well utilized. When those conditions are absent, the asset can absorb capital without delivering the intended strategic leverage.

Choose the form of control that best fits the bottleneck: own it where ownership creates durable advantage, or secure allocation, portability, remedies, and shared incentives that help partners improve together.

7. A research horizon

The next chapter will be written through new evidence: how commitments evolve, where partnerships expand, how packaging and memory investment develops, and which hybrid models create durable capability. This publication will update its record as that evidence arrives.