For years, the technology industry treated software as the highest layer of value creation. That assumption is now being challenged — not by a new SaaS category or a faster model, but by a structural shift in where future economic value will accumulate.
Recent reports tied to SpaceX suggest the company has framed a potential $28.5–29 trillion total addressable market spanning AI infrastructure, orbital computing, satellite connectivity, autonomous systems, and future space-enabled industrial economies. The headline sounds outrageous — until you realise SpaceX may no longer think of itself as a rocket company.
It increasingly looks like the early infrastructure layer for an AI-native civilization. And the more interesting question is not whether SpaceX deserves that number. It is what the existence of that number reveals about where economic value is migrating.
The constraint — models are not the bottleneck.
The market is starting to absorb something the technical community has known for months: large language models are likely to commoditize faster than expected. Capability gaps between frontier models are narrowing. Open-weight models are closing in on closed ones. Inference costs are falling on a quarterly cadence.
If models are not the constraint, what is? The honest answer reorders the strategic landscape. The real bottlenecks are now compute, energy, distribution, data, and physical infrastructure. That changes everything.
The winners of the next decade may not simply own the best models. They may own the infrastructure stack underneath intelligence itself — the layers that determine whether intelligence can scale, where it can run, who it can reach, and what work it can actually do.
The structure — the AI industrial stack.
In Issue No. 008 — The AI Industrial Stack — we set out a seven-layer model of the intelligence economy, extending Jensen Huang's framing with structural additions that account for how AI value compounds through industries, not just through models. That stack is the lens we apply here. Every meaningful AI-economy play is, in effect, a bet on one or more of these layers.
The point of this structure is not theoretical. It is that each layer has different economics, different defensibility, and different incumbents. A company can win at Layer 3 (Intelligence) and still lose value to whoever orchestrates Layer 5 (Industry GTM). A company can dominate Layer 1 (Compute) and have no exposure to where the largest GDP contributions ultimately accrue.
This is the lens that makes SpaceX's $29T claim legible.
The evidence — SpaceX, mapped to the stack.
Most observers still categorize SpaceX as a launch provider, a satellite company, or an aerospace manufacturer. That view is now structurally out of date. Read against the seven-layer stack, SpaceX is vertically integrating across multiple layers simultaneously — and unlike pure-play software firms, it is doing so in physical infrastructure where moats are harder to copy.
| Stack Layer | What SpaceX Is Building | Strategic Read |
|---|---|---|
L1 · COMPUTE Physical Infrastructure |
Launch economics for orbital data centres; long-term orbital energy potential. | Cost-per-kg-to-orbit collapse opens space-based compute as an economic option for the first time. |
L3 · INTELLIGENCE Autonomy & Robotics |
Reusable autonomous launch systems; high-volume industrial production capability. | Operational AI is already embedded in the core business — not a future bet. |
L5 · INDUSTRY GTM Global Distribution |
Starlink — a planet-scale connectivity network with millions of endpoints. | The distribution layer for any future AI economy that requires real-time global access. |
L6 · INDUSTRY Earth-to-Orbit Logistics |
Reusable transportation infrastructure; manufacturing and supply-chain integration. | A new industry being created — not an existing industry being entered. |
L7 · GDP Planetary Expansion |
Moon and Mars industrialisation as long-term economic frontiers. | The optionality on entirely new economic surfaces — the source of the headline TAM number. |
Whether all of this happens in ten years or thirty is secondary. The strategic positioning is already visible. SpaceX is not making a claim about rockets. It is making a claim about which layers of the future economy it intends to own. And that is the move worth studying — because it is a template.
The distribution lesson — Starlink is the most important asset.
Many investors still see Starlink as a side business. In reality it may become one of the most strategically important digital infrastructure assets in the world — for one specific reason. AI economies require constant connectivity, real-time data transfer, autonomous coordination, and globally distributed compute access. Starlink is the only privately-owned network that operates at planetary scale and can plausibly serve that demand.
Most AI firms today still depend on infrastructure owned by cloud hyperscalers, telecom providers, or governments. SpaceX increasingly owns the stack itself. That level of vertical integration is difficult to replicate — not because competitors are not trying, but because the time and capital required cannot be compressed.
The lesson is broader than space. As AI models become more accessible, differentiation moves elsewhere — toward proprietary workflows, first-party industry data, embedded distribution, operational integration, and ecosystem control. This mirrors every previous technology cycle. The biggest winners were rarely just technology creators. Amazon built logistics infrastructure. Microsoft built operating systems. Google built internet-scale data infrastructure. SpaceX is building the physical infrastructure for a future intelligence economy. The pattern is the same: the most defensible position is not at the technology layer — it is at the layer that distributes and orchestrates that technology across industries.
The pattern — what an AI-native industry actually looks like.
SpaceX is the extreme case — a single company moving across five of the seven layers in physical infrastructure. But the more important insight is not about SpaceX. It is about the pattern. The next wave of trillion-dollar companies may not look like traditional SaaS firms at all. They will look like entire industries restructured around AI-native operating models.
In each of these environments, AI agents perform operational work, autonomous systems coordinate decisions, infrastructure becomes intelligent, and industries operate through continuous machine-to-machine orchestration. Software stops being a product. It becomes embedded into the industrial fabric itself.
This is a fundamentally different economic model from the SaaS era. SaaS sold a license to a tool. AI-native industries own the operating layer of a vertical. That is a different category of company, with different unit economics, different moats, and different terminal values.
The implication — for the board, not the IT department.
Most companies still think they are adopting AI tools. That framing is a category error. The larger shift is that industries themselves are becoming AI-native. The strategic question is no longer which AI products do we buy. It is which layers of our industry do we intend to own as it restructures.
That means future market leaders may increasingly own intelligence layers, operational workflows, ecosystem distribution, and industry-specific data loops — not just software licenses. The companies who only buy AI tools become the ones whose margins get compressed by the companies who build AI-native operating layers around them.
This is the deeper implication behind the $29T number. It is not really a statement about rockets. It is a statement about the scale of the future intelligence economy — and about who will own the infrastructure beneath it.
SpaceX is one answer. The next ten trillion-dollar companies will be others. Most of them will not exist yet. The ones that emerge will share one characteristic: they will not have sold software. They will have orchestrated industries.