Software investments accelerate the brutal physical depreciation of industrial assets
The surface narrative across global industry today presents a tidy division of labor. On one side sits the intangible world of generative AI and software, commanding trillions in venture capital and sovereign initiatives. On the other lies the physical industrial base—automakers, telecom operators, and aerospace firms—wrestling with restructuring, plant closures, and volatile demand.
That division is an illusion.
Capital is not fleeing into software to build virtual playgrounds. It is fleeing heavy, slow-depreciating physical assets into code, only to find that advanced code instantly demands an even more brutal, immediate consumption of physical resources: silicon wafers, grid power, launch payloads, and specialized manufacturing tooling. What we are witnessing is not a transition to a digital economy, but an aggressive battle over who bears the burden of physical depreciation.
1. Whose Interests Are Served?
To understand multi-trillion-yen capital allocations into artificial intelligence, bracket the public relations statements about human potential and look at the cash flows.
When billions flow into frontier AI developers, the capital does not pool in software accounts. It is routed almost immediately into the balance sheets of semiconductor foundries (TSMC), chip designers (Nvidia), and utility-scale energy providers. The abstract code requires concrete assets: sub-3nm lithography, high-bandwidth memory packaging, and gigawatts of electrical infrastructure.
Who benefits? The asset owners at the absolute upstream bottleneck of physical physics.
Who loses? The mid-tier enterprises attempting to build downstream consumer software without proprietary hardware, and the legacy industrial operators who are burning working capital trying to subsidize both legacy tooling and digital transformation simultaneously. Capital is consolidating at the extremes: those who control the raw compute physics, and those who control end-user cash distribution. The middle layers are being hollowed out.
2. Why Now?
This friction is peaking now because the margin buffers that previously absorbed corporate inefficiencies have run out. The era of zero-interest-rate capital expenditure is over, and the first major waves of multi-year capital depreciation cycles from the post-pandemic era are hitting corporate balance sheets.
Consider low-Earth-orbit (LEO) satellite communications. A constellation like SpaceX’s Starlink operates on a harsh accounting reality: the satellites have an operational lifespan of roughly five years. The moment a Falcon 9 clears the pad, the depreciation clock starts ticking aggressively. A five-year asset life means roughly 20% of the constellation’s capital value must be amortized every twelve months.
To outrun that depreciation burden, an operator cannot wait for traditional enterprise sales cycles or negotiate ten-year municipal fiber contracts. They must capture steady, high-margin consumer cash flows immediately. SpaceX’s aggressive push into Direct-to-Cell service—bypassing traditional mobile network operators to connect directly to standard smartphones—is not an ideological crusade for global connectivity. It is a cash-flow imperative to service constellation churn before the hardware becomes orbital debris.
3. Constraints as Motivation
Corporate decisions that appear to be retreats or operational failures are almost always rational responses to severe balance sheet constraints.
Take Toyota’s recent decision to halt a planned next-generation EV program and pay tens of billions of yen in supplier cancellation compensations. To the outside observer, paying penalties for a canceled project looks like poor planning. But seen through the lens of capital preservation, it is a calculated containment of damage.
Building a dedicated EV line locks in billions in fixed assets that must be depreciated over seven to ten years. If that line runs at 30% utilization due to stagnant regional adoption, the structural drag on operating margins is catastrophic. Paying supplier indemnification is an immediate, one-time cash outflow. It stings in the current fiscal year, but it prevents an 8-year depreciation hemorrhage.
Volkswagen’s consideration of cutting up to 100,000 jobs stems from the exact same dynamic. It is not an admission that software-defined vehicles (SDVs) have failed, but an acknowledgment of a math problem: VW cannot fund the massive ongoing R&D expenditures required for centralized vehicle computing architectures while carrying the legacy fixed costs and pension liabilities of an internal combustion workforce.
Conversely, look at how software is being deployed defensively inside traditional manufacturing. Toyota’s implementation of generative models to draft vehicle technical specifications—reducing documentation lead times by 30%—is not about saving office supplies. In automotive development, ambiguities in technical specifications lead to downstream engineering rework. A defect caught during physical die-stamping or tooling testing costs tens of millions of dollars in scrapped steel and lost cycle time. Automating specifications is an insurance policy against the physical costs of downstream human error.
Compare this surgical use of software to the broad “consortium” model, such as Japan’s recent 44-company sovereign AI initiative. When companies assemble into massive, subsidized alliances to develop general-purpose foundational models, it is rarely driven by operational necessity. More often, it is an exercise in risk socialization for companies that cannot afford to build dedicated compute clusters on their own. These consortia primarily serve the cash flows of external system integrators and IT consultants, while diffusing the proprietary, tacit knowledge of the participating industrial members.
4. Grounded in Physics, Cost, and Institutions
The true test of this clash between physical scale and digital speed is taking place in Europe.
The European Union imposed tariffs of up to 35% on Chinese-built electric vehicles to protect domestic manufacturers. Under standard economic assumptions, an institutional barrier of that magnitude should neutralize foreign price advantages.
It hasn’t. Chinese manufacturers like BYD bypassed the tariff constraint through relentless physical vertical integration. BYD owns its lithium refining, its battery manufacturing, its semiconductor fabrication, and even its own dedicated roll-on/roll-off vehicle carrier ships.
A traditional automaker buys batteries with supplier markups, relies on Tier-1 systems integrators, and pays spot rates for oceanic shipping. BYD strips out every layer of external transactional margin. If a vehicle costs €14,000 to produce internally, a 30% tariff adds roughly €4,200. The imported vehicle can still retail in Germany at a lower price point than a European-built hybrid, while generating positive operating margin. Institutional protectionism cannot overcome a 40% structural advantage in manufacturing physics.
5. Falsifiable Conditions
This analysis rests on specific structural constraints. The hypothesis is invalidated if the following data emerges over the next 24 to 36 months:
- Legacy OEM Margin Resilience: Non-vertically integrated Western and Japanese automakers maintain operating margins above 8% while matching Chinese EV pricing in competitive neutral markets (e.g., Southeast Asia, Latin America) without government trade barriers.
- Consortium Efficiency: A multi-company, publicly subsidized AI consortium produces a commercial-grade foundation model that demonstrates lower cost-per-inference in specialized industrial applications than closed, vertically integrated enterprise models.
- Orbital Unit Economics Failure: Direct-to-cell LEO communication fails to offset constellation capital expenditures due to regulatory spectrum blocks and cell-site interference, resulting in Starlink scaling back launch frequency rather than expanding into commercial mobile billing.
6. Value Flows and Structural Shifts
The collision between capital-intensive physical infrastructure and rapid software iteration points to several clear shifts in corporate value capture:
- Winners: Upstream bottleneck owners (foundries, specialized power generators, raw material refiners) who supply indispensable physical inputs to both AI and electrification; and hyper-vertically integrated manufacturers who control their supply chains from raw atoms to final logistics.
- Losers: Mid-tier manufacturing assemblers who lack vertical integration and carry unhedged plant depreciation; terrestrial telecom network operators vulnerable to orbital bypass; and participants in broad, unfocused technology consortia whose capital is absorbed by intermediate system vendors.
In capital markets, these dynamics place structural pressure on the corporate credit and equity valuations of unhedged Tier-1 automotive suppliers and regional telecommunication infrastructure providers. Conversely, pricing power will increasingly concentrate in enterprise balance sheets that can fund high capital expenditure out of current cash flow without relying on debt markets.
7. The View from the Powertrain Line
As a systems engineer working on powertrain strategy inside a Japanese automaker, this dynamic is not theoretical—it is the weekly reality of our program reviews.
We balance vehicle mass against battery chemistry, analyze the raw kilowatt-hour cost of cell manufacturing, and fight for millimeter clearances in engine bays. From inside an automotive engineering division, software is not magic dust you sprinkle over a balance sheet to generate higher multiples. It is an engineering discipline that either reduces the physical mass and tooling scrap of a vehicle, or adds dead weight and warranty liability.
Watching capital markets behave as if software can permanently decouple from the laws of manufacturing physics is like watching a skier charge a backcountry slope without assessing the snowpack beneath. Code is fast, but the terrain is always physical. And the terrain always wins.
— Garryu
