Oracle’s media investments illustrate how AI leaders are vertically integrating to manage each compute provide and demand
Think about this: Oracle’s Larry Ellison invests in Paramount Footage and courts TikTok. At first look, it appears like a billionaire passion. As soon as aligned with Oracle’s multibillion-dollar AI infrastructure ambitions, nevertheless, the technique appears like a deliberate try to generate demand vertically.
Oracle is constructing huge AI information facilities with budgets rumored to exceed $100 billion.
Media manufacturing and social platforms corresponding to TikTok are compute-hungry and data-rich. They generate continuous, high-volume workloads.
Ellison is subsequently not simply shopping for content material. He’s engineering demand for his personal infrastructure.
By controlling each the provision of compute and the demand for it, Ellison goals to create a self-reinforcing system that shields him from the commodity entice that plagues pure infrastructure suppliers. This logic represents the spearpoint of how the AI economic system is diverging from the surroundings most companies inhabit.
From this angle, we will see why capital markets are behaving so unusually and why many investments might look a lot much less worthwhile in actuality than they do on paper.
Two diverging economies
The AI economic system vs. the legacy economic system
Does it really feel like each enterprise headline now options billions? Knowledge facilities, chip fabs, mannequin coaching runs, vitality vegetation flood the literature. Abruptly $10 billion sounds small, and $100 billion is the brand new critical.This isn’t the world that almost all companies inhabit.
We’re watching two economies diverge:
- The AI Financial system: Outlined by capital expenditures at a scale that used to belong solely to governments. The numbers are tens to a whole bunch of billions. Corporations like Nvidia, Microsoft, Meta, and Oracle are repriced not by their money flows as we speak, however by their potential to sit down on the middle of a remodeled manufacturing perform. Their inventory costs mirror perception that each unit of capital deployed into AI will ship extra output than the outdated economic system ever may.
- The Conventional Financial system: Manufacturing, retail, companies, and even mature tech. These corporations face compressed valuations, tight capital constraints, and personal fairness scrutiny over each greenback. They aren’t simply cyclical laggards. They’re being repriced as their future productiveness is discounted towards the potential for AI disruption.
From December 2022 onward, the efficiency hole has grown steadily wider. An “AI basket” of Nvidia, Microsoft, and Meta surged almost sevenfold from its base, whereas a “Legacy basket” of Walmart, Caterpillar, and Coca-Cola barely moved above its place to begin.
In distinction, the legacy names present a flat line. Their returns have been secure, modest, and largely disconnected from AI hype cycles. This quiet trajectory displays their function in mature industries that stay insulated from AI-driven revaluations.
The result’s a bifurcation in markets: two economies shifting on totally different curves. The AI economic system instructions premium multiples and capital inflows, whereas the legacy economic system continues to ship regular however unspectacular progress.
Constructing oil refineries to promote MP3s
Why would anybody commit $50 billion or $100 billion to AI infrastructure when the output (fashions and inference) is quickly commoditized? That is the paradox on the core of the AI economic system.
Huge fastened prices embrace information facilities, GPUs, vitality, cooling, networking, mannequin coaching, dataset acquisition, and software program engineering.
Speedy diffusion and near-zero marginal price then comply with. As soon as a mannequin structure is revealed, others replicate it. Open-source releases speed up cloning and fine-tuning. Competing fashions grow to be cheaper to coach as infrastructure improves.
In apply, companies are constructing enormous fixed-cost platforms to ship merchandise that behave like software program. Marginal price falls towards zero and replication danger rises sharply.
To make such funding believable, a agency should seize not solely infrastructure margins but in addition downstream worth. Ellison’s media and TikTok bets will be learn via this lens. He’s attempting to internalize demand, personal information, and lock in customers to his stack.
Situations: What occurs subsequent
The AI economic system continues to be younger, and its trajectory is way from settled. What appears like explosive progress as we speak may harden into secure utility-like returns, collapse into commodity pricing, or evolve into one thing solely new. The desk under outlines 5 doable paths ahead, every with a special mechanism, instance, and set of return implications. Collectively, they seize the strategic uncertainty that buyers and executives should navigate as they place bets on this quickly shifting panorama.
| Situation | Mechanism | Instance or Hypothetical | Return Implication |
| 1. Oligopoly and utility returns | Only some gamers survive, scale protects them | A handful of companies management AI compute globally, much like cloud suppliers | Returns stabilize within the vary of 8 to 12 p.c actual ROI. Worthwhile however not spectacular. |
| 2. Full commoditization | Core fashions grow to be open-source and standardized. Margins collapse. | Mannequin suppliers compete on price, worth shifts to functions | Infrastructure returns decline sharply. Most buyers lose capital. |
| 3. Regulatory lock-in | Mental property, licensing, and security guidelines elevate obstacles to entry | Governments mandate audits or limit mannequin distribution | Choose companies earn outsized returns below safety, however danger stays excessive. |
| 4. Vertical integration wins | Corporations mix provide, information, and demand in a single stack | Ellison’s Oracle to Paramount to TikTok technique succeeds | Returns come much less from mannequin edges and extra from ecosystem management. |
| 5. Paradigm shift reset | New architectures (quantum, neuromorphic, organic) displace as we speak’s fashions | Markets reset each decade, wiping out prior leaders | Excessive churn, destruction of capital, and occasional speculative winners. |
Illustrative instance (Situation 4): Suppose Oracle’s media and TikTok strikes let it amortize $50 billion of infrastructure price by charging “inner compute” utilization to its media division whereas additionally promoting surplus capability externally. Even when exterior margins decline, the interior load is locked and margin shift is feasible.
Beneath hybrid eventualities, probably the most resilient companies will likely be those who not solely construct sooner fashions but in addition personal information, lock in customers, and seize consumption straight.
Actionable playbook for executives and buyers
For AI and infrastructure aspirants
- Design your individual demand. Construct or purchase companies whose workloads your infrastructure will serve. Ellison didn’t simply construct information facilities. He purchased content material and distribution to feed them.
- Embed downstream worth seize. Don’t solely hire compute. Take a lower of enterprise outcomes corresponding to margin share in functions your compute permits.
- Time issues, and so does protection. You could have solely 12 to 24 months of premium margin earlier than competitors catches up. Plan your obstacles early, particularly in information, person expertise, and workflows.
- Hedge innovation danger. Allocate reserve capital for radical shifts corresponding to quantum that might out of date your present structure.
For conventional (legacy economic system) companies
- Don’t dismiss AI as irrelevant. Your valuation might already be discounting structural obsolescence. Face it consciously slightly than ready for capital to “normalize.”
- Companion or embed as an alternative of competing straight. Use AI platforms as leverage. Don’t try to win infrastructure races. Purpose to grow to be the most effective AI-powered model of your corporation.
- Discover area moats. In your business, determine information or workflow exclusivities that generalized AI can not simply replicate (for instance healthcare outcomes or provide chain idiosyncrasies).
- Speed up pilot to scale cycles. The window of differentiation is shrinking. Pilots that final 18 months will already be out of date. Use modular and speedy iterations.
Why Ellison’s strikes matter
Ellison’s Oracle plus Paramount plus TikTok technique is a microcosm of the logic that can outline winners within the AI age. These mixtures internalize demand, produce information, lock in workloads, and blur the road between provide and demand.
If profitable, such companies will likely be much less susceptible to margin collapse from commoditization as a result of they don’t seem to be simply compute distributors. They’re full-stack ecosystems.
If markets reward this technique, Ellison’s uncommon media bets might appear like foresight. If not, they could seem as extravagant diversions. Both approach, watching whether or not he can maintain that vertical system supplies a reside case research in how radical capital is being reallocated in actual time.