a16z’s *State of Markets II* report centers on two questions on everyone’s mind: Where is the flood of capital pouring into AI actually going? Across the full AI value chain, which players can turn revenue into consistent profits, and which ones are just burning cash as also‑rans?

A 90‑page a16z report titled *The $1 Trillion AI Buildout* has recently gone viral among investors.

Many fixate on fundraising news for AI startups and foundation‑model companies, yet they often overlook one reality: massive sums of capital have been reallocated across different segments of the AI value chain. Where capital flows is not where profits ultimately accrue. Hype around a sector does not guarantee it will produce the eventual winners.

Cutting through dense technical tables, we break down the report’s core arguments and unpack what this AI wave means for international students.

a16z’s *State of Markets II* report revolves around two questions top of mind for everyone: Where is the flood of capital pouring into AI actually going? Across the full AI stack, which players can convert revenue into consistent profits, and which ones are merely burning cash as also‑rans?

Unlike many relentlessly optimistic industry narratives, this report does not simply hype an explosion of AI applications. Instead, it soberly maps out the current AI landscape through the lens of capital expenditure, supply chains and enterprise ROI on AI spending.

I. Capital Is Flowing En Masse Into Foundational Infrastructure

The most obvious hallmark of this AI boom is not an explosion of novel AI applications, but a staggering surge in capital spending on compute infrastructure.

Combined projected capex for Microsoft, Amazon, Google, Meta and Oracle, the five major cloud providers, is set to reach nearly $777 billion in 2026. The vast majority of this capital is not going to model startups or AI application teams. It is directed toward GPUs, optical modules, storage, high‑speed interconnects, power systems, liquid cooling and data center construction.

The VC market is similarly bifurcated: 86% of US venture capital dollars are allocated to AI, while AI deals account for only 43% of total deal volume. Put simply, capital is highly concentrated, and the Matthew effect is pronounced. Countless smaller startups struggle to secure funding.

Many founders and investors still chase opportunities in the model and application layers, yet capital’s near‑term vote is unambiguous: build the foundation first, then layer on applications. As long as hard constraints on compute supply persist, infrastructure will continue to capture the largest share of capital budgets.

II. Near‑Term Profit Winners: Upstream Supply Chain, Not Foundation‑Model Firms

This is one of the report’s most sobering takeaways: in today’s AI value chain, the upstream hardware supply chain has the strongest ability to realize profits.

Chipmakers, optical module suppliers and supporting hardware vendors are capturing premiums driven by compute shortages. They benefit from solid order visibility and reliable collections, making them the clear near‑term winners of this cycle.

By contrast, foundation‑model players in the middle layer boast impressive revenue growth, yet their business model carries inherent heavy costs.

Training, continuous iteration and inference all consume massive ongoing compute resources. As revenue scales, capital outlays expand in tandem. Models are rapidly commoditizing, capability gaps between competitors are narrowing, and inference costs keep trending down.

This means model companies that sell tokens and API access will find it increasingly difficult to sustain high gross margins, and their moats can erode quickly.

AI application teams, the most numerous group, face the hardest path to profitability. Many AI products on the market are essentially wrappers built on top of foundation models. Enterprise clients are willing to pay for pilots and trials, but only a small share of projects deliver clearly quantifiable gains or cost savings. Budgets may be approved, yet the foundation for recurring purchases and price increases remains fragile. Much of the revenue these applications capture comes from experimental client budgets rather than stable long‑term contracts.

To sum up the current profit dynamic: upstream players capture windfall gains, model vendors compete on scale, and application teams fight to validate real value.

III. Long‑Term Value Will Likely Accrue to Agents and Outcome‑Tied Applications

If the upstream supply chain wins in the short run, where will value reside over the long term?

a16z’s analysis does not stop at compute windfalls. Its view is that lasting value will not be captured by hardware or generic foundation models. Instead, it will belong to AI applications and AI Agents embedded within real workflows and tied directly to business outcomes.

Most prevailing pricing models today charge by token or usage volume.

This structure has an inherent flaw: costs and risk sit with the application buyer, who can easily compare quotes and switch model providers at any time. The Agent paradigm shifts the pricing logic from “compute consumption” toward revenue sharing and outcome‑based compensation. AI is no longer just a plugin tool; it deeply integrates into high‑value business workflows and directly drives conversions, cost reduction and efficiency gains. When product value translates directly into measurable returns for customers, it becomes far less vulnerable to price cuts from underlying model providers.

A critical distinction here: wrapper applications that call foundation models to deliver generic functionality have low barriers to entry. Agent applications that penetrate vertical industries, reengineer workflows and accumulate proprietary domain data and business rules gradually build defensible moats.

Foundation models resemble a universal utility like electricity or water. Essential as they are, utility providers rarely capture all the incremental economic value. That surplus will ultimately be contested by players closest to end‑user operations and paying use cases.

IV. What Is the Biggest Risk in This AI Capital Cycle?

The report highlights an underappreciated risk: today’s hundreds‑of‑billions‑dollar compute buildout rests on a core assumption — enterprises will keep converting AI spending into measurable business returns.

If, after running numerous pilots, companies conclude AI fails to deliver promised gains, willingness to pay will contract rapidly. Once demand fails to keep pace with expanding supply, premiums for compute, chips and optical components will evaporate, triggering a broad de‑risking and deflation across the whole value chain.

This is not to say AI lacks value. Rather, capital expansion has outrun validation of client ROI. Infrastructure takes a long time to build, yet enterprises have limited patience for returns.

Closing Thoughts

The report offers a simple evaluation framework for everyone assessing AI projects: do not judge a venture merely by whether it sits within a hot sector. First identify its position along the value chain and the source of its earnings. Is it profiting from temporary supply scarcity? Is it generating model revenue sustained by continuous cash burn? Or is it delivering measurable, recurring business value for clients?

The compute feast continues, yet the feast will not belong to upstream suppliers forever. The short game is about supply, the medium game about cost optimization, and the long game about use cases. Capital flows reveal today’s answer; profit capture reveals the final one.

This conclusion is also relevant for families planning study abroad and job searches.

Avoid blindly crowding into roles tagged with trendy labels such as foundation models or general AI. Hardware and compute engineering offer abundant near‑term openings but are exposed to severe cyclical swings.

Cross‑functional professionals who understand industry operations and can deploy AI to solve tangible business problems are more recession‑resistant. They are also the talent overseas companies are willing to hire consistently over time.

When selecting majors and hunting for internships, anchoring yourself to genuine business value creates a far more stable career path than chasing fleeting trends.