Hey there, bargain hunter.

This week, two of the most powerful technology companies on earth told you exactly what they believe the next decade looks like — and they said it in the only language that counts in investing: committed capital.

Amazon disclosed plans to spend roughly $200 billion on capex in 2026, the bulk of it aimed at AI infrastructure for AWS. Meta signed a $21 billion AI cloud deal with CoreWeave running through December 2032. These are not press-release ambitions. They are binding financial commitments backed by real cash flows and publicly filed documents.

When two companies spending a combined quarter-trillion dollars in a single year agree on where the money goes, pay attention.

The Scoreboard

• AWS revenue run rate: $142B, up 24% YoY — fastest pace in 13 quarters
• Amazon chip portfolio (Graviton, Trainium, Nitro): $20B+ run rate, triple-digit YoY growth
• Amazon free cash flow: Fell from $38B to $11B in 2025 — intentionally, on $50.7B in infrastructure spend
• CoreWeave 2025 revenue: $5.13B, up 168% YoY; 2026 guidance: $12–13B
• CoreWeave revenue backlog: $66.8B at year-end 2025 — more than 4x the start of the year
• Meta 2026 capex guidance: $115–135B, nearly double 2025 — AI infrastructure is the primary driver

What the Market Is Really Saying

Markets move on the delta between expectations and reality. For most of 2025, the AI infrastructure trade carried one nagging question: is this genuine monetizable demand, or is it a Field of Dreams capital cycle?

Jassy’s shareholder letter answered it. AWS demand continues to exceed available capacity. That is not a demand problem — it is a supply problem. In a high-margin, recurring-revenue business, a supply problem is exactly the kind of problem you want a company to have. Every gigawatt of compute that comes online is generating revenue on day one. Each incremental gigawatt translates to roughly $3 billion in annual cloud revenue.

The Meta/CoreWeave deal drives the same point home from a different angle. Meta is building its own data centers. It has the capital to build more. And yet it locked in $21 billion in external capacity through 2032. Why? Because inference at Meta’s scale — hundreds of millions of daily active users across Facebook, Instagram, and WhatsApp — requires distributed compute that its own facilities cannot always absorb at peak load. That single data point is the entire investment thesis for the specialist AI cloud market.

Is It Cheap?

Amazon is not cheap on trailing P/E. The right frame is earnings power trajectory. Operating income grew 17% through the investment cycle. AWS sits at a $142B run rate, yet 85% of global IT spend still runs on-premises. That is the runway. The chip business — a $20B run rate that simultaneously lowers AWS’s own infrastructure costs by tens of billions per year — is an asset the market has not fully priced.

CoreWeave is expensive on every traditional metric. But the valuation debate should center on backlog coverage, not trailing earnings. A $66.8B backlog against $5.13B in 2025 revenue represents more than 13 years of current-run-rate revenue under contract. The risks are real: customer concentration, aggressive leverage (including $4.25B in new debt at roughly 10% — effectively junk-bond pricing), and dependence on Nvidia GPU supply. The bull case is that a backlog anchored by Meta, Microsoft, Google, and OpenAI does not evaporate when the AI cycle wobbles.

Bull / Base / Bear

Action Plan

Amazon — core position, accumulate on weakness. The balance sheet is fortress-grade. FCF compression through 2026 is the price of entry for a 3–5 year hold. Treat pullbacks of 5–10% as scale-in opportunities. The payoff thesis is FCF re-expansion in 2027–28 as capacity monetizes and chip margin scales.

CoreWeave — satellite position, maximum 3–5% of portfolio. Use the $66.8B backlog as your fundamental anchor. If the backlog shrinks materially in upcoming quarterly reports, that is your exit signal — not the stock price alone. Size for high conviction, not high comfort.

The Cheap Investor Scorecard

• AWS growth rate — Re-acceleration above 25% is the bull signal. Below 18% warrants a reassessment.
• Amazon FCF trajectory — Target $25–35B by 2027. Any further deterioration without revenue acceleration is a red flag.
• Amazon chip run rate — Watch for $30B+ guidance. This is the hidden margin lever the market undervalues.
• CoreWeave revenue backlog — Must hold above $60B. Contraction is the single most important exit signal.
• CoreWeave contracted power conversion — 3.1 GW contracted vs. 850 MW active. Track quarterly conversion as the near-term revenue proxy.
• Hyperscaler capex guidance — If Amazon, Meta, Microsoft, or Google cut 2026–27 capex meaningfully, reprice the entire trade.

The Bottom Line

The AI infrastructure buildout has moved from narrative to arithmetic. Amazon is committing $200 billion. Meta is locking up $35 billion in external cloud capacity through 2032. These are not signals of speculative enthusiasm. They are signals of operational necessity.

If AI workload demand continues to outpace hyperscaler capacity — and every current data point supports that premise — then AWS is the most defensible large-cap beneficiary, and CoreWeave is the highest-torque, highest-risk expression of the same structural theme.

The $200 billion is Jassy’s bet. The question for you, bargain hunter, is whether you want to sit across from it — or alongside it.

The data says alongside.

— The Cheap Investor Editorial Desk