Three Cash-Burning Stocks and the Same Hidden Pattern

On a single Monday in early August, Cathie Wood bought three stocks that share a problem almost no other investor would want: they're spending themselves into the red.

CoreWeave burned $20.6 billion on capital expenditures over the trailing twelve months. Amazon spent $169 billion. Rocket Lab is about to add $8 billion more in debt to fund its transformation. Combined, these three companies are hemorrhaging free cash flow — but the market is punishing them for building infrastructure that could dominate for years.

The "bargain hunting" headline misses what's actually going on. This isn't a dip-buying pattern. It's a bet that the market has misread a massive capital deployment cycle as a cash flow crisis.

Let's start with the most extreme case: CoreWeave.

The AI infrastructure company went public at $40 early last year. Revenue has roughly doubled each quarter since then — $2.6 billion in Q2 2026, up 112% from a year earlier. It generates $1.5 billion in adjusted EBITDA on that revenue, a 59% margin that shows the underlying business is highly profitable.

But the GAAP picture looks nothing like 59%. CoreWeave reported a $626 million net loss in Q2. Its free cash flow margin sits at -170%. The company carries $72 billion in debt.

The disconnect is entirely capital deployment. CoreWeave is spending $9.4 billion per quarter on GPUs, racks, and data center capacity to feed a $104 billion revenue backlog. Management guides to between $35 billion and $39 billion in annual capex. That's more than three times its full-year revenue of $12.4 billion to $13.2 billion.

Here's the mechanism: CoreWeave raises debt, buys chips, builds capacity, and books it as capex. Revenue comes in as customers consume that capacity over contracted periods. The losses are structural while the buildout continues but temporary in the sense that they reverse once capacity deployment slows relative to revenue growth. CEO Michael Intrator called Q2 an "inflection point" as scale began translating into operating leverage. Adjusted operating income rose to $128 million at a 5% margin.

The question for any investor: is the buildout finite, and will demand fill the capacity? The $104 billion backlog — with a Meta commitment of $21 billion and multi-year agreements with companies like Anthropic — suggests demand is not the issue. The risk is execution and financing. Interest expense jumped to $640 million in Q2 from $267 million a year earlier. CoreWeave survives if the math eventually works: $39 billion in infrastructure spending that generates tens of billions in contracted revenue over multi-year terms.

Amazon's story looks different but follows the same pattern — massive capex, negative free cash flow, and a market that hasn't fully priced in the return profile.

Amazon reported $200.6 billion in Q2 revenue — the first quarter ever to cross that threshold. AWS grew 37%, its fastest pace in 18 quarters, to $42.2 billion. AWS operating income rose 64% to $16.6 billion at a 39% margin. The company's consolidated operating income of $27.5 billion was up 43%.

Then came the capex number: $169 billion trailing twelve months, up 64% year over year. Amazon raised its full-year 2026 capex guidance from $200 billion to $220 billion. Free cash flow swung to a $7.6 billion outflow from $18.2 billion a year earlier.

The market reaction tells you everything about the narrative gap. When Amazon first announced the $200 billion capex plan in February, the stock fell 11%. When it raised that guidance by another $20 billion after Q2 earnings, the stock rose 15%. The capex didn't change. What changed was disclosure quality.

For the first time, CEO Andy Jassy gave the payback period for the AI servers Amazon was buying: less than three years. The asset life is five to six years. That means the servers pay for themselves in roughly half their useful life, then generate pure cash flow for the remaining period. The market had been pricing a cash flow crisis instead of a favorable investment return.

AWS supports this with a $496 billion backlog at a 6.4-year weighted-average remaining term. The AI and chips businesses within AWS each exceed $25 billion in annualized run rate with triple-digit growth. Amazon has $78 billion in cash and only $6 billion in net debt. Operating cash flow hit a record $161 billion on a trailing basis.

Amazon trades at 21x trailing earnings but nearly 40x forward estimates — a reflection that the earnings denominator is being distorted by the investment cycle. The PEG ratio of 0.23 suggests the forward multiple, while elevated in absolute terms, is still well below the revenue growth rate. This isn't a company running out of cash; it's a company that generates $161 billion in operating cash flow and is choosing to spend most of it on assets that pay back in under three years.

Rocket Lab is the hardest thesis of the three.

The company generates $234 million in quarterly revenue, up 62% year over year, with a backlog that surged 137% to $2.36 billion. But it's also running a 33% negative operating margin, burning cash, and trading at roughly 49x trailing revenue — a valuation that demands extraordinary future execution.

Then Rocket Lab announced it would acquire Iridium Communications for $8 billion.

Iridium is a profitable, cash-generative satellite network with $872 million in annual revenue and a 57% EBITDA margin. It serves 2.55 million active subscribers across defense, aviation, maritime, and IoT markets, backed by globally harmonized spectrum that competitors can't easily replicate. The deal closes mid-2027.

This acquisition changes the Rocket Lab thesis from a pure launch infrastructure play to a vertically integrated space company — designing, building, launching, and operating satellite constellations. It's the same playbook SpaceX executed with Starlink: use your launch capability to build and replenish your own network, eliminating third-party launch costs and capturing recurring subscription revenue internally.

But $8 billion is enormous for a company with $2.1 billion in cash and a $38 billion market cap. The deal is structured as $27 per share in cash plus stock, with a $3.6 billion bridge loan facility in place. The stock portion carries a collar banding from $67.50 to $112.50 per Rocket Lab share, which means if Rocket Lab's stock declines further, it has to pay more in equity — a dilution risk that the collar is designed to limit but not eliminate.

The Iridium deal also adds integration risk. Rocket Lab has already been absorbing smaller acquisitions while developing its Neutron rocket, which is on track for first flight in Q4 2026. Adding an $872 million revenue business with a complex satellite constellation and 2.55 million subscribers is a different order of operational challenge.

The math that justifies it: Iridium's $495 million in EBITDA at a 57% margin would dramatically improve Rocket Lab's consolidated profitability profile once the deal closes. Combined with Neutron's entry into service — which opens high-margin government launch contracts — the integrated company could look like a fundamentally different entity than the unprofitable launch company Rocket Lab is today. The 90-mission launch backlog and a $397 million U.S. Space Force contract support the launch pipeline. The risk is that execution stumbles on any front — Neutron delivery, Iridium integration, or continued dilution — and the high revenue multiple compresses fast.

What ties these three together isn't that they're cheap. Amazon isn't cheap at $2.8 trillion. Rocket Lab isn't cheap at 49x revenue. CoreWeave isn't cheap at nearly 10x enterprise value to sales.

What ties them together is that they're all in the middle of building something expensive and long-lived, and the market is reading their cash flow statements at face value instead of through the lens of asset quality and contracted demand.

CoreWeave's $104 billion backlog against $39 billion in annual capex. Amazon's sub-three-year payback on servers that last five to six years. Rocket Lab's path to adding $495 million in profitable EBITDA through Iridium while building its own launch engine.

The pattern is capital-intensive companies where the market sees cash outflow and the investor sees contracted return. That's not bargain hunting. It's the same investment as it has always been when infrastructure is being built at scale: buy the company deploying the capital when the numbers look ugly, and hope the capacity becomes real revenue.

The difference today is the size. Three companies — one a $2.8 trillion mega-cap and two high-growth names — collectively representing more than $300 billion in combined market value are executing this playbook simultaneously. That concentration of infrastructure capex is worth tracking, regardless of whether you agree with the bet.

The risk is the same across all three: if the capacity doesn't convert to revenue, or if the financing costs compound faster than the revenue growth, the losses become structural instead of temporary. CoreWeave's interest expense trajectory, Rocket Lab's Iridium integration, and Amazon's server utilization are the three variables that determine whether this is a profitable buildout or an expensive overreach.