David Tepper has spent years betting on the biggest beneficiaries of artificial intelligence.
Now he’s moving further down the stack.
Appaloosa Management, led by Tepper, added $107 million of CoreWeave (CRWV) stock and $38 million of SpaceX stock to its technology-heavy portfolio in the second quarter. Alphabet, Amazon, Micron Technology, and Taiwan Semiconductor Manufacturing were Appaloosa’s biggest positions.
The two new holdings are completely unique.
CoreWeave leases powerful AI compute power. SpaceX launches rockets, runs Starlink and is building a nascent AI business.
But they share one key similarity.
Both require huge amounts of physical infrastructure to keep up with demand that is outstripping current capacity.
So Tepper’s buying isn’t so much a bet on two hot AI stocks as it is a bet on one of the industry’s biggest constraints: building enough infrastructure to sustain the boom.
David Tepper is moving deeper into AI infrastructure
Tepper already has significant traditional AI exposure through Appaloosa’s holdings.
Amazon made up around 15% of the reported portfolio at the end of June, while Micron made up nearly 15%, Taiwan Semiconductor made up roughly 10%, and Alphabet made up roughly 8%. Nvidia was also one of the fund’s biggest holdings.
CoreWeave takes that approach down to the infrastructure layer.
The company generated $2.58 billion of second-quarter revenue, while revenue backlog reached about $104.2 billion as of June 30.
That backlog was up from $99.4 billion in the previous quarter and excluded more than $25 billion of additional commitments signed early in the third quarter.
Related: Bank of America tweaks CoreWeave stock forecast after earnings
That’s an insane amount of contractual demand for a firm that is still at CoreWeave’s current size.
Its customers include AI developers and big organizations. CoreWeave claimed it grew ties with firms such as Databricks, Cognition, and Runway and gained business from Caterpillar and Grammarly.
Demand isn’t the obvious issue.
It is worth meeting it.
CoreWeave’s $104 billion backlog comes with a massive bill
The $104 billion CoreWeave figure may not be the most illuminating.
That would be $35 billion to $39 billion.
That’s how much CoreWeave now intends to spend on capital expenditures in 2026, compared with its earlier prediction of $31 billion to $35 billion. Alone, capital investment in the second quarter was more than $9.4 billion, compared to $6.8 billion in the preceding quarter, Reuters reported.
The expenditure reveals the odd economics of Tepper’s gamble.
Giant deals are possible, since there’s still a shortage of high-end AI compute capacity at CoreWeave. But it needs data centers, networking equipment, electricity, and pricey accelerators to be set up first to turn those contracts into cash.
This scenario creates a potentially profitable loop if demand for AI keeps increasing.
Such an investment also constitutes a major financial risk.
“Neocloud” suppliers of AI, such as CoreWeave, combine strong leverage with a substantial dependence on a relatively limited number of major clients and on Nvidia technology, Reuters noted. So a change in demand, price, or chip technology in AI might damage corporations that have plunked down billions on infrastructure years in advance.
That comparison makes sense, particularly given that during the second quarter, more than half of CoreWeave’s backlog was already in customer delivery.
Tepper isn’t betting just that consumers want AI compute. The backlog already answers that.
He’s betting CoreWeave may provide substantial profits after paying the big expenditure necessary to supply it.
David Tepper just made two AI bets with one thing in commonEston Parker/ISI Photos / Getty Images
SpaceX gives Tepper another version of the same bet
SpaceX solves the infrastructure issue from space.
The startup promotes itself as building infrastructure for space, connectivity, and AI. Its activities include launch services, Starlink, and a burgeoning artificial intelligence enterprise.
SpaceX’s first quarterly report since going public underlined exactly how different those companies already are.
Starlink earned $4.3 billion in sales in the second quarter, up 66%, and $1.7 billion in operating profit. SpaceX’s AI business brought in around $2.6 billion in revenue, quadruple the previous-year figure, although it remained unprofitable.
More Manager Buy/Sells:
Michael Burry increases his bet against popular chip giant
Warren Buffett reveals he broke his own investing pattern
Mark Cuban bets on MLB with Athletics minority stake
The enterprises are linked by the expense of bringing them to market.
Because SpaceX launches its own rockets, it can launch Starlink satellites at a scale rivals cannot. Starship might extend that lead much farther if SpaceX can make the vehicle fast, reusable, and drastically drive down the cost of delivering mass into orbit.
That might enable larger Starlink installations and perhaps far more extensive orbital-computing infrastructure.
So SpaceX is just another capital-demanding gamble on limited capacity.
The distinction is that its infrastructure is situated above the Earth, not within a data center.
Tepper’s two purchases reveal where the AI trade may be moving
Tepper’s positions in CoreWeave and SpaceX are small compared with Appaloosa’s biggest holdings.
That’s important.
The purchases look more like early bets than portfolio-defining convictions, but collectively, they reflect a change in where some knowledgeable investors think the next AI opportunity lies.
The initial phase of the boom benefited semiconductor producers, cloud giants, and model creators.
The next phase might be more about organizations that can offer the actual infrastructure those businesses use.
CoreWeave has a $104.2 billion backlog, showing that users are already booking large quantities of future work.
SpaceX’s Starlink and AI and launch companies are another way to tackle capacity and connection issues.
The potential is huge precisely because developing that infrastructure is tough.
So is the risk.
CoreWeave has to invest tens of billions, with no idea what the AI hardware economics will look like many years from now. SpaceX must keep pouring billions into Starship, satellites, and computer infrastructure while showing that those companies can support a value now in the trillions.
Tepper is not merely betting that AI keeps growing.
He is betting that the companies supplying the infrastructure that AI cannot grow without will capture an increasing share of the value.
Related: CoreWeave stock sinks as mag 7 move rattles investors