Tesla's Spending Pincer Has No Exit

On Tesla's Q2 earnings call Tuesday, Elon Musk said two things that cannot both be true in the way shareholders need them to be. He made his confidence in the AI investments explicit.

This is a massive capex year, but I'm confident that all the things that we are investing in will yield incredible returns. — Elon Musk

Then, in the same call, he described the production ramp for Optimus — the humanoid robot he has called "potentially one of the most important products in human history" and projected could eventually generate $10 trillion in revenue ^[1][2].

the production ramp would be "agonizingly slow." — Elon Musk

Confidence in the destination, admission the journey is slowing. The two statements sit in tension, and the tension is the company's central problem. The arithmetic is straightforward. Tesla told investors it would spend $25 billion on capital expenditures in 2026, almost all of it directed at AI infrastructure, robots, and autonomous driving ^[1]. By mid-July, it had spent only $2.5 billion ^[3]. Even that partial outlay produced negative $1.1 billion in free cash flow — Tesla's first quarterly cash burn in more than two years ^[4]. The CFO had warned in April this was coming, calling it "the right strategy to position the company for the next era" ^[1]. But the market's reaction has been less patient: Tesla shares are down 18% this year, the worst performance among the Magnificent Seven, and the stock trades at 163 times forward earnings ^[3]. Now here is the pincer. Investors are not telling Tesla to slow down. They are telling it to spend more. The $2.5 billion deployed so far is seen as too little, too slow, in an AI arms race where Alphabet alone plans $180 billion in capital expenditures this year — roughly seven times Tesla's entire budget ^[5]. Morgan Stanley warned in April that "with capex doubling and FCF turning negative, investors will need clearer evidence that unsupervised autonomy is around the corner to support the stock's valuation" ^[6]. Three months later, the evidence has not arrived. Tesla's Austin robotaxi service launched in January with what the company described as "a few unsupervised vehicles" mixed into a fleet still shadowed by safety monitors in chase cars ^[7]. Musk had claimed in January that "self-driving cars is essentially a solved problem at this point" ^[7]. By July, the revenue timeline had not shortened. So Tesla faces a double bind that no amount of AI-pivot rhetoric resolves. Spend too little, and the 163-times-forward-earnings multiple — a valuation that prices in an AI future, not a car company — collapses under its own assumptions. Spend enough to actually compete with the Alphabets of the world, and the core electric-vehicle business cannot fund it. The EV business is not collapsing: Tesla's U.S. market share held at 54 to 58 percent in the first quarter even as the broader EV market contracted, and the company is explicitly prioritizing margins over volume ^[8]. The AI5 chip reached tapeout 45 days ahead of schedule, a genuine technical achievement that prompted UBS to upgrade the stock from sell to hold ^[9]. But a well-run car company with a promising chip program is not what the 163-times multiple is pricing. That multiple is pricing a revolution in artificial intelligence and robotics, and the revolution is not arriving on schedule. Enter the SpaceX merger. In the days since the earnings call, Gene Munster of Deepwater Asset Management raised his estimate of a Tesla-SpaceX combination to 90 percent probability ^[10]. Gwynne Shotwell, SpaceX's president, said a merger "might make Elon's life a little easier" ^[10]. Musk himself acknowledged the convergence.

As you can tell from the many collaborations on so many fronts with SpaceX, there's more and more overlap. — Elon Musk

But when asked directly about a merger on the earnings call, he deflected.

We can’t talk about, you know, combining companies and that kind of thing on an earnings call—it has got to be done with the appropriate process. — Elon Musk

The merger is being positioned as a solution to the resource-allocation conflict at the center of Tesla's pincer. And the entanglement between the two companies is real enough to make the logic seem plausible. Tesla converted a $2 billion investment in xAI into SpaceX equity after SpaceX acquired xAI in February, giving Tesla a stake of just under 1 percent in the rocket company ^[11]. Tesla generated $573 million in revenue last year from sales to xAI and SpaceX — $430 million from Megapack batteries to xAI, $143 million from Cybertrucks to SpaceX ^[12]. The two companies are partnering with Intel on Terafab, a $55 billion chip-making facility in Austin designed to produce a terawatt of compute hardware annually ^[13]. The web of cross-company transactions is dense enough that a merger would, in one sense, simply formalize what already exists. But the merger does not resolve the underlying arithmetic. SpaceX has its own AI-infrastructure ambitions, and it recently raised $20 billion to $25 billion in bonds to fund them — a debt load that contributed to a 30 percent plunge in its stock from the peak ^[14]. Two companies whose AI-infrastructure ambitions already exceed what either can independently finance do not become more financeable by combining. They become one larger entity with the same gap between what the market demands and what the operations can produce. Shotwell was precise in her phrasing: a merger "might make Elon's life a little easier." That may be true. But making one person's life easier is not the same as making two stretched balance sheets into one solvent one. The pincer survives the merger because the pincer is not an organizational problem. It is a numbers problem. Tesla must spend billions it does not have to justify a valuation built on a future that has not arrived, and no corporate restructuring changes the fact that the spending required exceeds the cash the business generates. The escape does not compute.