Elon Musk Didn't Lose Half a Trillion — He Just Changed What the Market Can Price

Forbes pegged Elon Musk's net worth at roughly $715.6 billion on July 27, 2026. On June 16 — barely seven weeks earlier — the same tracker put him at $1.45 trillion. The swing was almost entirely SpaceX stock: up from an IPO close of $161.11 on June 12 to a 52-week high of $225.64, then down to $114.92 at the close on August 6.

The market has been obsessed with the number. It shouldn't be. The number is just a lagging mirror of what SpaceX has become — and the real question is whether the market is misreading the architecture shift behind the volatility.

The company the market doesn't know how to value

SpaceX is no longer a rocket company. Not in the way that matters for valuation.

In February 2026, SpaceX acquired xAI, Musk's AI startup, in a merger that valued the combined entity at $1.25 trillion. By the time SpaceX filed for its IPO in April — targeting a $1.75 to $2 trillion valuation — the company was pitching itself as the next generation of AI infrastructure. The IPO raised $75 billion, the largest in history, and Musk's 38% ownership stake made him the world's first trillionaire.

But the market doesn't know what to do with a company that is simultaneously a launch provider, a compute landlord, and a chip manufacturer. The $115 price level you see today — down roughly 49% from that June high — isn't just post-IPO normalization. It's the result of a market trying to apply traditional valuation frameworks to a business model it hasn't seen before.

The compute deals tell the story

Here's what's actually happening beneath the stock price:

Anthropic signed a deal to pay $1.25 billion per month through 2029 for access to all available compute at SpaceX's Colossus 1 data center — that's more than 220,000 Nvidia GPUs and 300 megawatts of dedicated power. Google will pay $920 million per month from October 2026 through June 2029 for approximately 110,000 Nvidia GPUs. A third customer, open-source model provider Reflection, agreed to $150 million per month starting in July for access to Nvidia GB300 chips inside Colossus 2.

Put plainly: SpaceX has lined up roughly $2.32 billion per month in committed compute lease revenue. Annualized, that's $27.8 billion per year. This is before any additional deals, before Terafab comes online, and before Musk's vision of orbital data centers moves from concept to execution. The company reported a net loss of $4.28 billion in Q1 2026. The compute business is the bridge from capital-intensive rocket operator to high-margin infrastructure provider.

This is the same transition pattern I've tracked across the AI stack: once a company builds sufficient infrastructure, the recurring revenue multiple jumps. Hardware sets the ceiling; recurring compute contracts set the multiple. SpaceX is doing for AI compute what hyperscalers did for cloud hosting — building the physical layer, then monetizing capacity at scale.

Terafab is the supply chain signal that matters

If the compute deals tell you who needs the capacity, Terafab tells you why the global supply chain can't fill the gap.

On August 6 — yesterday — Tesla and SpaceX announced they will initially invest $16.8 billion to build Terafab, a semiconductor fabrication plant in Grimes County, Texas. Filings reveal the total projected spend could reach $119 billion across a multi-phase buildout. The facility will cover chip design, manufacturing, packaging, and testing under one roof, over 100 million square feet.

Intel is joining the project, contributing its forthcoming 14A process node. Tesla broke ground in April on a separate research fab at its Giga Texas campus as a precursor.

Musk has been explicit about why this is happening: external suppliers can't produce enough hardware to satisfy demand, and geopolitical risk around Taiwan Semiconductor Manufacturing Company creates unacceptable supply-chain vulnerability. Chip analyst Ben Bajarin called it a "15-year strategy" to control the supply chain.

The supply signal here is unambiguous. A $119 billion capex commitment, split between a company that just went public and one trading at $333 per share with a $1.32 trillion market cap, is not a speculative bet. It's a capacity commitment that says demand for compute will outstrip global supply for the next decade. That's the same bottleneck argument that drove hyperscaler capex to record levels — except now one player is trying to own the entire stack from silicon to satellite.

SpaceX is also writing its own AI training software. Musk claims a custom training stack in C, optimized for a cluster of 220,000 GB300 GPUs connected via 800G NICs, with speed improvements over JAX described as "over an order of magnitude." Whether or not you take that at face value, the architecture play is clear: control the silicon, control the software, control the compute layer.

The leverage risk

However, the same infrastructure thesis that makes SpaceX compelling also concentrates enormous execution risk in one person's capital allocation.

SpaceX stock has been hammered by short sellers since the IPO. Data from S3 Partners showed approximately 95% of available shares on loan — roughly 34% short interest — in the days before the insider lockup expired on August 6. Musk has publicly warned short sellers on Twitter, posting messages on July 17 and August 4 cautioning against sustained bearish positions. When management feels the need to tweet at bears, the stock is not yet doing the convincing.

More importantly, the Terafab commitment transforms Musk's capital exposure. Between Tesla's $1.32 trillion market cap, SpaceX's public valuation (roughly $1.5 trillion at current prices, though market cap math is fluid given the share structure), and a joint capex pipeline that could approach $119 billion, the leverage concentration is extraordinary. That doesn't mean the thesis is wrong. It means the margin for error is thin.

The lockup expiration on August 6 is another pressure point. Hundreds of millions of previously restricted shares became eligible to trade, adding supply to a stock already trading near its 52-week low of $104.83. The price is at $114.92 — a recovery of about 10% intraday, but still down sharply from June levels.

The Tesla context

Tesla adds a different layer of complexity to the Musk allocation question. Musk's position in Tesla has been growing, not shrinking. On June 16 — the same day SpaceX went public — he exercised 303.9 million stock options from his 2018 compensation package at $23.34 per share. Tesla withheld 17.5 million shares to cover the exercise cost. The net result: Musk's Tesla stake grew to approximately 20% of the company, his highest ownership level ever.

The catch: those shares are restricted until 2033. The $115.9 billion in paper gains from that transaction can't be liquidated. And Tesla's own business is facing headwinds — the stock is down 25.8% year-to-date, and Q2 2026 earnings, reported July 22, missed badly: $0.33 EPS versus a $1.03 forecast, on revenue of $28.24 billion.

Tesla's fundamentals tell a story of a company that's growing but not accelerating. Revenue is up 11.75% year-over-year with a 26.13% quarter-over-quarter jump — the sequential momentum is there, but the operating margin sits at 5.0% and ROIC at 3.3%. The stock trades at 346 times trailing earnings and 417 times forward earnings. That's not a value play. It's a pure vision multiple.

Where does the capital go?

The debate is not whether SpaceX is executing an ambitious infrastructure play. The compute deals, the Terafab announcement, and the in-house software development make that clear. The debate is whether the current SpaceX share price — down roughly 49% from its June high, near 52-week lows, with the lockup just expiring — represents a setup worth positioning into, or whether the leverage risk and execution uncertainty warrant patience.

I believe SpaceX is on the right side of the compute bottleneck transition. The company is positioning itself between the chip layer and the model layer — the exact bottleneck that's been constraining AI progress. Anthropic, Google, and Reflection aren't placeholder customers. They're customers that are spending billions because the global compute shortage is real and immediate.

But the stock at $115 carries two distinct risks. First, the market is still deciding what category SpaceX belongs in — and until that narrative crystallizes, the multiple will swing violently. Second, the Terafab capex commitment means Musk is deploying capital at a scale that, if execution stumbles, creates a double whammy of stock decline and real-world stranded investment.

In my opinion, the SpaceX setup is better than it was at $225. The 49% decline from peak has compressed the "Elon premium" and brought the stock closer to where a patient investor could build a position. But the lockup expiration and 34% short interest mean the volatility is not over. A small position sized for turbulence — not a core allocation — seems appropriate for someone who believes the compute infrastructure thesis.

For Tesla, the picture is different. Musk has locked himself in with 20% ownership and restricted shares he can't sell until 2033. The missed Q2 earnings, single-digit operating margins, and 346x trailing P/E mean the Tesla story is entirely dependent on autonomous driving and robotics execution that hasn't materialized in the financials yet. I believe the long-term thesis — Optimus robots, Cybercab, energy storage — could still compound. But the near-term return profile on Tesla looks back-half weighted, and much of that upside is tied to the same Musk vision premium that's already reflected in the price.

Wedbush analyst Dan Ives has put 80% to 90% odds on a Tesla-SpaceX merger by early 2027. If that happens, the valuation question becomes even more complicated — but also more interesting. A combined entity that owns rockets, AI compute, edge chips, autonomous vehicles, and robotics would be the most vertically integrated technology infrastructure company in the world. Musk's capital allocation has been converging toward that outcome for two years.

The break point for my bullish view: if SpaceX fails to execute on Terafab, if the compute deals lose customers due to performance shortfalls, or if the stock falls below $90 on sustained fundamental weakness rather than short-seller pressure. Any of those would signal the infrastructure thesis is overstretched.

The question isn't whether Musk is building something transformative. The question is whether the current prices — SpaceX at 49% off peak, Tesla at 25% off year-to-date highs — offer enough margin of safety for the concentration risk you're taking. I think SpaceX at these levels is worth a position. Tesla is worth waiting for better evidence that the auto business is reaccelerating before adding more capital.