InvestAnswers Weighs Whether SpaceX Compute Could Justify a $100 Trillion Valuation

The host opens by saying this week's video departs from the usual format. "We're not going to do Bitcoin micro strategy because nothing basically has happened," he says, adding that every week feels the same. Bitcoin is promised only at the end of the video.

He says the SpaceX compute topic has occupied him for a long time and thanks viewers who joined a live poll.

The host says he has been working on SpaceX compute since before the IPO. "I've been burning my brains on this one since even before the IPO, and I'm just going to look at the business, the rumors, the mergers, the odds, the valuations, and yet another price prediction model," he says.

He notes he has built about 12 or 13 price prediction models so far and that the compute topic alone is especially interesting to him. He says he has done the numbers and analyzed the competition, and he adds that the content is "not financial advice, just a guy on the internet."

The host says Musk's posting has been unusually intense. "He has never, ever in history sounded this bullish on SpaceX," he says. He adds that Musk now talks about SpaceX being worth over $100 trillion in 10 years, which the host describes as about 50 times the current level.

The host says Musk backed the claim with post after post, and that the video is supported by many real tweets viewers can look up. He also suggests reading the comments under a tweet for additional context.

The host points to a SpaceX compute roadmap for Q4 2026 and says SpaceX is ramping faster than anyone else. In his account, SpaceX is the only company that can build coherent compute, energy and cooling at this scale, and it can build a facility in about 20 weeks.

He says SpaceX can set up 100,000 GPUs in about 90 days, which he compares with Jensen Huang's remark on speed. "Most other competitors takes 2 to four years to do this," he says, calling this a major problem for rivals.

The host says the demand for compute is going "completely parabolic." He argues the bottleneck is not demand but the ability to scale and deliver compute. "What's not going parabolic is the ability to scale to deliver that compute," he says.

He claims no one else has been able to string together more than 30,000 GPUs coherently, while Musk does 220,000 GPUs repeatedly. He presents these as real facts rather than hopium.

The host lists Colossus 1 as having 150,000 H100s, 50,000 H200s and 30,000 GB200s. Colossus 2 is listed with 110,000 GB200s and 440,000 GB300s. He calls the total "monstrous."

He says another 220,000 GB300s were to be fully operational the following week, with another 220,000 expected in November and possibly another 220,000 by late December. "He or she who has the most compute wins," he says, which he presents as the simple story.

The host describes a site in Brownsville, Texas, near Starbase, where he says SpaceX is building quickly. He cites a source he credits as a friend of the channel for details including about 444 to 445 acres, water, land, buildings, and 400 megawatts of wind turbines on site for one mini heart, with more planned.

He estimates the clusters could generate at least $22 billion a year from 220,000 GB300s and $30 billion a year from 220,000 Vera Rubins. He says the Vera Rubins are coming very soon and that he has rumors from good sources that the company may build these in 20 weeks. He calculates six buildings could produce $132 to $180 billion in annual revenue, which he says is roughly twice Tesla's revenue after nearly two decades of operation.

The host turns to a Ron Baron forecast that SpaceX could be worth $50 trillion in 10 years, which he describes as 25 times from current levels. He says Baron has large holdings in both SpaceX and Tesla and is a long-time believer in Musk. The host adds that Starlink alone might be worth $14 to $15 trillion.

The statement the host finds most significant comes from Musk: "Unless there's a major world war, SpaceX probably gets to more than hundred trillion in less than 10 years." The host calls the numbers bonkers and says he will examine the valuation model and price predictions later.

The host says the S-1 filing lists a total addressable market of $28.5 trillion, and that he read the S-1 twice. He says he will verify the figure independently. The host reports that Musk responded, "We are wrong. It's vastly higher than that."

The host explains that he builds many price models from total addressable market, then derives market share, revenue and profitability. If the TAM is wrong, he says, every model he has built is also wrong and understated.

He also cites a Morgan Stanley view that the market is ignoring the real engine, including inference from space and terrestrial data centers. He says an analyst asked a room whether anyone owned SpaceX and not one hand went up.

The host says Terafab is a project that he claims nobody believes will be pulled off. He reports that last week TSMC, Nvidia and Intel were said to be working with the project. He says chips and power will still be bottlenecks.

He says the output value of the factory could place it among the top 20 national GDPs, and that the first phase involves $17 billion of investment and 3,000 jobs. He describes plans for full vertical integration, including the company's own memory production.

The host cites John Erisman, whom he describes as good at predicting profit and revenue growth, as saying Starlink revenue could grow 925% in five years. The host says that is roughly a tenfold increase. He reports that Musk responded that it is "easy peasy."

The host argues that Europe, China and Russia want their own satellite networks but are far behind, citing Starship payloads of 100 to 150 tons. He says the world is so far behind that it is "actually unfair."

The host says the FAA is reviewing a Starbase expansion that would approve up to 456 launches per year. Because booster and ship land separately, he says this corresponds to 912 landings a year, which he calls more than one a day. He notes possible restrictions on weekends and evenings due to weather.

He stresses the expansion is not approved yet. He also refers to a recent market dip and a CoreWeave controversy, reporting Musk's response to a Zero Hedge article about CoreWeave's risks. The host says the difficulty lies in building the hardware, power, cooling and connectivity, not just raising money.

The host says North America plans 33 to 35 gigawatts of compute next year, but may not be able to turn all of it on because power and transformers may be missing. He says the shortfall could be as high as 13 to 15 gigawatts. He also mentions China's planned 40 to 45 gigawatts with a possible shortfall of about 2.

He argues that chips alone are not enough without facilities and power, and that if demand keeps rising and no one else can deliver compute, compute prices could rise sharply, letting SpaceX earn more per gigawatt.

The host says SpaceX could earn maybe $50 billion or $60 billion per gigawatt once built and running. "As compute demand goes through the roof, and nobody can deliver in time, the demand will explode, and there's only one company in the catbird seat to take advantage of this, and that is SpaceX," he says.

He says this explains why SpaceX has been building rapidly with large engineering teams, and that the upcoming earnings call will show what has been achieved. He then turns to a merger update and separates facts from rumors.

The host cites a Barclays model that treats SpaceX acquiring Tesla in an all-stock deal as the likelier structure, while stressing that this was speculation and not a deal. He notes Tesla's annual meeting on December 3 has no merger on the ballot.

He says Tesla holds a multi-billion-dollar stake in XAI, which became SpaceX shares, and that Musk has hinted at the combination in posts. He reports that prediction-market odds on Polymarket and Kalshi spiked over the last few days to nearly 60% for a merger around spring 2027, which he calls about five or six months away.

The host relays a rumor from a source that SpaceX has locked up 30 to 40% of Nvidia's Vera Rubin allocation, with some going to space next year. He says a radiation-capable version is being built for orbit and mentions a rumored $40 billion Nvidia financing in talks with Apollo.

He cites ARK's estimate that Tesla and SpaceX could reach 66 to 67 gigawatts by 2031, which he says feeds into his valuation model. He reports that Musk said he would be very disappointed if that were all they reached. The host concludes that the companies' capabilities support being very bullish, while noting he is only discussing compute.

The host's base case is 66.5 gigawatts of online compute by 2031, which he says Musk expects to be higher. He says this compute business alone would support about $825 per share, which he calls a 5.65 times increase. In a bull case that doubles gigawatts online, he estimates nearly $4,000 per share, or 21.7 times.

He stresses these figures are compute only, with no value attributed to launches, Starshield, Starlink or other businesses. He also computes that 655 shares, which he describes as a retirement holding he has used before, would be about $540,000 at $825 and about $2.4 million at the bull-case price.

The host presents two polls. In the first, 78% of respondents chose 100 SpaceX shares over 0.18 Bitcoin, with each valued at about $15,000. In the second, 74% chose SpaceX and 26% chose Bitcoin.

The host says he believes AI is the faster horse, while Bitcoin is there in case the world goes south. He closes by thanking community members and saying he expects an interesting earnings period.