SpaceX IPO Update: Google and Cursor Deals, Other Valuations, My Modeling Process, and AI Tools

I rarely publish “follow-up” articles or videos on this site, but I felt it was warranted for the SpaceX IPO since we’re witnessing financial history.

Also, we received dozens of great comments and questions across the valuation video and article, so I wanted to respond to at least a few of them here.

I will also discuss a few valuations from other sources, such as Professor Damodaran at NYU and Morningstar, as well as my own process for researching and building the SpaceX valuation.

And yes, I will value either OpenAI or Anthropic once their S-1’s are publicly available.

Some parts will be easier (just 1 segment rather than 3 different divisions), while others will be harder (long-term unit economics).

We will eventually add modified versions of all these case studies (closer to step-by-step walkthroughs) to the Venture Capital & Growth Equity course as well.

OK, now back to SpaceX:

Video Version of This Update

Files & Resources:

• SpaceX IPO Update – Presentation Slides (PDF)

Video Table of Contents:

• 0:00: Introduction
• 2:04: Part 1: IPO Update, Google, and Cursor
• 4:48: Part 2: Other Valuations: Damodaran and Morningstar
• 7:53: Part 3: My Modeling Process (Starting Points, Starlink)
• 12:33: Part 4: Process for Other Segments
• 13:33: Part 5: AI Usage in Modeling
• 15:14: Part 6: 1-Year Predictions and Uber Comparison
• 16:54: Recap and Summary

SpaceX IPO Performance Update

As of Wednesday morning, SpaceX has traded up to almost $200 / share following its IPO at $135 / share, which implies a valuation of over $2.5 trillion.

While this is absurd, it still falls short of my expectation for $5 trillion – but we’ll see what happens by the end of the week.

It could crash, but if you look at Tesla’s performance over the past 5 years, an equally plausible outcome is that it thrashes around within a wide range and is more of a “volatility maxxer” than anything:

Of course, even dumber things could happen, such as a Tesla / SpaceX merger, which would be the equivalent of AOL / Time Warner in January 2000 at the peak of the dot-com bubble.

We also got a few questions about how the recent Google compute deal (for over $900 million per month) and SpaceX’s just-announced acquisition of Cursor affect things.

The short answer: “They may increase the company’s implied value by $50 – $100 billion with some very generous assumptions, but it’s still not that much for a company that’s overpriced by at least $1 trillion.”

For reference, our forecasts in the AI segment already included a very generous increase in “compute rental” revenue and AI solutions revenue, so it’s not even clear how much these deals improve the forecasts.

However, to play devil’s advocate, if you assume significantly higher “Unused Compute Capacity Revenue” in line with the Google deal and add a bit more on top of both this one and the Anthropic deal, the valuation potentially increases by $20 – $40B:

The initial margins here are still negative, which limits the impact, but if you:

• Assume positive margins on these specific contributions (compute rental deals).
• And factor in the impact of Cursor’s few billion per year in revenue, presumably at very low or negative margins.
• And assume no negative impact on the rest of SpaceX due to reduced compute capacity.
• And subtract the $60B SpaceX is spending to acquire Cursor.

Then these deals could potentially increase SpaceX’s implied value by $50 – $100 billion, which would normally be a lot, but this is now a $2.5 trillion company, so… not really.

I certainly don’t see how these deals could boost the overall valuation by $1 trillion or even $500 billion.

Other Valuations: Damodaran and Morningstar

Here’s a quick summary of Damodaran’s valuation and Morningstar’s vs. the one on this site:

Damodaran is far more optimistic, while Morningstar is closer to my view (note that in the Morningstar case, their Expansion/Moonshot Revenue is in a separate scenario and does not add to the baseline Revenue shown here).

The long-term margin assumptions are similar between all 3 models, but the Revenue, Discount Rates, and treatment of “Expansion” or “Moonshot” opportunities differ.

Damodaran has 2036 Revenue at about twice my number ($420 billion vs. $211 billion), and he uses a much lower Discount Rate, closer to 8%, through the entire forecast period.

He defines “Reinvestment” as CapEx – Depreciation +/- Change in Working Capital, and it’s within range of the 2036 number in our model.

He also assumes major new business lines from opportunities like orbital data centers and lunar tourism, producing $100 billion in Revenue by 2036, while I assume no contributions there.

My main issue here is not whether he’s too optimistic or speculative, but that he doesn’t go into much detail on the market size/share or unit economics driving this forecast.

For example, what is Starlink’s Total Subscriber count by Year 10 or 20? How does this compare to other telecoms? What percentage of the commercial vessel market will it capture?

For a high-growth company like this, you need “bounding” assumptions to make sure you don’t end up with 1 billion Starlink subscribers.

The lack of scenarios for such a speculative company is also an issue, given the wide range of potential outcomes and the time frame involved.

With the Morningstar valuation, the overall approach is more defensible because they build in different cases and probability-weight each one, landing on a $780B estimated Equity Value, or $63 per share.

Their overall 10-year numbers are also closer to ours in their Base Case (these exclude the “Moonshot” opportunities in one of their cases):

I would like to see more details on their unit economics, CapEx, and Discount Rate assumptions, as it’s difficult to judge the full analysis without those.

My Financial Modeling Process

I also got a few good questions about my financial modeling process for this exercise.

My first tip is “Don’t pick this type of company” if you’re new to valuation or modeling.

It’s too complex, and too much judgment is required for the assumptions.

If you’re new, pick something much simpler, like Walmart or a restaurant company with a single segment and low growth rates.

If you want to torture yourself by picking a company like SpaceX, though, I recommend starting with the smallest part that is still useful OR the most important part.

In this case, I knew that Starlink would be the most important part, since it was already EBITDA- and cash-flow-positive.

It also had the most supporting data and the most solid comparable companies, so I decided to start there.

Initially, I planned to multiply the Subscriber Count by the Average Revenue per User (ARPU) to determine Revenue.

But this didn’t work because Starlink has a “Commercial” business selling to ships and airlines that’s separate from the Consumer division:

Unfortunately, the S-1 provided very limited information on the Commercial division, so I poked around online to find estimates of the monthly rates they charge vessels.

The SpaceX / Starlink website also had some quotes.

The prices spanned a wide range, but I settled on an estimate of $10K per month per vessel, under the logic that this was close to various external estimates and produced a reasonable number of “Implied Vessel Customers”:

At this point, I started making some rough assumptions for how the Market Share or Market Penetration in each segment would change over time.

But a key question kept coming up: What was a reasonable *maximum* for the Market Penetration?

Yes, Starlink has grown its customer base at an impressive rate, but no company can double its Subscribers or Market Penetration forever.

To get reasonable numbers, I turned to the comparable public companies.

I used Capital IQ to gather sets of comparable telecom firms, and I used some of their subscriber counts as reasonable “upper bounds” on the eventual Market Penetration for Starlink:

Based on this, I decided it was reasonable for Starlink to reach anywhere from 20% to 40% of its addressable market over 20 years:

I continued that same process for other drivers, such as the EBITDA Margins and CapEx.

I also changed the model setup by removing the initial hard-coded assumptions and linking the inputs to the “Control” tab instead, which made it easier to set up and test different cases.

With the overall valuation for this segment, I wanted to see substantial but not overwhelming differences.

For example, $150B vs. $230B vs. $330B might be reasonable, but something like $0 vs. $1 trillion vs. $2 trillion would not be:

I tweaked some of the market share assumptions but eventually landed on a set that produced decent spreads.

In most models and valuations, we prefer to create scenarios around 1 – 2 key drivers, such as the market share or market penetration here.

If you vary more than that, it’s difficult to “calibrate” the model and get reasonable output across all the cases because there are too many conflicting variables.

SpaceX: The Other Segments

After finishing Starlink, I did something similar for the other segments.

AI is similar, but Revenue is split into 3 categories rather than 2, and the “Neocloud” and “AI Tools” segments depend heavily on compute capacity, which is supply-constrained and limited by physical construction timelines.

But I felt a “Market Share” approach still made sense because construction is underway on tons of data centers, and SpaceX will capture only a small percentage of the total.

For the Space division, I decided against a Market Share / Market Size approach because it already owns nearly the entire market.

Its share is unlikely to change much, and real external demand growth is quite low, so it seemed reasonable to use modest growth rates here:

AI Tools for Financial Modeling

We also got a few questions and comments about how to “integrate AI tools” or “use AI for financial modeling.”

I published some thoughts on AI in financial modeling last year, and I’m aware of the advancements since then.

But my opinion hasn’t really changed; if you are learning these concepts for the first time, you should do it manually because “putting in the reps” is the only way to learn.

In my experience, most people who claim to “use AI” for their modeling work do not actually check the output or fully understand what they’re doing, which creates problems when they have to explain or justify it.

Just for fun, I experimented with the premium version of Grok for this SpaceX valuation and went through at least a few hundred prompts to research companies, data sources, and various ideas.

It saved me some time here and there, but it’s not as if it reduced a 100-hour task down to a 50-hour one; it saved maybe a few hours over ~2 weeks.

My recommendation is to use these tools to dig into companies, do research, and cross-check numbers, but avoid using them to “think for you,” as you will never be able to explain or justify anything that way.

What’s Next for SpaceX?

Overall, the IPO so far has matched my expectations.

SpaceX is ridiculously overvalued at $2 trillion, let alone $2.5 trillion, but I would not be surprised if it keeps going up.

The 1-year and 2-year outlooks are murkier because there’s a long history of hyped companies going public, “popping,” and then falling substantially.

With the Uber valuation in 2019, I felt more comfortable being openly negative because the market was slightly more rational, and fewer retail bag holders were involved.

But given the current markets and Musk’s involvement, I don’t feel confident about anything.

And if Tesla and SpaceX merge, buckle up because that could just be the “January 2000 AOL / Time Warner merger peak bubble” moment.