Key points:

• Big payday for early backers

• AI infrastructure and apps become the main growth drivers, guarded by space capacity and compute power moats

• The long-run multiple compression is imminent, but there’s significant runway for growth thanks to a large TAM

• Our verdict - below

Sending multiples to orbit

Let’s take a look at SpaceX’s financials from S-1:

Though the official price tag isn’t known yet, the company is reportedly aiming for a $75 billion raise at a $1.75 trillion valuation, which corresponds to:

• $598 price per share ($1.75T / 2,926 million shares)

• 94x EV/S multiple ($1.75T / $18.7B revenue)

We also finally know the share prices of early funding rounds. Spoiler - early investors are taking home loads of money:

94x EV/S multiple is already in the stratosphere, and it’s a matter of time before it collapses to some long-run equilibrium, a common dynamic for tech companies after going public. To estimate the equilibrium, though, we need a peer group to compare against. Which is tricky, because after a series of mergers, SpaceX became four companies in one: an AI company (xAI), a telecommunications company (Starlink), a social media company (X.com), and - finally! - a space launch company.

How does all this fit under one roof? The company argues it’s actually one machine powered by three “foundational competitive advantages” - Space, Connectivity, and AI - that compound. The narrative: use cheap multi-use rockets to deploy satellites and space compute clusters; use the established telecom network and compute power to deliver cheap AI inference to customers on Earth (including itself); use that inference capacity to drive AI adoption at an unprecedented scale.

The point is, SpaceX is no longer pitching itself as a pure launch provider but as a vertically integrated infrastructure player across those three sectors, expecting most of its future growth to come from AI:

Within AI, the company targets two critical layers of the stack: the compute infrastructure and the front-end applications. On the compute side, it’s relying on global demand for data centers and is already renting out its Colossus cluster to Anthropic. On the applications side, it’s pushing an enterprise AI platform, Macrohard, to automate enterprise workflows and capture a slice of the impending agentic economy.

This suggests we’re evaluating an AI company. Having that figured out, we can compare SpaceX to NVIDIA (the clearest example of an AI hot stock), Tesla (included in our most recent AI stocks overview and whose valuation incorporates the “Elon Musk” factor), and the NASDAQ median (just for reference):

At those pricing levels, the only clear strategy for IPO investors to make their money back is waiting until the company outgrows the multiple compression, which seems imminent. This is similar to how venture investors view startup investing, and with startups, the long-term growth story is what matters. Let’s take a look from that angle.

Hit the target no one else can hit

Technically, startup valuation comes down to predicting:

• The terminal revenue (being measured top-down as “TAM x share” or bottom-up as “units x price”)

• Exit EV/S multiple (derived from where the markets have historically traded)

• The probability of the terminal revenue target getting hit (this parameter is implicit, vague, and the source for most of the errors, yet the most important)

Putting everything together, Terminal Revenue * Exit EV/S multiple * Probability = Expected Enterprise Value, or E(EV). For SpaceX investors to earn anything back, that number has to be greater than $1.75T.

To deliver a 15-25 percent annualized return1, the company’s market cap needs to reach between $7T and $16T by 2036:

That won’t happen without revenue skyrocketing, with the exact number required depending on the long-run EV/S multiple. Using Tesla’s value of 10.08x (which conveniently factors in Musk’s insane productivity and the AI & hardware mix), required revenue falls between $702B and $1.6T:2

Only one company has ever reached that mark - Amazon, with its FY 2025 revenue of $717B, and whose founder is pouring resources into Blue Origin, the second launch operator in history to successfully reuse a rocket.

While this seems distant compared to SpaceX’s most recent $18.7B mark, it doesn’t completely defy statistics: 10 years ago, when Bezos was 52 (Elon’s 54 right now), and Amazon was 22, the company’s reported annual revenue was $136B, while Tesla’s most recent results were $95B in revenue (with $4B in net profit3).

The last piece of the valuation puzzle is the conditional probability that Musk can pull it off, expressed as P(Revenue > $702 billion | Multiple milestones jointly reached). If anything falls short and the revenue fails to reach the target, the model might crash back to Earth, together with realized returns for IPO investors.

What’s the single biggest thing that might go wrong? There are numerous stronger contenders for the largest segment of the TAM: the $22.7T enterprise AI market. Even if AI agents spread like wildfire, xAI has to win against Microsoft, OpenAI, Google, Anthropic, and Amazon, as well as vertical AI vendors like Salesforce, ServiceNow, Palantir, Oracle, and SAP.

While Musk can still out-innovate everyone over time, as he has already demonstrated with his multi-industry track record, relying on that assumption to achieve a bare 15% annual return makes little economic sense.

Our verdict

As Musk likes to say, the company’s mission is to make life multiplanetary and reach stars. Is making our civilization’s dream come true worth underwriting an obviously overpriced IPO? For non-financial reasons, many people would say “yes”, which guarantees oversubscription. For a rational investor, however, patience is a better strategy.

Given aggressive pricing, many factors need to align for this IPO to deliver good long-run performance, making the entire construction too shaky. Our verdict: if you want to ease the FOMO and support the mission, put a little, but if you want to make money, wait for a better entry price4.

Investors love to debate whether AI stocks are in a bubble. To make up our mind, we mapped 77 companies across the AI industry, and the picture is mixed. Two-thirds of them look overpriced right now, but if we factor in projected revenue growth, everything flips: suddenly, two-thirds look underpriced, and some really interesting opportunities pop up. Let’s dig in.

AI industry map: from atoms to bits and back

First, let's unpack the AI industry. Here is how value moves within the AI stack - from raw materials and energy all the way down to AI-powered applications. But while end-customer value flows from atoms to bits, shareholder value flows in the opposite direction (along with revenue and profits). The diagram arrow shows the latter.

We break the industry down into three main layers: Physical upstream, Software & models, and Downstream apps. We borrowed this upstream-downstream logic from the oil and mining industries. Let's zoom into each.

LAYER 1: The physical upstream is where AI models (and many apps) live in the physical world - and how they get their power. It breaks down into:

• Materials. The raw resources used to build computer chips, servers, switches, cables, and cooling systems. The US Department of the Interior publishes a comprehensive list of these minerals to spark debate about how to fix supply chain risks. Here are the tickers we track (with more coming soon):

• Energy. Data centers need cheap, reliable, and ideally clean power. Tickers:

• Hardware. The AI economy runs on fast chips, and it needs lithography equipment to build them. Two giants hold near-monopolies here: ASML (for lithography machines) and Nvidia (for GPUs). Tickers:

• Data centers. This is where the actual computing power lives - the most critical resource for companies building AI models. While data centers rely on lots of software, this subgroup focuses strictly on hardware infrastructure. Tickers:

LAYER 2: Software & models is the absolute core of the industry. Breakthroughs in model architecture are what triggered the current AI boom, supported by specialized software for compute cluster management and chip design. This layer includes:

• Cloud platforms. To run AI models, you need a highly specialized software stack sitting on top of data center hardware. Since most data center operators don't want to build this software layer themselves, specialized AI clouds step in - and many end up building their own infrastructure too. Tickers:

• Chip design. Electronic Design Automation (EDA) tools: a small but stand-alone software subgroup. Tickers:

• Data. This is the thinnest layer in the stack, but managing data workflows for AI is still a real, stand-alone software business. Tickers:

• Models. This is the brain of AI, and it’s a pure battlefield. It’s where the world's top talent and massive capital are being poured. Most frontier companies here are still private, though secondary market valuations for Anthropic and OpenAI are already closing in on $1 trillion. Tickers:

LAYER 3: Downstream apps is the user-facing layer driving mass adoption. It includes:

• Smart devices & edge AI. This is where robots, cars, and gadgets get an AI brain. Many legacy hardware giants are adding customer-facing AI to their tech, which puts them straight into this category. Tickers:

• Consumer & prosumer AI apps. Chatbots, image generators, and software tools are exploding in popularity simply because the underlying models have gotten good. Tickers:

• Enterprise AI software. Unlike consumer and prosumer tools above, this category is all about software built for corporations and governments. Tickers:

Now we have 77 companies (65 public and 12 private) on our watchlist across 11 sub-industries. Let's look at how underpriced they actually are - in other words, their upside potential.

AI stocks: upside estimates

We can evaluate 65 of our 77 companies - specifically the ones that are publicly traded and file regular financial reports. To calculate upside potential, we look at three years of quarterly statements, consensus revenue projections from major investment banks (like Morgan Stanley, Citigroup, and JPMorgan), and daily stock prices. Our core assumption is simple: a stock price reflects the market's estimate of future cash flows left over for shareholders. We can express this with a standard formula:

We can rewrite this as a valuation multiple:

Then replace FCFE with revenue and margin:

For fast-growing companies, margin numbers can be noisy. That’s why we swap the P/FCFE multiple for EV/Sales:

This ratio is usually stationary1 (meaning it tends to fluctuate within predictable boundaries), making it a reliable anchor for estimating fair value. As you can see below, Nvidia’s EV/Sales multiple generally trades right around its 1-year median:

Our upside metric is equal to the gap between a stock's current EV/S multiple and its historical median:

Right now, only 25 out of 65 public companies show a positive upside. But here is the trick: if a company's revenue grows while its stock price stays flat, its multiple drops. That widens the gap and increases the upside. Once we factor in future revenue projections, the number of underpriced stocks jumps from 25 to 46.

Among the underpriced (positive upside):

• NVDA

• CRWV

• MSFT

Among the overpriced (negative upside):

• ASML

• TSMC

• DOCN

Apparently, we won't make an investment based on this metric alone. It’s just a screener to help us see the big picture and figure out which companies are worth digging into.

The full breakdown of AI stocks by sub-industry is below. The actual upside estimates and underlying math are reserved for our paid subscribers. Let us know in the chat if you want this sent over as an Excel or Google Sheets file!

Read more

While there are myriad investment ideas and strategies, we focus on equity investments in innovative growth companies because this is where the best investment returns are found. For example, from 1991 to 2026 (which includes the Dotcom Bubble of 2000), the NASDAQ-100 produced an annualized total return of 15.65% vs. 11.25% for the S&P500. There are several fundamental economic reasons for such outperformance, which we explore below.

Equity outperforms all other asset classes

Have you ever wondered which is the highest-yielding asset class in the world? Equity, or ownership stakes in public and private companies, consistently outperforms everything else globally. This outperformance is well-documented by many research teams and papers: for example, the research report from Goldman Sachs or this 200-year data panel from Deutsche Bank:

Financial market economists call this statistical phenomenon an Equity premium puzzle and still debate why it exists. The reason is rooted in how the economics of returns work across all asset classes:

• Cash simply doesn’t produce any yield.

• Bills and bonds yield a risk-free rate + a risk premium (which is lower than an equity investor would demand since debt is inherently less risky).

• Commodity prices tend to grow at the inflation rate (with occasional price spikes in some commodities that then revert to their average levels).

• Real estate combines rental yields (roughly equal to fixed-income yields) and price gains (roughly equal to inflation). If we add them together, the resulting yield can be high, which may put pressure on the thesis, according to some data, but both those components are still bound to the aforementioned economic laws.

• There are also cryptocurrencies, which have delivered unmatched performance since the creation of Bitcoin in 2009, but it might still be too early to say where they’re heading compared to traditional asset classes, which have been around for hundreds of years1.

At the same time, equity returns are linked to the profits of the underlying businesses, which tend to grow at the GDP growth rate but can often outpace it, as companies naturally fight for larger market shares and stock exchanges act as a selection mechanism to pick winners:

The global competition and the pursuit of greater profits both fuel capitalism to invest more and more resources in R&D, new technologies, means of production, the marketing of new goods and services, and distribution infrastructure. And equity is technically the only way to capture those ever-increasing profits and the primary source of first-generation wealth worldwide: 59% of millionaires made their first million through entrepreneurship (i.e., owning equity in a business). And almost all billionaires worldwide derive their wealth from ownership stakes in their own companies.

If we don’t want to start a (new) business yet want to have a tailwind for our returns, we should look at others’ businesses. We have to start looking inside the investable equity universe.

Revenue growth explains most of equity returns

Which direction do we go once we’re inside that universe? To answer that question, it’s worth understanding what the main driver of profit growth and hence equity returns is.

Shareholders want their companies to make money to survive and thrive. And making money requires selling products to customers. This brings revenue, which is the primary regular source of cash inflows. The cash is used to cover operating expenses and capital spending, which are the primary destinations of cash outflows. The difference between all outflows and all inflows available for distributions to shareholders through dividends and buybacks is called Free cash flow to equity (FCFE) and converges to accumulated net profits in the long run.2

If we divide net profits by revenue, we get what’s called the margin. Profits originate from revenue and flow to shareholders’ pockets through margin. Because margin can never exceed 100 percent of revenue and is typically lower in practice, the only way a company can accumulate large net profits is by substantially increasing its revenue. In 2014, researchers from BCG confirmed this intuition, and a team from Morgan Stanley arrived at similar results:

The result also extends into the future per J.P. Morgan’s Long-Term Capital Market Assumptions:

That doesn’t, of course, mean that profits are not important. At the end of the day, shareholders are expecting distributions, and without profits, there’s no free cash flow to pay those distributions from. We’re just saying that significant profits aren't possible without significant revenue.

On average, companies’ revenue grows at the same rate as GDP, which is 2-4% in most countries. If a company grows faster, that means it’s taking over some other player’s market share, represented by that player’s revenue divided by the total GDP. Because GDP is ultimately the sum of all spending in an economy, the company that wants to outgrow the market must fight to get a bigger share of that spending.

If we want higher returns, we should look more closely at growth companies that expand faster than GDP by taking over competitors’ market share.

Innovations are critical to outsized growth

If a company wants to win a larger share of customers’ budgets, it must introduce a superior new product at a better price, or it will struggle to lure customers away from their existing suppliers. Of course, unless it’s a monopoly, but today monopolies are battled and are considered rare.

To find a recipe for making superior products, we can look at entrepreneurs with consistent track records of creating them, of whom Elon Musk is the best example. In a video recorded at SpaceX HQ in 2017, Sam Altman, the then-CEO of Y Combinator, has the following dialogue with Musk:

Not being useful is what kills most young companies (which die before growing into big companies, hence fewer big companies either), so this conversation directly relates to developing and launching new products. The marginal customer utility that such a product creates can be regarded as a formal definition of its superiority. How to achieve that superiority? By fostering innovations.

What is innovation? Basically, any practical implementation of a new idea that creates improvements in goods, services, and distribution. Rather than just occasionally innovating, companies should maintain a constant high pace of innovation:

Though Musk’s original commentary focused on technology companies, he later generalized it to apply to any company trying to stay ahead of the competition. Over time, the pace of innovations translates into a deep advantage of a company’s product over substitutes, enabling it to outsell and outgrow its peers.

Indeed, the study by NASDAQ supports the hypothesis that companies that spend more on R&D (innovation) demonstrate higher shareholder returns:

The NASDAQ-100 index itself, widely known as a collection of innovative stocks, has significantly outperformed the S&P500 on a total return basis, even if we start to count before the Dotcom Bubble:

What’s the main driver behind innovations? We can learn the answer from an industry that made recognizing and supporting innovations its everyday business: venture capital investing. VCs argue that to produce more innovations, it’s critical to have outstanding people on a team, which they aim to capture by measuring the “talent density” of the startups they screen for investment. We’ll explore this idea in depth in our follow-up notes.

What’s next

The key takeaway for now is that to capture faster-than-average growth and premier investment opportunities, we need to identify companies with a high pace of innovation and a high density of talent. In our next article, we’ll explore where innovations and talent (and therefore growth) tend to concentrate, and how big the investment returns could potentially be3.