3 Stocks to Play SpaceX’s $28.5 Trillion Pivot

Elon Musk just filed paperwork with the SEC that contains the largest number ever printed in an IPO document.

$28.5 trillion.

That's not a typo. That's what SpaceX is telling prospective investors what it believes to be its Total Addressable Market opportunity.

For context: the entire U.S. economy clocks in around $32 trillion a year. SpaceX is claiming the right to chase a market almost as big as everything America produces.

And here's the wildest part. Most of it has nothing to do with rockets.

When you think SpaceX, you think Starlink. You think Falcon 9. You think Mars.

That's the company you used to know.

The new SpaceX — the one Musk is taking public this summer at a reported $1.75 trillion valuation — is something else entirely.

According to its S-1 filing, more than 90% of SpaceX's $28.5 trillion addressable market is tied to artificial intelligence for businesses.

Not consumer chatbots. Not Grok answering trivia at parties. AI that lives inside the world's largest companies and runs their operations: Logistics, defense systems, industrial automation, financial modeling, and healthcare.

Here's the part most retail investors miss.

When SpaceX prices its IPO this summer, you won't get in at the offering price. Goldman will. Fidelity will. The hedge funds will.

By the time SpaceX shares hit your brokerage screen, the easy money is already booked. The "pop" you read about happens before retail can click buy.

Worse — there's no guarantee it pops at all. SpaceX lost roughly $4.9 billion last year. The xAI unit alone burned through $6.4 billion. The company spent $12.7 billion on AI capex in 2025 — more than rockets and satellites combined. Wall Street analysts are already throwing around terms like "220x EV/EBITDA," which is a polite way of saying the IPO will be priced for absolute perfection.

So, there's a better way to play the $22.7 trillion enterprise AI tidal wave. And it doesn't require waiting on an IPO that hasn't even priced.

The thing about a $22.7 trillion gold rush is that it doesn't start when the gold gets pulled out of the ground. It starts when somebody builds the saloons. When somebody sells the picks.

These three companies are already doing it. None of them is a household name. All three are getting paid right now while the talking heads on CNBC argue about Nvidia's next earnings call.

Comfort Systems USA installs heating and air conditioning.

I'll say that again, because it's the most boring sentence you'll read today. Comfort Systems USA installs HVAC systems.

But here's what nobody's telling you. Every AI data center being built in America right now needs industrial-grade cooling. The chips that run modern AI models throw off so much heat that without specialized thermal management, they melt. Literally.

The numbers are starting to show what that means for Comfort Systems' business.

Last quarter, Comfort Systems reported revenue of $2.9 billion, up 56% year over year. Same-store growth came in at 51%. Gross margins hit an all-time record of 26.3%. Earnings per share more than doubled to $10.51. The company's backlog reached a record $12.5 billion — that's roughly the entire market cap of most mid-cap tech stocks.

Where's all that work coming from? Data centers. Technology customers now account for 45% of revenue, up from 33% a year ago. Hyperscalers and enterprise AI builders are signing contracts faster than the company can break ground.

The stock is up 77% in 2026 already. It's outperforming Nvidia year-to-date. And it's still trading at a fraction of the multiple Wall Street pays for the headline AI chip names.

Comfort Systems doesn't need an IPO pop to make money for shareholders. It just needs Big Tech to keep building data centers, which is exactly what every hyperscaler has committed to do through 2030.

If you want to see what enterprise AI actually looks like in the real world — not in a SpaceX filing, but on an active factory floor — visit a Walmart distribution center.

Symbotic's robots are in all 42 of them.

Symbotic builds AI-powered warehouse automation. The company's platform uses machine learning to orchestrate fleets of autonomous robots that move pallets, sort inventory, and build truck loads. It's the dirtiest, least glamorous version of AI imaginable. It also works.

Walmart liked the technology so much that it sold its own in-house robotics business to Symbotic for $200 million, then turned around and committed another $520 million to expand the partnership into store-level fulfillment. That single deal added more than $5 billion to Symbotic's backlog and opened a new $300 billion U.S. market for micro-fulfillment.

Walmart owns a big chunk of the company. So does SoftBank.

Symbotic turned profitable in early 2026, beating analyst estimates by 625% in one quarter. Gross margins on technology deployments run 80% to 85%. To put that in perspective: those margins are higher than Tesla's. They're higher than just about anyone's.

This isn't a speculative AI bet. This is enterprise AI, deployed at scale, getting paid every single quarter.

And it's still flying mostly under the radar.

You can't have AI without data. Massive amounts of clean, labeled, evaluated data.

That's where Innodata comes in.

Innodata is a small-cap data engineering firm that supplies the training and evaluation data behind some of the biggest AI companies on Earth. Five of the Magnificent Seven are customers. Foreign sovereign AI programs are signing contracts. Federal agencies are starting to bite — the company even brought a retired four-star general onto its board to lead its government push.

In 2025, Innodata's revenue jumped 48% to $251.7 million. Adjusted EBITDA soared 68% to $57.9 million. Cash on the balance sheet nearly doubled to $82.2 million. Management is guiding for at least 35% growth in 2026 — a number the CEO has hinted is deliberately conservative.

Here's the kicker.

Innodata is already helping companies build the exact thing SpaceX says it wants to sell — agentic AI that can run inside enterprises. They provide model evaluation, agent testing, adversarial security, and even physical AI datasets for robotics. While SpaceX is filing a TAM number, Innodata is filling purchase orders.

The company is profitable. The market cap is under $2 billion. Most retail investors have never heard the name.

Total addressable market figures in S-1 filings tend to use lofty numbers. Uber claimed $5.7 trillion when it went public in 2019. SpaceX just printed a number nearly five times bigger.

The headline is designed to grab attention. And it worked.

But strip away the marketing, and the number underneath is real. Enterprises are spending. Hyperscalers are building. The supply chain that powers all of it — the cooling, the data, the deployment — is already in motion.

SpaceX IPO and TAM figures

Comfort Systems USA (FIX)

Symbotic (SYM)

Innodata (INOD)