What Happens When You Ask Bridget to Research the Biggest IPO in History?

Written with Bridget, powered by Context Analytics & the Elastic Filings MCP

I drove from St. Louis, MO to Anchorage, AK in a converted van while working remotely with a Starlink mini on the dash. Through national parks, the Yukon, stretches of Canada where the nearest town was hours away, and for most of the trip, the connection was more reliable than it had ever been sitting in an office in Chicago or St. Louis. That got me curious about the business behind it, and curious enough to dig into whether the biggest IPO in history is worth the price. 

So when SpaceX filed its S-1 in May 2026 targeting a $1.75 trillion IPO valuation (the largest in history). I had a simple question: is this a good investment? I wanted to dig into the S-1, but I didn't have time to read 458 pages. With Bridget, I didn't have to.

Bridget is Bridgewise's conversational AI, an all-in-one chatbot for investments, built from the ground up for regulatory compliance and capital markets precision. We've been expanding what she can see. Context Analytics’ social sentiment data is already live, and corporate filings intelligence is close behind. This analysis was an early look at what that combination can do. My goal here isn't to tell you whether to buy SPCX. It's to show you what happens when a regular person with a real-world observation gets access to institutional-grade tools.

SpaceX's S-1 shows $18.7B in 2025 revenue, with Starlink as the only profitable segment. The AI division is burning cash. The Mars clause is real. Bridget found all of it in under a minute.

What the S-1 actually says

Most IPO coverage is based on press releases and analyst notes. Bridget AI reads the actual filing. Using the parsed corporate filings database, which indexes SEC documents including the full SpaceX S-1. The core numbers are pulled directly from the source.

SpaceX runs three business segments: Space (rocket launches and government contracts), Connectivity (Starlink), and AI (xAI/Grok, X, and the Colossus data centers). The xAI and X entities were merged into SpaceX in February 2026 through a common-control reorganization, meaning Elon Musk held controlling stakes in all entities, so no goodwill was recorded. SpaceX retroactively combined all three companies' financials in the S-1.

One detail buried in the filing that most coverage missed: SpaceX acquired FCC-licensed AWS-4 and H-Block spectrum from EchoStar in November 2025, specifically to power Starlink Mobile's direct-to-cell capability. That's not just a product feature. That's a multi-billion dollar regulatory moat. You can't launch a satellite-to-mobile service without licensed spectrum, and SpaceX just bought it.

FROM THE S-1 VIA ELASTIC FILINGS MCP: 2025 REVENUE BY SEGMENT

Source: SpaceX S-1 filing, May 20 2026, via Bridget corporate.

The Average Revenue Per User (ARPU) question: red flag or rational strategy?

One of the first things Bridget flagged from the S-1 was that Starlink's ARPU drop from $99/month in 2023 to $66/month in Q1 2026, a 33% compression in three years, even as subscriber count grew from 2.3 million to 10.3 million. I asked if that was a red flag.

That answer changes how you read the number. ARPU as a standalone metric looks like erosion. ARPU as a deliberate strategy to capture a market that traditional Internet Service Providers (ISPs) wrote off looks like a land grab.

Here's the part that clicked for me on the road: in rural Canada and Eastern Alaska, Starlink dishes were on houses, not as a luxury upgrade, but as the only internet option available. Traditional ISPs will never run fiber to those communities. The unit economics don't work. Starlink's unit economics do, at $45–$65/month and no physical infrastructure investment required from the customer.

According to the GSMA’s State of Mobile Internet Connectivity 2025 report, 3.4 billion people remain offline, with 70% living in rural areas where ground-based infrastructure will never reach (ITU/GSMA, 2025). That's not a slice of the existing Total Applicable Market (TAM), that's a market that doesn't exist yet on any analyst's model. Starlink is the only technology with a credible shot at it.

STARLINK: ARPU VS. SUBSCRIBER GROWTH (FROM S-1)

ARPU compression driven by deliberate international pricing expansion. Starlink added ~1M subscribers every two months in early 2026.

Is the $1.6T connectivity TAM real?

SpaceX claims a $1.6 trillion TAM for its Connectivity segment: $870B for broadband, $740B for mobile. Most third-party estimates put global broadband at $565B–$1.15T. Bridget walked me through the gap.

The broadband TAM as traditionally defined assumes the market already exists. Starlink's TAM includes people with no viable option. The addressable market for people who currently have no reliable internet is enormous, and it doesn't show up in legacy market sizing models that count existing ISP subscribers.

SPACEX TOTAL ADDRESSABLE MARKET: $28.5T TOTAL (FROM S-1 VIA ELASTIC FILINGS MCP)

93% of SpaceX's claimed TAM is AI, primarily enterprise AI applications. Whether you believe the $1.75T valuation lives or dies in this column.

Can Starlink replace your cell service?

Starlink Mobile crossed 10 million users in early 2026. But the path to displacing traditional carriers isn't obvious. The S-1 doesn't give a clean answer. What it does show is the EchoStar spectrum acquisition. SpaceX spent billions buying FCC-licensed spectrum specifically to enable satellite-to-cell handoffs. Combined with the T-Mobile partnership already live for dead-zone coverage, the most likely near-term model is hybrid: terrestrial towers in cities, Starlink satellites everywhere else.

Thinking about it from the van I wasn't in a dead zone that T-Mobile couldn't reach because of some failure. I was in terrain where building towers is economically impossible. No regulatory change fixes that. No 6G rollout touches it. Starlink does.

LIVE SENTIMENT, CA S-FACTOR FEED (JUNE 8, 2026): SPACE-NATIVE VS. LEGACY CARRIERS

SMean = 20-day average S-Score (Z-score vs. historical baseline). Space-native stocks show 30-40x stronger long-term sentiment than legacy carriers heading into the SpaceX IPO. Live data from Context Analytics S-Factor feed.

The moat nobody's talking about: orbital real estate

Every successful tech company has exploited a structural gap between what the technology could do and what the regulations had caught up to. Amazon didn't collect state sales tax on online orders for years, not because they were cheating, but because the rules hadn't been written yet for what they were doing. By the time the rules caught up, they'd already built the infrastructure advantage that made it irrelevant.

SpaceX is doing the same thing in low Earth orbit, and most retail investors aren't thinking about it.

Scientists estimate LEO can safely hold roughly 100,000 satellites before orbital debris and collision risk become unmanageable at scale. SpaceX is currently operating ~9,600 satellites, just below 10% of that capacity. They're not just running a satellite internet company. They're the dominant occupant of a finite, non-renewable resource that every future tech company depending on space infrastructure will need to negotiate with.

LEO ORBITAL CAPACITY: ESTIMATED SAFE CEILING ~100,000 SATELLITES

Sources: ITU filings, FCC records, UCS Satellite Database; Brad Stone, The Everything Store (2013). SpaceX also filed for a million-satellite orbital AI mega constellation. That filing is a land claim, not just a product roadmap.

The regulatory parallel to Amazon goes deeper: ITU orbital slot allocations operate on a use-it-or-lose-it basis. You file for slots, you deploy, and you establish priority. SpaceX filed to deploy a million-satellite orbital AI data center mega constellation. That's the same playbook Amazon ran with same-day delivery infrastructure, and nobody else has the rocket to compete with it.

Data centers in space: the Anthropic deal as a preview

In May 2026, two weeks before the S-1 was filed, SpaceX announced that Anthropic would pay $1.25 billion per month or $15 billion a year to lease compute capacity at Colossus, SpaceX's terrestrial AI data center in Memphis. From a company that SpaceX's own AI division competes with directly.

The 90-day out is real. This isn't locked revenue. But it does validate something more important: external AI companies are willing to pay premium rates to access SpaceX's compute infrastructure. That's the terrestrial version of a thesis that gets more interesting in orbit.

Think about why data centers are so hard to build on Earth right now. It comes down to three things:

• Energy: Data centers are power-hungry, grid-constrained, and facing growing carbon scrutiny. In orbit, that changes entirely. Space-based solar arrays receive approximately 1,361 watts per square meter compared to a maximum of about 1,000 watts per square meter at Earth's surface, and operate at 28-30% efficiency compared to 15-22% for terrestrial panels. No clouds, no night cycle, no atmospheric losses.
• Cooling: Up to 40% of a data center's electricity use goes to cooling, according to the National Renewable Energy Laboratory — consistently one of the largest operational costs in the industry. In space, the void handles it passively. There's no chiller to run, no water to consume, no heat exchanger to maintain. (source)
• NIMBY: There's a growing wave of communities pushing back against data centers near their homes, citing noise, energy consumption, and environmental impact. In orbit, that's not a conversation anyone is having.

25% cheaper AI compute, with no energy bills, no cooling costs, and no zoning board. The maintenance problem is real: there's no patching a server rack via truck roll when it's at LEO altitude. But SpaceX is the only company with a rocket cheap enough to make the math work even with maintenance factored in.

So, should you buy it?

I asked Bridget the most obvious question.

That's an honest answer. Not a cop-out. Bridget can't score a stock with no price history. That's a feature, not a limitation. What she can do is pull the S-1, surface the financials, compare the sentiment data on public proxies, and give you everything you need to form your own view. Which, for a $1.75 trillion decision, is probably more useful than a hot take anyway.

The bottom line

Starlink reaching 10 million subscribers faster than T-Mobile or Comcast reached theirs in their early years isn't a quirk of the data. It's what happens when you solve a problem that nobody else could solve, not because they weren't smart enough, but because the physics of ground-based infrastructure made it uneconomical. Starlink doesn't have that constraint.

The $1.75T valuation is a hinges on the AI segment delivering. That part is genuinely speculative. But the connectivity story (the one I watched play out across 4,000 miles of North American wilderness) is real, it's growing, and the traditional models for measuring its TAM are almost certainly understating it.

Bridget can't tell you whether to buy SPCX at $135. What she can do is read the S-1, pull the sentiment data on every public proxy, surface the Anthropic deal terms, and help you understand what you're actually investing in. That's a different kind of research than what most retail investors have access to. It's also the right kind for a company this complex.

Trading opens June 12. Do your homework.

This analysis is a glimpse into what's possible when institutional-grade filings intelligence meets real-time social sentiment data.

Learn more about how Context Analytics powers next-generation investment research and AI platforms at www.contextanalytics-ai.com.

This post is for informational purposes only and does not constitute investment advice. Context Analytics and Bridget do not issue buy/sell recommendations on securities without established price history. All S-1 data sourced via the Elastic Filings MCP from SEC EDGAR. Sentiment data from Context Analytics S-Factor feed, June 8, 2026.