SpaceX, the Force-Fed Valuation, and the AI-Lab Echo: A Skeptic's Field Guide

The short version

The “force-fed” thesis is largely correct in mechanism but overtaken by events on price. SpaceX rode a private-market valuation escalator (~$210B mid-2024 → ~$350B Dec 2024 → ~$400B July 2025 → ~$800B Dec 2025) set by twice-yearly insider tender offers, absorbed xAI in February 2026, and IPO’d on June 12, 2026 at $135/share (~$1.77T) — the largest IPO in history — popping to a ~$2T+ market cap on a thin ~4–5% float before falling ~20% once options/short-selling arrived. Public price discovery has now begun, and the most disciplined independent valuations (Damodaran $1.25–1.35T; Morningstar ~$780B / $63 a share) sit 25–55% below the IPO price. The market was force-fed; the fundamentals do not support the mark.
Strip away the narrative and SpaceX is two good businesses (a launch monopoly + Starlink) wrapped around a cash-incinerating AI bet. 2025 consolidated revenue was ~$18.7B (but only because xAI/X were retroactively consolidated), with a ~$4.9B net loss. Starlink is the only profitable segment (~$11.4B revenue, ~$4.4B operating profit). Launch is a genuine monopoly but small (~$4.1B). xAI lost ~$6.4B in 2025 and consumed ~$12.7B of capex. At ~$1.7T the stock trades at ~90–112x trailing sales — versus ~1.8x sales / ~16x EBITDA for Lockheed Martin. That gap is the “hopes and dreams” premium, quantified.
OpenAI and Anthropic are running the identical playbook and are MORE exposed to the critique, not less. Both are marked at AGI-optionality multiples (OpenAI $852B on ~$25B ARR; Anthropic $965B post-money on ~$47B run-rate) set by a tight circle of repeat backers (SoftBank, Nvidia, Amazon, a16z, GIC, Coatue) whose “circular” compute-for-equity deals inflate apparent demand. SpaceX at least owns hard assets, a launch monopoly, and a cash-gushing Starlink; the AI labs have explosive revenue growth but massive losses, unproven unit economics, and a financing structure resembling a closed loop. Honest verdict: SpaceX is overpriced but real; the AI labs are the purer expression of the force-fed phenomenon.

What stands out

SpaceX valuation trajectory — the escalator, then the launch

Primary rounds (small, infrequent): Series N raised $1.9B in 2020 at $46B; $850M in 2021 at $74B; $750M in January 2023 at $137B. SpaceX raised only ~$10–12B of primary equity in its entire life — striking for a company that briefly touched a ~$2T market cap.
Tender-offer escalator (the real price-setting mechanism): June 2024 ~$210B ($112/share); December 2024 ~$350B ($185/share; $1.25B tender of which SpaceX bought back $500M); July 2025 ~$400B ($212/share); December 2025 ~$800B ($421/share; ~$2.56B insider sale). The valuation roughly 4x’d in ~18 months on transactions that cashed out small slices of insider stock.
xAI merger (Feb 2, 2026): All-stock; combined entity valued ~$1.25T (SpaceX ~$1T, xAI ~$250B) — the largest private merger ever. This bolted an AI-narrative premium and a cash-burning business onto the rocket company right before the IPO.
IPO (June 12, 2026): 555.6M Class A shares at $135 = ~$75B raised (largest IPO ever), ~$1.77T valuation. Stock opened ~$150, ran to an all-time high of $225.64 on June 16, then fell ~20% over two sessions to ~$175–192 by June 18 once SPCX options began trading June 17 and gave bears their first practical tool. Only ~4–5% of shares were floated; ~95% locked up (first windows late July/Aug, standard lock-up ~Dec 2026, Musk’s stake to June 2027).

The fundamentals (from the S-1)

Consolidated revenue: 2023 ~$10.4B → 2024 ~$14.0B → 2025 ~$18.7B (+33%); Q1 2026 ~$4.7B (+15%). Critical caveat: the 33% growth is partly an artifact of common-control accounting that retroactively consolidates xAI and X into all prior periods.
Segment revenue 2025: Connectivity/Starlink ~$11.4B (+~50%, 61% of revenue); Space/launch ~$4.1B (+~8%); AI/xAI ~$3.2B (+22%).
Profitability: Consolidated net loss ~$4.9B in 2025 (vs. a ~$0.8B net profit in 2024 before xAI consolidation); adjusted EBITDA ~$6.6B. Starlink operating profit ~$4.4B (only profitable segment). Space roughly breakeven (Starship R&D ~$3B). AI operating loss ~$6.4B. Q1 2026 net loss ~$4.28B.
Capex: ~$20.7B in 2025 (of which ~$12.7B was AI infrastructure); Q1 2026 AI capex alone ~$7.7B (a ~$30B annualized run-rate). Free cash flow deeply negative.
Starlink operating metrics: Subscribers 2.3M (2023) → 4.4M (2024) → 8.9M (2025) → 10.3M (Q1 2026). But monthly ARPU fell from ~$99 (2023) → ~$91 (2024) → ~$81 (2025) → ~$66 (Q1 2026) as growth shifted to price-sensitive markets. Starlink gross margin improved 37% → 48% (2024→2025).
TAM claim: The prospectus headlines a $28.5T total addressable market, of which $26.5T is AI (including $22.7T of “enterprise applications”). Damodaran: “SpaceX has the largest TAM of any company in history… This estimate borders on fantasy.”
Governance: Dual-class. Musk holds all Class B super-voting shares and controls ~85% of votes (as-filed) / ~82% post-offering. His 2026 comp package (1 billion performance shares) vests partly on “a permanent human colony on Mars with at least one million inhabitants.”

What the multiple implies vs. comparables

At ~$1.7–1.8T on ~$18.7B trailing revenue, SpaceX trades at ~90–112x sales — among the highest for any large-cap in US history; MarketWatch noted it leaves “virtually zero room for error.”
Defense/aerospace comps trade at a fraction: Lockheed Martin ~1.8x EV/revenue, ~16x EV/EBITDA, ~17x forward P/E; Northrop Grumman ~20x forward P/E, ~15x EV/EBITDA; RTX ~17x EV/EBITDA. SpaceX trades at roughly 50x the revenue multiple of the defense primes.
Satellite-comms comps (Viasat, EchoStar, Eutelsat/OneWeb) trade at low-single-digit revenue multiples and ~20% EBITDA margins — far below Starlink’s reported ~50–63% EBITDA margin, which is the bull’s strongest card.

The independent valuations — the reality check

Aswath Damodaran (NYU “Dean of Valuation”): DCF enterprise value ~$1.22T, equity $1.3T (~$100/share). Verbatim: “At the rumored pricing of $1.8 trillion… it is too richly priced for my tastes, given my valuation of $1.25–$1.35 trillion for the equity.” His model already bakes in generous assumptions: $320B blended long-run revenue, ~50% blended operating margin, an 8.37% cost of capital, and a doubling of his AI revenue target to $160B. He still lands ~28% below the IPO price.
Morningstar: ~$780B / $63 a share — “overvalued,” less than half the IPO price.
The gap between the ~$1.7T mark and these ~$0.78–1.3T fundamentals-based estimates IS the force-fed/narrative premium. New Market Pitch’s anchor analysis pegs the “Musk effect”/narrative premium at roughly $450B–$970B (≈25–55% of the IPO valuation), midpoint ~$700B.

The fuller picture

Part 1 — The hard valuation: a monopoly, a utility, and a money pit

SpaceX is genuinely three businesses, and the first two are excellent. Launch is a near-monopoly: SpaceX flew the overwhelming majority of the world’s orbital mass in 2025, with reusable Falcon 9 economics no competitor approaches (NewStreet estimates “at least a 10-year lead”). But it is small (~$4.1B revenue, +8%), and ~66–75% of flights are internal Starlink missions that book no external revenue. Starlink is the actual value engine: a SaaS-like, recurring, high-margin utility that doubled subscribers to 10.3M and turned free-cash-flow positive in 2024. It generated ~$11.4B revenue and ~$4.4B operating profit in 2025 at an EBITDA margin reported in the 50–63% range — multiples of the ~20% margins at legacy satellite operators.

The problem is twofold. First, ARPU is falling fast — from ~$99/month (2023) to ~$66 (Q1 2026) — because incremental subscribers come from price-sensitive markets (Africa, SE Asia, LatAm). Growth is real but the unit economics are diluting, and Starlink’s revenue growth decelerated from ~95% (2024) to ~27–50% (2025, depending on source). Second, competition is finally arriving: Amazon’s Leo (formerly Kuiper) targets a mid-2026 consumer launch with claimed 1 Gbps speeds and AWS integration; Eutelsat/OneWeb revenues were up ~60% in H1; Telesat Lightspeed and others are scaling. Starlink’s lead is large but the LEO broadband market is now, in Quilty Space’s words, “a legitimate race.”

Then there is xAI — the swing factor and the skeptic’s smoking gun. Bolting xAI on in February 2026 converted a barely-profitable hardware company into a money-losing AI conglomerate. xAI lost ~$6.4B operating in 2025 on ~$3.2B revenue and consumed ~$12.7B of capex; in Q1 2026 it burned ~$7.7B of capex in a single quarter. The Colossus compute center is leased to Anthropic for ~$1.25B/month through May 2029 — an ironic arrangement where Musk’s AI lab keeps the lights on by renting GPUs to a rival. Damodaran’s central worry is precisely this: that SpaceX “will overreach in the AI business,” with Musk’s voting lock-in leaving shareholders unable to restrain capex even “after it becomes clear that the AI market is much smaller than anticipated.”

The bear case, consolidated:

Valuation gap: ~$1.7T mark vs. ~$0.78–1.3T defensible DCF. The market is paying 30–120%+ over fundamentals.
Government/customer concentration: ~30–40% of revenue from NASA/DoD/Space Force; lifetime federal commitments ~$22B. The June 2025 Trump–Musk feud (Trump threatening to cancel contracts; Musk threatening to decommission Dragon) showed the political tail risk is live, not theoretical.
Key-man + governance risk: Musk controls ~82–85% of votes; all 11 of xAI’s original co-founders departed before the IPO; the comp package ties vesting to a Mars colony. There is no governance brake.
Capex intensity & FCF: Deeply negative free cash flow; the constellation needs perpetual replenishment and the AI buildout is open-ended.
Float/lock-up overhang: ~95% of stock unlocks in stages through 2027; the scarcity premium that drove the pop can reverse.
No S&P 500 inclusion until four consecutive GAAP-profitable quarters (not expected before late 2027), limiting forced passive demand beyond the early-July Nasdaq-100 add.

The genuine bull case (intellectual honesty requires stating it):

The launch monopoly and Starlink are real, defensible, high-margin businesses with a decade-plus moat. Damodaran — no hype man — still lands at $1.3T, within shouting distance of $1.7T if you believe the AI optionality.
Starlink direct-to-cell (via the ~$17B EchoStar spectrum acquisition) and Starshield/defense revenue are credible, large, under-monetized adjacencies.
The Tesla precedent: Musk-led companies have repeatedly made “absurd” valuations look cheap in hindsight. Shorting the narrative has bankrupted smarter people than the average bear.
The uncertainty is arguably skewed to the upside (Damodaran’s own framing): more ways to surprise high than low if Starship’s sub-$100/kg economics unlock genuinely new markets.

Part 2 — The force-fed mechanism: how a few trades became a “market cap”

The mechanics matter more than the moralizing. Private megacap “valuations” are not market-clearing prices; they are marks set by a tiny number of insider transactions and then reported by the press as if they were market caps.

Tender offers as price-setting: SpaceX ran these twice a year, cashing out a small slice of employee/early-investor stock at a price negotiated between the company and a hand-picked set of “authorized buyers” — historically a tight Musk-orbit circle (a16z, Founders Fund, Gigafund, Sequoia). Musk himself noted in December 2024 that “almost no investors wanted to sell shares” at $350B — i.e., the mark was set on low volume by sellers who didn’t want to sell. The jumps (67% in six months; a doubling July→Dec 2025) were explicitly easier “because it’s easier to place small amounts of stock at aggressive valuations than large amounts.”
Reflexivity/circularity of marks: The same repeat funds appear round after round, and each markup increases the carrying value of their existing stake — a self-reinforcing loop where the markers benefit from the marks. The $1.5T IPO target itself was used to make the $800B December round “more palatable,” which in turn became the springboard for the IPO.
Secondary markets & SPVs: Forge Global, EquityZen, and Hiive let accredited investors buy slices, often through stacked SPVs with layered fees and contractual restrictions (ROFRs). Forge’s own “Forge Price” for SpaceX hit ~$1.03T by April 2026. These platforms create the appearance of liquid, tradable, priced assets while real price discovery (short-selling, mark-to-market, quarterly earnings) is absent. Forge was acquired by Schwab for $660M — down from a $2B SPAC price — a tell on how much of the “private-markets-democratization” story was itself hype.
Retail amplification & the premium-to-NAV absurdity: The closed-end fund Destiny Tech100 (DXYZ) is the purest illustration. Per Acadian Asset Management’s analysis, DXYZ listed on the NYSE March 26, 2024 at a $4.84 NAV and rose to $105 by April 8, 2024 — “a price that was 21X too high; the premium to NAV was more than 2,000%,” the highest CEF premium ever observed (beating the prior 1,235% record of Capital Administration Co., per DeLong & Shleifer 1991, as cited in DXYZ’s own SEC Form 497AD). Per Destiny’s Dec 31, 2025 results, NAV reached $19.97/share, “up 210% from $6.44… at the end of the fourth quarter of 2024,” on a ~$434.0M portfolio in which SpaceX was 16.2%. Morningstar called it a mechanism for retail to “enrich others at the expense of themselves.”
Why private marks persist where public scrutiny would not: No short-sellers, no daily mark-to-market, no quarterly earnings, illiquidity, and a powerful selection effect in who gets to transact (only insiders sell, only approved buyers buy). The result is a valuation that floats free of the disciplining forces of a public market — exactly the “force-feeding” the thesis describes. The SpaceX IPO is the natural experiment: within days of public trading and the introduction of options (the first short-selling tool), the stock fell 20% and Future Fund’s Gary Black noted it had “traded more like a meme stock than one driven by fundamentals.”

This is straight Damodaran “narrative vs. numbers”: for young companies, value is driven by TAM and unit-economics stories, not bottom lines, and the SpaceX prospectus is “a soaring story, but with weak links and multiple distractions.” The force-feeding is the propagation of the story-derived mark as if it were a fact.

Part 3 — Macro: an asset-inflation / liquidity phenomenon, late cycle

The private-megacap phenomenon is downstream of a decade-plus of cheap money. Combined PE/VC AUM went from ~$600B in 2000 to >$10T today; the 2020–21 ZIRP blow-off deployed capital at ~2x the long-run average at peak valuations. Crucially, the best companies now stay private far longer, capturing in private markets the appreciation that used to accrue to public investors — which is why a SpaceX or OpenAI can reach hundreds of billions before any public scrutiny.

The current cycle has a specific accelerant: the AI capex super-cycle. The five largest US tech spenders are guiding to ~$635–690B of 2026 capex (≈$2.1T across 2026–28), driving capex intensity to ~34% of revenue — more than double the late-1990s internet-buildout peak — and pushing aggregate free cash flow negative for the first time in ~35 years. AI investment is ~1–1.5% of GDP, comparable to past infrastructure waves but with shorter asset lives (GPUs depreciate fast).

Where are we (mid-2026)? Late, and showing classic top-signals, but not obviously broken:

Bear signals: index concentration into a handful of mega-caps; AI capex outrunning AI revenue with no measurable GDP impact yet; the Fed listing AI as a top systemic risk; circular financing inflating apparent demand; GMO/Hussman-style models projecting near-zero-to-negative real returns for US large-caps over 7–12 years; the Nasdaq shedding >7% from a June 1 all-time high right as SpaceX listed.
Bull/stabilizer signals: unlike 2000, the spenders are the most profitable companies in history, largely funding capex from cash flow (Russell 3000 capex/FCF below 1x vs. ~4x in 2000); chips sold out 18–24 months forward (real pre-paid demand); strong S&P earnings breadth.

Honest read: a liquidity-and-narrative-driven late-cycle environment in which a small number of “winner” stories absorb a disproportionate share of capital — a textbook setup for the force-fed dynamic — but one where the leaders have real cash flows, so a financing-stop bust (the dot-com pattern) is less likely than a grinding de-rating. The marginal risk is exactly that “financing stop”: if debt-funded capex (Morgan Stanley sees >$400B hyperscaler debt issuance) meets disappointing AI revenue, the reflexive loop reverses.

Part 4 — OpenAI and Anthropic: same playbook, weaker assets

OpenAI: $157B (late 2024) → $300B (March 2025 SoftBank-led $40B round) → $500B (Oct 2025 secondary) → $852B (late March 2026, after a record $122B raise at $730B pre-money anchored by SoftBank, Nvidia, Amazon). Revenue ~$13B (2025) to ~$25B annualized (Feb 2026); the $500B mark implied ~167x 2025 revenue. But it loses prodigiously: ~$8–8.5B cash burn in 2025, projected ~$27B in 2026 and ~$63B in 2027, not cash-flow positive before 2030. It carries >$1T in disclosed compute commitments (Stargate $500B, Oracle >$300B, Azure $250B, AWS ~$138B). Secondary marks reportedly reached ~$880B.

Anthropic: $61.5B (March 2025) → $183B (Sept 2025 Series F) → $380B (Feb 2026 Series G) → $965B post-money (Series H, “$65 billion… led by Altimeter Capital, Dragoneer, Greenoaks, and Sequoia Capital”). Run-rate revenue went from ~$1B (early 2025) → $9B (end-2025) → ~$14B (Feb 2026) → “crossed $47 billion earlier this month” (the Series H announcement) — among the fastest revenue ramps in software history. Backed by Amazon (~$8B+) and Google, with the now-familiar compute-for-equity structure: Anthropic committed >$100B to AWS and $30B to Azure while Amazon/Nvidia/Microsoft invest in Anthropic.

The circular-financing concern is the AI-specific version of the force-feeding mechanism. Nvidia invests up to $100B in OpenAI → OpenAI commits to buy Nvidia GPUs → routed through Oracle/CoreWeave who buy more Nvidia chips. Microsoft/Nvidia invest ~$15B in Anthropic → Anthropic spends ~$30B on their compute. The same dollar can appear as Nvidia revenue, OpenAI funding, and Oracle backlog simultaneously. As Azeem Azhar put it, a “financial ouroboros”: revenue that is “really just the same stock of money going in circles.” The February 2026 stall of the Nvidia-OpenAI deal triggered a multi-company selloff precisely because the market glimpsed how reflexive the loop is. The only number that proves real demand is revenue from outside the circle.

SpaceX vs. the AI labs — who is more “force-fed”?

Dimension	SpaceX	OpenAI / Anthropic
Hard assets / moat	Launch monopoly, satellite constellation — real, defensible	Model weights + brand; moat eroding (DeepSeek, open-source, price wars)
Profitable core	Yes — Starlink ~$4.4B op profit	No — both burn billions
Revenue multiple	~90–112x sales	OpenAI ~30–167x; Anthropic ~20–40x
Cash flow	Deeply negative (driven by AI bet)	Deeply negative (structural)
Financing structure	Mostly equity + IPO cash	Heavily circular compute-for-equity
Narrative anchor	Mars / AGI / orbital data centers	AGI / “transformation of all knowledge work”

Verdict: the AI labs are MORE exposed to the force-fed critique, not less. SpaceX has a genuine, monopolistic, cash-generating core (Starlink + launch) onto which an expensive AI option has been grafted — its overvaluation is a premium-on-a-real-business problem. OpenAI and Anthropic are pure optionality plays: their entire valuations rest on capturing a multi-trillion-dollar AGI TAM that does not yet exist, financed by a circular structure that manufactures the appearance of demand. If the liquidity tide goes out, SpaceX has Starlink to fall back on; the labs have burn rates and compute commitments. The one caveat in the labs’ favor: their revenue is genuinely growing faster (Anthropic ~10x annually) than almost anything in history, so the “grow into the multiple” path is at least conceivable — whereas SpaceX’s launch and Starlink growth is decelerating and its AI growth is the slowest of the major labs.

What to watch

The picture that emerges, staged and conditional:

SPCX is a flow vehicle, not a value holding, through ~Q1 2027. The stock is governed by float mechanics (4–5% float), the early-July Nasdaq-100 add (~$8–10B passive demand, but modeled at only ~1.6–2.2% price impact), and a staged lock-up — tradable, knowable catalysts. Damodaran’s framing: this is a trader’s stock right now, where intrinsic value is “perhaps even irrelevant.”
The fundamental pressure window is the lock-up cascade, not now. ~95% of shares unlock in stages from late July 2026 through June 2027; the scarcity premium that drove $135→$225 can reverse hard as supply hits, with the December 2026 standard lock-up expiry the key date. The countervailing risk: this is a Musk vehicle prone to violent narrative-driven squeezes.
The independent anchors, not the tape, mark fair value. Damodaran’s $1.3T (~$100–105/share split-adjusted) and Morningstar’s ~$780B ($63) bracket the corridor — value below ~$100/share, rich above ~$180–200. Thresholds that would raise fair value: Starship V3 reaching operational sub-$100/kg economics; Starlink direct-to-cell monetizing the EchoStar spectrum; Starshield/defense scaling. Thresholds that would lower it: continued ARPU erosion, an AI-capex blowout with no enterprise traction, or a renewed government-contract political rupture.
For the AI labs, the signal is second-order, not in the private marks. The circular-financing loop means a financing-stop shows up first in the infrastructure layer (Nvidia, Oracle, CoreWeave debt/CDS spreads), not in illiquid OpenAI/Anthropic secondaries that cannot easily be exited. Rising infrastructure-provider CDS spreads are the early-warning system — the lenders, not just the deals.
The retail-trap vehicles are structurally rigged against the buyer. DXYZ and similar premium-to-NAV closed-end funds make the premium the product. The SPV/secondary route (Forge/Hiive) at least prices closer to mark, at the cost of illiquidity and layered fees.

What would change the thesis:

More bullish on SPCX if: four consecutive GAAP-profitable quarters arrive (S&P 500 inclusion catalyst, ~late 2027); Starship V3 hits operational reusability and sub-$100/kg; xAI capex shows enterprise revenue traction rather than pure burn.
More bearish if: ARPU keeps falling while subscriber growth decelerates; Amazon Leo takes meaningful share; the AI-capex super-cycle shows a financing stop (hyperscaler capex guidance cuts, rising data-center debt spreads); or a Musk–government rupture threatens contracts.

Caveats

The world moved faster than the “$350B” framing this thesis started from. As of mid-2026 the live facts are: the xAI merger (Feb 2026), a $1.77T IPO (June 12, 2026), and a ~$2T peak followed by a ~20% pullback. The “force-fed at $350B+” thesis is now best understood as describing the private escalator that preceded public price discovery — and that discovery is, so far, partially vindicating the skeptics (the stock fell once shorting became possible, and independent valuations cluster 25–55% below the IPO mark).
Many SpaceX/xAI/Starlink figures are estimates or freshly disclosed and not yet tested under sustained public scrutiny. Segment numbers come from the S-1; pre-IPO figures relied on Quilty, Payload, Sacra, and Novaspace estimates that varied materially (2024 total revenue was variously estimated $13.1B–$14.2B). The ~33% 2025 growth is inflated by retroactive common-control consolidation of xAI/X.
The bull case is not trivial. Damodaran — the most credible independent skeptic — still lands at $1.3T and explicitly would not short. Musk-company valuations have repeatedly defeated fundamentals-based bears (Tesla). “Overvalued” is a statement about price relative to a probability-weighted range, not a prediction of imminent decline.
Source quality varies. Tender/round figures are well-corroborated (Bloomberg, CNBC, Reuters). Some secondary-market and forward-revenue figures come from interested parties (brokers, pre-IPO platforms, company press releases) and should be discounted accordingly. Forward TAM claims ($28.5T) are explicitly aspirational per the prospectus’s own disclaimers and should never be treated as established fact.