SpaceX Google deal: $30bn AI cash-flow test for IPO

SpaceX handed investors a cleaner number to argue over on Thursday: a $30 billion agreement to sell Google computing capacity before the rocket-and-satellite company tries to float at a $1.75 trillion valuation.

Formally, the SpaceX filing with the US Securities and Exchange Commission is large enough to move the IPO conversation out of pure Musk-premium territory. Google is slated to pay $920 million a month from October 2026 through June 2029 for access to about 110,000 Nvidia GPUs, with a lower fee during the ramp-up period. For a company being marketed as a rockets, Starlink and AI-infrastructure conglomerate, the contract changes the pitch from compute may matter to here is a cloud-style revenue line.

The first hard question comes from the same document. Escape hatches include termination rights after Dec. 31, 2026, and a one-month grace period if SpaceX fails to deliver the required GPU capacity. Investors can treat the deal as a stronger IPO story, but not yet as settled evidence of durable revenue.

The odd thing is that either Musk is correct and the S-1 is materially misleading, or the S-1 is correct and Elon is up to his old hijinx.
Source: Eric Talley, Columbia Law School, quoted by CNBC

That governance risk matters because demand is already running hot. Reuters reported that the IPO book was roughly two times oversubscribed, with about $150 billion of demand against the target raise. In that setting, the Google contract is more than another customer win. It gives bankers a fresh way to defend a valuation that would put SpaceX among the largest listed companies in the US on its first day of trading.

What the filing changed

Until the Google agreement, the SpaceX roadshow had a familiar shape: enormous ambition, scarce public financial history and a valuation target that forced investors to price several businesses at once. This contract narrows one part of that exercise. It translates the AI-infrastructure story into a scheduled payment stream, tied to a customer that is both a long-time investor and one of the largest buyers of compute in the world.

Bankers can carry the arithmetic through an IPO model without much translation. At $920 million a month, the contract points to about $11 billion of annualized revenue once fully in effect. The headline value of about $30 billion through mid-2029, reported by Bloomberg, gives them a tangible figure to place beside Starlink growth, launch cadence and xAI-adjacent data-center optionality.

Valuing SpaceX is still hard. Morningstar analysts, cited by CNBC, have argued that SpaceX is worth less than half of its $1.75 trillion target, partly because the range of outcomes around xAI and related compute assets is wide. A signed Google contract narrows the left tail for that business line, but it does not erase the question of how much multiple investors should pay for revenue that depends on chip delivery, power availability and customer urgency.

Customer identity matters here. Google is not buying a speculative AI story from the outside. It is paying for bridge capacity to support Gemini Enterprise, its agent platform. A Google spokesperson told TechCrunch the agreement was short-term and timely, meant to meet surging customer demand. That wording frames SpaceX as a supplier of scarce infrastructure during a bottleneck, not merely as a Musk-controlled company bundling narratives ahead of an offering.

This is a short-term, timely agreement to ensure we have bridge capacity to meet surging customer demand for our agent platform, Gemini Enterprise.
Source: Google spokesperson, TechCrunch

Public investors know how to pay for scarcity. They are less generous when scarcity turns into commodity supply. The contract gives SpaceX a valuation bridge only if the company can show that the bridge becomes a repeatable compute business rather than a one-off emergency lease.

The cancellation clause

Fine print drives the skeptic’s case. The SEC filing says SpaceX must make the full GPU capacity available by Sept. 30, 2026, after which the $920 million monthly payments begin. If SpaceX misses that delivery date, Google gets a one-month grace period before it can walk away. After Dec. 31, 2026, either side can terminate with 90 days’ notice.

Those are not minor details for an IPO buyer. A recurring revenue line with three years of locked-in payments deserves one multiple. A bridge-capacity lease with mutual cancellation rights deserves another. Musk sharpened that distinction when he said in a separate dispute over SpaceX’s Anthropic compute arrangement that the company had not committed to leasing Colossus for years, describing it as a 180-day lease with 90-day notice thereafter.

Formal Google filings should carry more weight than a social-media post about a different customer. The pattern is useful anyway. SpaceX is asking investors to value AI infrastructure as a durable business while its own disclosures show flexible commercial terms. Customers buying frontier compute may not want to lock themselves in when model architectures, chip supply and in-house capacity are changing quickly. Flexibility cuts both ways, and IPO investors usually pay less for revenue that can leave quickly.

Here the analyst and skeptic perspectives meet. Morningstar’s valuation concern is not that SpaceX has no growth options. Too many of them require generous assumptions at once. The Google contract improves one assumption: near-term demand for SpaceX-controlled compute. It weakens another if investors had expected multi-year, take-or-pay cloud revenue with limited cancellation risk.

Why Google is paying up

Google’s side of the transaction may be more revealing for the broader market. Alphabet has been trying to reassure investors that its AI spending can defend search, build a credible agent business and generate enough revenue to justify the data-center bill. CNBC reported this week that Alphabet expected capital expenditure to reach as much as $190 billion this year, double last year’s spending, while the company explored fresh financing for the buildout.

Set against that backdrop, a $920 million monthly lease reads as urgency. Google is not short of engineering talent or balance-sheet capacity. It is short of immediately usable high-end compute at the moment when agent products are moving from demos into enterprise sales pitches. If it is willing to rent capacity from SpaceX at that scale, Nvidia demand is no longer just a hyperscaler procurement story. It is becoming an inter-company cash-flow market, where those with clusters can sell time to those whose customer demand is outrunning internal supply.

SpaceX can monetize chips before the market has a clean way to value the asset base. Google gets capacity without waiting for every data center, power interconnect and GPU deployment to arrive on its own schedule. Nvidia benefits indirectly if the deal reinforces the scarcity premium around advanced GPUs, even if the transaction itself reallocates capacity that has already been bought.

That is the AI capex trade in miniature. For two years, investors have asked whether the big cloud companies are overbuilding. This deal shows the opposite risk for at least one buyer: not enough capacity, soon enough, to serve enterprise demand. The catch is that a bridge lease also implies the shortage may be temporary. If Google closes its own capacity gap, SpaceX needs either renewed demand, another customer or a lower valuation multiple for the same GPU estate.

The IPO read-through

SpaceX can now tell investors that AI infrastructure reaches beyond an internal Musk-ecosystem transfer. Google is an external customer, and the monthly payment number is too large to dismiss. Heavy oversubscription, as reported by Reuters, gives the company room to defend its offering price and market the IPO as a rare chance to buy into the Musk empire beyond Tesla.

None of that ends the valuation argument. MarketWatch has warned that the proposed valuation would require exceptional performance just to keep pace with broad equity benchmarks, and Morningstar’s published skepticism gives institutional buyers cover to push back. The Google contract makes the bull case cleaner, but it also makes the diligence list sharper: delivery dates, GPU sourcing, power costs, customer concentration, related-party exposure and how management describes termination rights.

Disclosure remains a separate issue. SpaceX is entering the market as a company with several narratives packed into one stock: launch monopoly, Starlink cash generation, AI compute, xAI optionality and the Musk control premium. When public comments appear to compress or qualify formal filings, the risk is not only legal. It is valuation noise. Investors need to know whether they are buying contracted cash flow, strategic optionality or both.

The bullish reading is straightforward. SpaceX has found a way to turn its data-center bet into pre-IPO proof. Google needed compute. SpaceX had capacity or a path to provide it. The result could add billions of dollars of annual revenue and make the listing look less like a leap of faith.

Caution points to a narrower conclusion. SpaceX has won a valuable, time-sensitive lease in a market distorted by AI scarcity. That is still a major deal. It is not the same as proving that AI infrastructure deserves to sit beside rockets and satellites as a stable public-company pillar.

The tension is the point. The Google contract gives SpaceX the cash-flow evidence investors wanted, then asks them to decide how much of that evidence survives when the scarcity passes.