Why did revenue from a major customer change near-term expectations?

  • Based on disclosed information, Deutsche Bank estimates that major customer B accounted for 34.5% of SpaceX’s revenue in the third quarter of 2026, equivalent to 1.68 billion dollars per month, above the bank’s original model based on a contract value of 1.25 billion dollars per month.
  • Deutsche Bank had expected an expanded renewal only in the first quarter of 2027. It now speculates that the renewal may have come early, that SpaceX sold the customer more computing capacity, and/or that it obtained a price above the original 31 dollars per watt. It remains unclear whether the higher rate covered the entire quarter or began only in September.
  • On that basis, the bank raised its fourth-quarter revenue forecast by approximately 1 billion dollars to 23.6 billion dollars, above the market consensus of 22.4 billion dollars, and believes expectations for the next few quarters may still be too low.

What is the rationale for low-cost expansion, and how difficult is the timetable?

  • Management aims to reach nearly 3 gigawatts by the end of 2026, then add another 10 gigawatts. Deutsche Bank believes a total of 13 gigawatts is possible within two to three years, but questions the path to reaching that scale by the end of 2027.
  • Deutsche Bank suggests that vertical integration can reduce construction costs and shorten schedules, for example through direct control of projects, conversion of existing buildings, modular construction and the use of battery storage to optimize backup power arrangements. Savings from converting older buildings may not be consistently replicable.
  • The energy strategy also involves trade-offs: simple-cycle gas turbines have lower upfront costs and can be deployed faster, but are less efficient over the long term. Deutsche Bank suggests that new data centers may use gas turbines alongside energy storage to generate their own power behind the meter.

How much funding does rapid expansion require?

  • In Deutsche Bank’s scenario, 10 gigawatts of computing capacity are added in 2027 at a cost of 35 billion dollars per gigawatt. Together with 25 billion dollars in spending on space and connectivity businesses, this brings total capital expenditure to 375 billion dollars.
  • Under the bank’s scenario, factoring in approximately 70 billion dollars in cash, approximately 150 billion dollars in operating cash flow the following year and a minimum cash balance of 20 billion dollars, the funding gap is approximately 175 billion dollars. Deutsche Bank expects the gap to be filled mainly through debt, including GPU-backed financing.
  • The bank estimates that, at current spot prices, the payback period for investment in computing capacity is 8 to 9 months, making rapid construction financially attractive. It also expects spot prices could normalize at the end of 2027 or in 2028; if prices remain firm, it calculates that the new cloud business’s annualized exit revenue could exceed 200 billion dollars by the end of 2027.

How does more computing capacity translate into model and valuation growth?

  • SpaceX plans to build a training cluster of 440 thousand next-generation Nvidia Vera Rubin GPUs by the end of 2027. Deutsche Bank estimates its power consumption at approximately 1.3 to 1.4 gigawatts. The bank believes Grok’s rate of improvement could accelerate if investment in training compute increases and scaling laws continue to hold to some extent.
  • Grok Bot reached 800 thousand paid weekly active users earlier this month, more than doubling from the previous month; deployment of the enterprise version has just begun. Deutsche Bank believes this could drive significant growth in Grok inference annual recurring revenue outside the new cloud business in the second half of 2027.
  • Deutsche Bank maintained its Buy rating and raised its price target from 235 dollars to 250 dollars. In its sum-of-the-parts valuation, the AI business is valued at approximately 6 times projected 2034 revenue, versus approximately 4.5 times previously. The report also notes that Starship execution, reliance on key individuals and regulatory approvals will all affect the deployment and growth of long-term projects.