Space Exploration Technologies Corporation (SpaceX) (SPCX) Report Interpretation

Overview

The report examines six recent SpaceX developments and links them to Bernstein's broader investment thesis: high launch cadence and full Starship reuse are needed to deploy orbital data centers at scale, while terrestrial compute, proprietary AI capabilities, semiconductor capacity and Starlink adoption provide additional growth paths. Bernstein remains positive but uses a slower launch trajectory than management.

Core views

Bernstein's central thesis is that SpaceX's value depends heavily on achieving a high Starship launch cadence, which in turn requires both more launchpads and full reusability. The announced Starbase Louisiana site on Pecan Island covers 125,000 acres and is intended to become a self-sustaining spaceport with propellant production, power generation, deep-water shipping, vehicle-processing facilities and an airport. SpaceX has committed at least $100 billion over time and at least 3,000 jobs. The site is expected eventually to have at least 10 launchpads and support more than 30 flights per day. The report's opening summary references first launches in 2028, while the detailed discussion says construction is expected to begin in 2027 and the first Starship launch is planned for 2029, underscoring that the schedule remains unsettled. Bernstein views Louisiana as important infrastructure for orbital data centers. The 10 planned pads would supplement five pads expected to be operational by the end of 2027—two in Texas and three in Florida. At three launches per pad per day, all 15 pads could support well over 15,000 launches annually. If five Louisiana pads were operating in 2031, the resulting 10-pad network could provide roughly 10,000 annual launches at that rate, broadly consistent with SpaceX's projection of launching about 50 GW of orbital data-center capacity in 2031. Bernstein nevertheless assumes the company misses its stated launch timeline by roughly one year and then grows more slowly over the long term. Management's target of one launch per day by the end of 2027 would require only slightly more than one weekly launch per pad across the five pads expected by then, but Bernstein still uses a more conservative trajectory because SpaceX has historically delivered major capabilities later than initially planned. The next technical milestones concern orbit, stage recovery and reuse. Bernstein expects Launch 14 within the next few weeks, but Elon Musk has said it will not attempt the previously hoped-for second-stage catch. Subject to FAA approval, Launch 14 is expected to reach orbit and deploy fully operational Starlink V3 satellites that can interconnect with the current V2 Mini constellation. A previous flight deployed 20 Starlink satellites below orbit to test connectivity before they re-entered. The catch may instead occur on Launch 15, with actual reuse targeted for early 2027. The FAA is still assessing safety and environmental conditions, although the report says the Department of Transportation's easing of environmental-assessment requirements could help approval. Bernstein also notes uncertainty about whether SpaceX changed its favorable view of the Launch 13 heat shield after recovering the vehicle. These steps matter because the investment case requires rapid, repeatable launches rather than isolated successful flights. SpaceX's planned foundry near Bastrop, Texas, is intended to make blades and vanes for industrial gas turbines and remove a potential bottleneck in behind-the-meter power for terrestrial compute. The report says the market is almost entirely controlled by Howmet and Precision Castparts, with Howmet dominant. SpaceX has targeted putting power in place for 20 GW of terrestrial compute capacity by the end of 2027, but Bernstein doubts that sufficient high-end blade capacity can be built quickly enough to affect the power buildout over the next 18 months. It expects the foundry to be an internal supplier, not a broad competitor in turbine casting and forging. If capacity is rushed, the report expects management to prioritize timing rather than economics, making it unlikely that the operation would match Howmet simultaneously on cost, scale and quality. In AI software, SpaceX has completed its acquisition of Cursor and continues to emphasize improvements in Grok 4.6. Bernstein had assumed that orbital data centers would primarily sell raw compute capacity; stronger proprietary intelligence could add another layer of revenue and margin. That opportunity is uncertain because OpenAI has said it intends to withdraw Cursor's access to its model, while Anthropic has said Cursor can continue using Claude, and Cursor and OpenAI remain in discussions. Bernstein therefore treats enhanced AI monetization as potential upside rather than an established forecast. The Terafab project adds a semiconductor-manufacturing component to SpaceX's compute ambitions. SpaceX, Tesla and the Texas governor announced on August 6 that the facility would be built in Grimes County. The ultimate aspiration is chip output sufficient for roughly 1 TW of compute demand, a level the report describes as significantly above current global supply. The initial phase will be much smaller but is still expected to involve about $16.8 billion of combined SpaceX and Tesla capital investment and at least 3,000 employees. The scale illustrates the ambition of the compute strategy while also reinforcing semiconductor availability and execution as constraints. Starlink's airline business is the clearest near-term evidence of commercial leadership discussed in the report. Bernstein counts 46 airlines signed with Starlink after Royal Air Maroc joined and Qatar Airways expanded its rollout. Starlink has more than 11,000 satellites in orbit, compared with 396 for Amazon's competing LEO constellation, which Bernstein says is not yet sufficient to operate the service. Delta and JetBlue are the only two cited airlines that selected Amazon's system rather than Starlink; Delta is using Viasat while awaiting that constellation. Bernstein expects Starlink's airline presence to continue expanding. As of August 31, 2026, SPCX closed at $143.69, against a $248 target and reported 73% upside. The table lists a $1,950.28 billion market capitalization, $1,889.63 billion enterprise value and a 52-week range of $225.64 to $104.83. One-month absolute performance was 32.6%, or 30.0 percentage points relative to the SPX. Bernstein's summary forecasts show revenue rising from $18,674 million in F25A to $46,402 million in F26E and $150,375 million in F27E, while adjusted EBITDA rises from $6,584 million to $21,317 million and $85,583 million. Adjusted EPS is shown at negative $0.20 for F26E and $1.30 for F27E, and EV/adjusted EBITDA declines from 287.0x in F25A to 88.6x in F26E and 22.1x in F27E. The $248 one-year target uses a sum-of-the-parts valuation based on 12-month-forward EV/EBITDA applied to 2031 EBITDA. Bernstein uses 2031 because orbital data centers are the cornerstone of its valuation and would not be represented adequately by applying a multiple to 2028 earnings, which would mainly capture launch, terrestrial compute and near-term Starlink. It calculates segment values in 2030 using 2031 estimates and discounts each segment back to mid-2027. The resulting valuation remains exposed to the explicit risks that Starship fails to scale or achieve full reuse, orbital data centers cannot be operationalized, semiconductor or other supply constraints emerge, launch incidents or regulation interrupt deployment, or compute demand proves weaker than expected.

Analysis framework

Bernstein starts with six discrete August developments, then connects each to the company's longer-term economic model. It translates launchpad counts and assumed launches per pad into potential annual launch capacity, compares that capacity with orbital data-center deployment needs, and deliberately models a slower path than management. It also assesses supplier concentration in turbine components, compares satellite-network scale in airline connectivity, and treats AI software and semiconductor manufacturing as possible extensions of compute monetization. Finally, it combines segment forecasts through a sum-of-the-parts EV/EBITDA framework and discounts the long-dated values back to the one-year target date.

Methodology notes

• Valuation methodsSOTP (Sum-of-the-Parts) Valuation

  Sum-of-the-parts valuation

  Bernstein values SpaceX's segments separately so that launch, connectivity and AI-related businesses can reflect different economics and maturity profiles, then combines the segment values.

• Valuation methodsEV/EBITDA valuation

  12-month-forward EV/EBITDA applied to 2031 EBITDA

  The report applies enterprise-value multiples to 2031 segment EBITDA because it believes an earlier year such as 2028 would omit the orbital data-center business, then discounts the resulting 2030 values back to mid-2027.

• Industry AnalysisSupply-demand framework

  Launch-capacity requirement versus orbital deployment demand

  Bernstein converts launchpad availability and launches per pad into annual capacity and compares that capacity with the launches required to deploy SpaceX's projected orbital data centers.

• Industry AnalysisIndustry Concentration Analysis

  Industrial gas-turbine blade supplier concentration

  The report evaluates the foundry decision against a market dominated by Howmet and Precision Castparts, using the small supplier base and technical difficulty of high-end blades to judge likely timing and competitiveness.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• SpaceX (US.SPCX)

  Primary covered company; Bernstein links its valuation to launch cadence, full Starship reuse, orbital and terrestrial compute, AI monetization and Starlink growth.

  Strengths

  Large planned launch infrastructure, historical ability to deliver major technical capabilities, leading Starlink airline adoption, more than 11,000 satellites in orbit and potential integration of launch, connectivity, compute and AI.

  Weaknesses

  Management timelines have historically slipped, full reuse has not yet been achieved, the Louisiana site requires substantial development, and the foundry may not affect the near-term power buildout.

  Comparison

  Starlink has 46 airline signings and more than 11,000 satellites, versus 396 satellites cited for Amazon's competing LEO constellation; the planned foundry would also enter a component market dominated by Howmet and Precision Castparts.

  Risks

  Failure to scale reusable launches, operationalize orbital data centers, secure semiconductors and other supplies, avoid launch or regulatory interruptions, or sustain expected compute demand.

Key data

• SPCX closing priceUSD 143.69Close on 31 Aug 2026
• Price targetUSD 248.00One-year target; 73% upside from the reported close
• Starbase Louisiana site125,000 acresPlanned self-sustaining Louisiana spaceport
• Louisiana investment and employmentAt least $100 billion and at least 3,000 jobsSpaceX commitment over time
• Louisiana launch capacityAt least 10 pads and more than 30 flights per dayEventual planned capacity
• Potential network launch rateWell over 15,000 launches per yearBernstein calculation for 15 pads operating at three launches per pad per day
• 2031 orbital data-center projectionRoughly 50 GWCompany projection associated with about 10,000 annual launches in Bernstein's capacity illustration
• Terrestrial compute target20 GW by end-2027Power capacity targeted by the company
• Terafab initial investmentApproximately $16.8 billionExpected combined SpaceX and Tesla capital investment, with at least 3,000 employees
• Terafab ultimate ambitionApproximately 1 TW per yearLong-term chip-production scale cited by the report
• Signed Starlink airlines46Bernstein count after Royal Air Maroc joined and Qatar Airways expanded
• Satellite comparisonStarlink more than 11,000; Amazon LEO 396Satellites in orbit cited by the report
• Revenue forecast$18,674m / $46,402m / $150,375mF25A / F26E / F27E
• Adjusted EBITDA forecast$6,584m / $21,317m / $85,583mF25A / F26E / F27E
• EV/adjusted EBITDA287.0x / 88.6x / 22.1xF25A / F26E / F27E

Impact & implications

Bernstein believes Louisiana materially expands the physical capacity required for SpaceX's orbital-compute strategy, while the next Starship flights will test whether the company can progress toward the full reuse needed to make that capacity economically useful. The foundry and Terafab projects show an effort to control power and chip bottlenecks, and Cursor and Grok could improve compute monetization beyond raw capacity. Starlink's airline adoption supports the connectivity business, but the valuation still rests on long-dated launch and orbital data-center assumptions rather than only current operations.

Risks

• Starship launch capacity may not scale dramatically enough because the required full reusability has not yet been achieved.
• SpaceX may be unable to operationalize orbital data centers at the scale assumed in the valuation.
• Semiconductor availability and other supply-chain constraints could delay deployment.
• Launch incidents or regulatory restrictions could interrupt the launch plan.
• Compute demand could prove lower than prevailing expectations.

What to watch

• FAA approval and whether Launch 14 reaches orbit and deploys operational Starlink V3 satellites.
• Whether Launch 15 attempts the second-stage catch and whether actual reuse begins in early 2027.
• The construction schedule, first-launch timing and launchpad ramp at Starbase Louisiana.
• Progress toward management's target of one launch per day by the end of 2027.
• Whether the Bastrop foundry can produce turbine blades within the next 18 months and support the 20 GW power target.
• The outcome of Cursor's model-access discussions with OpenAI and continued access to Anthropic's Claude.
• Terafab's initial buildout and the availability of semiconductors for SpaceX's compute plans.
• Further Starlink airline signings and the progress of competing connectivity constellations.