AI Compute Price Increases and Capacity Expansion Open Upside, Target Price Raised to USD 248

Overview

SpaceX reported second-quarter results after the close on August 4, 2026, with revenue, EPS, and segment operating profit all significantly better than market expectations. The report argues that more aggressive ground-based AI infrastructure buildout, the compute pricing premium validated by Google and Anthropic contracts, and continued growth in Starlink broadband together improve the long-term revenue and earnings outlook. Bernstein raised its 2031 revenue forecast from USD 554bn to USD 655bn, but higher capex, debt financing needs, and a decline in the valuation multiple for AI-related assets meant the target price was only modestly raised from USD 239 to USD 248.

Core views

First, the main upside driver for the AI business is not a higher launch frequency for orbital data centers, but a larger scale of ground-based compute buildout and higher per-watt pricing. Second, full Starship reuse remains the core pivot for long-term valuation; Launch 13 showed positive progress on the heat shield, but second-stage recovery, regulatory approval, and scaled launches still need to be validated. Third, Starlink consumer, enterprise, and government broadband remains strong, with user growth, stable ARPU, and margin improvement supporting cash generation. Fourth, the report remains cautious on building a proprietary ground-based mobile network, arguing that partnering with major U.S. mobile network operators or adopting an MVNO model is more feasible.

Analysis framework

The report combines actual second-quarter results, management capacity and pricing guidance, Google and Anthropic commercial contracts, Starship launch and reuse progress, and Starlink user and ARPU metrics to update forecasts for the Space, Connectivity, and AI businesses respectively; it then uses a sum-of-the-parts approach and a long-term EV/EBITDA framework to discount each segment’s future value back to the target valuation date.

Methodology notes

• Valuation methodologySum-of-the-parts valuation (SOTP)

  Value the Space, Connectivity, and AI businesses separately and then aggregate enterprise value.

  Each segment is assigned appropriate valuation parameters based on its long-term revenue, profitability, capital intensity, and risk, and segment values are then discounted to mid-2027 to derive a one-year target price.

• Relative valuationForward EV/EBITDA

  Estimate 2030 enterprise value based on 2031 EBITDA.

  The report argues that orbital data centers are the core of long-term value, and using only 2028 earnings would not reflect this business; therefore, it uses 2031 EBITDA and lowers the AI valuation multiple from 23x to 21x.

• Operating driver modelCompute capacity and per-watt pricing model

  AI revenue is mainly determined by available compute capacity and revenue per watt.

  The model raises both ground-based compute installed capacity and near-term per-watt pricing, while assuming compute prices gradually decline thereafter as supply catches up with demand.

Asset mapping & comparison

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

• SPCX.US

  Core covered stock; the report maintains an Outperform rating and raises the target price.

  Strengths

  AI compute demand and pricing are strong, Starlink broadband growth is steady, and major second-quarter financial metrics exceeded expectations.

  Weaknesses

  Near-term free cash flow is under pressure, with increasing reliance on high capex and external financing.

  Comparison

  The USD 248 target price implies potential upside of about 129% versus the closing price of USD 108.27.

  Risks

  Starship reuse failure, delays in orbital data center implementation, declines in compute demand or pricing, and supply chain and regulatory constraints.

• AI

  The main source of long-term revenue and valuation upside.

  Strengths

  Google and Anthropic contracts validate demand and premium pricing capability, and capacity expansion is rapid.

  Weaknesses

  Construction costs are extremely high, with the report assuming about USD 47bn of investment required for each additional 1GW of capacity.

  Comparison

  Bernstein lowered the AI valuation multiple from 23x to 21x, but upward revisions to capacity and pricing assumptions still increased the business value.

  Risks

  Compute supply-demand conditions loosening, declines in per-watt pricing, semiconductor shortages, rising financing costs, and construction progress falling short of expectations.

• Connectivity

  The only business segment currently generating cash, covering Starlink consumer, enterprise and government broadband, and mobile businesses.

  Strengths

  Users reached 12m, ARPU is stable, operating margin is 39%, and enterprise and government customer growth is strong.

  Weaknesses

  Building a proprietary ground-based mobile network requires high-density terminals and complex infrastructure, and the implementation path remains unclear.

  Comparison

  Enterprise and government businesses may contribute revenue close to or above the consumer business in the future, and typically have better economics and customer stickiness.

  Risks

  The mobile business lacks a clear implementation path, operator partnerships have yet to be reached, and V3 satellite deployment may be delayed.

• Space

  Supports Starlink and orbital AI infrastructure through Starship launch capability and full reuse.

  Strengths

  Launch 13 showed progress on the heat shield, and second-quarter launch volume, revenue, and losses were all better than expected.

  Weaknesses

  Second-stage reuse has not yet been fully validated, and the target of one launch per day is higher than Bernstein’s current forecast.

  Comparison

  The company plans to have five launch pads operational by the end of 2027 and achieve one Starship launch per day.

  Risks

  FAA approvals, launch accidents, heat shield or engine issues, production bottlenecks, and delays in reuse progress.

Key data

• Second-quarter revenueUSD 7.81bnAbove market expectations of USD 6.55bn.
• Second-quarter diluted EPSUSD -0.09Significantly better than market expectations of USD -0.24.
• Second-quarter segment operating profitUSD -143mBetter than market expectations of USD -1.73bn.
• AI business revenueUSD 2.56bnAbout 23% above market expectations, mainly driven by Google and Anthropic contracts.
• AI nominal compute capacity1.4GWExpected to exceed 2GW by the end of 2026.
• Ground-based AI capacity forecast8GW in 2027; 15.5GW in 2028; 18.6GW in 2029Bernstein significantly raised its ground-based compute buildout assumptions starting in 2027.
• 2027 blended compute pricing assumptionabout USD 25/WManagement believes pricing for some compute contracts may remain at USD 30–50/W.
• Starlink users12mIncreased by 1.7m sequentially in Q2, with ARPU stable at USD 66.
• Connectivity business revenue and marginUSD 4.29bn; 39%Revenue was above market expectations of USD 3.88bn, and margin was above the expected 37%.
• Space business revenueUSD 962mAbove market expectations of USD 874m; 38 spacecraft were launched in Q2.
• 2031 revenue forecastUSD 655bnPrior forecast was USD 554bn; the company moved forward its USD 1tn revenue target from 2031 to 2030.
• Target price and potential upsideUSD 248; 129%Prior target price was USD 239, with the reference closing price being USD 108.27 on August 5, 2026.

Impact & implications

If SpaceX can expand ground-based compute capacity as planned and continue selling compute at prices above historical assumptions, the AI business’s revenue and earnings leverage will increase significantly; Starlink’s stable subscriber growth, ARPU, and expansion among enterprise and government customers provide high-quality cash flow support. Long-term valuation remains highly dependent on full Starship reuse and the commercialization of orbital data centers. Because expansion requires massive capex and additional debt financing, even with higher operating forecasts, valuation improvement will be partially offset by the cost of capital, balance sheet pressure, and a decline in AI valuation multiples.

Risks

• Starship may fail to achieve full reuse or may be unable to substantially increase launch frequency as planned.
• The commercialization and operational progress of orbital data centers may fall short of expectations, weakening the basis for long-term valuation.
• AI compute demand may be lower than expected, or increased supply may cause per-watt pricing to decline faster.
• Ground-based AI expansion may result in excessive capex, debt financing needs, and balance sheet pressure.
• Semiconductor supply shortages or other supply chain bottlenecks may limit compute and satellite deployment.
• Launch accidents, FAA approval delays, or other regulatory restrictions may affect the pace of operations.
• Starlink’s mobile business faces high execution difficulty in building its own network and has not yet secured clear operator partnerships.
• The expiration of share lock-up periods may create short-term share price pressure.

What to watch

• Whether Launch 14 can secure FAA approval, deploy operational V3 Starlink satellites, and attempt second-stage capture.
• Starship heat shield inspection results, second-stage reuse progress, and the feasibility of the end-2027 target of one launch per day.
• Whether compute capacity can exceed 2GW by the end of 2026, and the pace of ground-based AI expansion from 2027 to 2029.
• Whether per-watt pricing in Google, Anthropic, and subsequent customer contracts can support the USD 30–50/W range.
• Whether AI capex, debt issuance, and changes in free cash flow match the capacity expansion plan.
• Starlink net user additions, USD 66 ARPU, enterprise and government revenue, and margin trends.
• The actual improvement in network capacity and service performance from V3 Starlink satellite deployment.
• Whether SpaceX reaches partnerships with major U.S. mobile network operators or forms an executable MVNO path.