How Fast Are Space Launch Costs Falling?

Key Takeaways

• Average launch cost fell from $87,023/kg in 1960 to $3,868/kg in 2025 in the study.
• Each doubling of cumulative payload is associated with about a 21.2% fall in average cost.
• Lower costs remain conditional on scale, competition, reuse, regulation, and orbital sustainability.

What the New Data Says About Space Launch Costs

In 1960, the average cost of sending one kilogram to low Earth orbit was $87,023 in 2024 dollars. By 2025, Alessio Terzi and Francesco Nicoli estimate that figure had fallen to $3,868, a decline of more than twentyfold. Their July 2026 PNAS Nexus study, From Sputnik to Starship: Estimating the Experience Curve of Space Launch Technology, assembles 4,405 launches covering more than 330 rocket configurations from 1960 through 2025. The dataset spans 16 spacefaring geographical entities and standardizes missions to a common low Earth orbit equivalent so different rockets and eras can be compared on the same basis.

That long time horizon matters because space launch costs did not fall smoothly. The study finds a sharp decline during the Moon-race years, followed by a period in which average cost rose somewhat after Saturn V disappeared and then stayed broadly stable until renewed declines began around the early 1990s. The study’s launch-level data show another change that can be missed when attention stays on headline averages: launch-cost dispersion narrowed over time. Early launches were scattered across a very broad cost range, whereas later launches cluster much more tightly near the lower end.

The metric also needs careful interpretation. The authors use a standardized estimate of unit flyaway cost, which includes manufacturing, recurring engineering, sustaining tooling, quality control, overhead, and related recurring expenses, but excludes research and development spent to create the vehicle design. They convert all values to 2024 U.S. dollars and normalize performance to low Earth orbit using engineering conversion coefficients. Some historical data gaps are filled with cost information from the closest configuration in the same rocket family, with those imputations documented in the supporting dataset.

That means the $3,868 figure is not a universal customer quote. It is an analytical estimate designed to compare launch technology over time. Actual buyers purchase a specific mission, orbit, schedule, integration service, reliability profile, and contract. New Space Economy’s discussion of payload dollars per kilogram makes the same distinction from a buyer’s perspective: a low theoretical dollar-per-kilogram figure can misstate mission economics when payload mass, orbit, timing, and unused capacity differ.

Why Wright’s Law Fits Space Launch Costs

The central finding is an experience curve rather than a simple calendar trend. Wright’s Law, named for Theodore Wright’s 1936 work on aircraft production, relates unit cost to cumulative experience or production. Terzi and Nicoli adapt the idea to launch by using cumulative payload sent to orbit as the scale variable. Across the full sample, the statistical fit produces an R-squared of 0.791. For 2000 through 2025, the fit rises to 0.953, meaning the relationship between cumulative payload and average cost becomes much tighter in the modern period.

Their estimated coefficient implies that a 1% increase in cumulative payload is associated with an average 0.34% decrease in launch cost per kilogram. When cumulative payload doubles, average cost falls by about 21.2%. The post-Cold War experience curve is steeper: the study estimates a roughly 44% cost decrease for each payload doubling after 1989, compared with about 17% during the Cold War. These figures describe an observed historical relationship, not a physical law that guarantees future prices.

The distinction between an experience curve and a learning curve is important. An experience curve can reflect manufacturing learning, process changes, larger production runs, new vehicle generations, competition, shifts in customer mix, and movement of payload toward cheaper rockets. A narrower learning curve asks how costs change within a particular rocket family as accumulated experience grows. When the authors isolate that within-family effect, they estimate cost reductions of roughly 6% for each doubling of cumulative payload within a family, rising to about 8% in the post-Cold War period.

A January 2026 National Bureau of Economic Research working paper by Ruibing Su, Chenyu Yang, and Andrew Sweeting reaches the subject from a different direction. Its dynamic model allows past launches to improve reliability and lower costs and examines procurement, market structure, United Launch Alliance, Falcon 9, and Vulcan Centaur. That research does not replicate Terzi and Nicoli’s global historical estimate, but it supports the broader idea that repeated launch activity and procurement design can shape cost and reliability over time.

What Actually Drove the Historical Cost Decline

The strongest challenge to the familiar story of launch economics comes from the study’s decomposition of cost change. Popular accounts often attribute cheaper access to space mainly to better rocket technology or booster reuse. Terzi and Nicoli find that market reallocation has been at least as important. They divide the historical record around 1995, when launch-cost variance had settled at lower levels, and separate within-family cost change from movement of payload among rocket families.

Before 1995, the aggregate decline was driven by reallocation toward different launch systems, and the study finds that within-family costs actually rose. The interpretation is consistent with a period when governments tested many designs, abandoned some, retained others, and made choices for strategic or national reasons that were not centered on minimum cost. After 1995, both within-family learning and reallocation contributed to lower average costs. Over roughly the past three decades in the dataset moving more payload onto cheaper rockets contributed more to the aggregate decline than cost reduction inside the same rocket family.

That result changes how Falcon 9 should be understood. Falcon 9 matters because of engineering, production, booster recovery, high launch cadence, and operational standardization, but it also matters because a very large share of payload shifted onto a system with lower recurring cost. The distinction is more than academic. If a low-cost provider gains share, the average cost across all launches falls even if competitors do not improve at the same pace. If customers later move toward higher-cost national systems for security or industrial-policy reasons, the global average can rise even without any technological regression.

The study identifies 2010 as a structural break in the experience curve and associates that date with Falcon 9’s introduction and NASA’s Commercial Orbital Transportation Services program. The statistical break does not prove that one vehicle or one procurement program caused the entire acceleration. It does line up with an important institutional change. NASA’s official COTS history documents the agency’s use of funded partnerships and milestone-based development to create commercial cargo transportation capabilities. NASA later described COTS as the development and demonstration phase that preceded purchased cargo services under its Commercial Resupply Services program.

New Space Economy’s review of SpaceX launch cadence and reuse places Falcon 9’s economics inside a larger operating system of recovery, refurbishment, manufacturing, mission planning, launch infrastructure, and internal Starlink demand. The cost effect of reusable hardware becomes stronger when the entire industrial system can repeatedly put that hardware back into service.

How Launch Learning Compares With Solar Power and Steamship Freight

The study’s most striking comparison is with solar photovoltaic modules. Solar has one of the best-known technology experience curves. Between 1975 and 2019, the researchers report that module costs fell by about 20.2% for each doubling of installed capacity. From 1975 to 2023, solar module prices fell by 99.8%, compared with an 85.3% reduction in the launch-cost series. Yet solar reached that outcome through a much larger increase in cumulative deployment.

When the authors compare Wright’s Law relationships using indexed cost and cumulative scale, they find the space-launch experience curve is steeper than solar’s over the comparable period. This does not mean rockets became cheaper than solar panels in any ordinary price sense. The technologies have different units, markets, supply chains, and engineering constraints. The comparison says that launch cost declined faster relative to the increase in cumulative payload than solar module cost declined relative to the increase in installed solar capacity.

A second comparison reaches back to transatlantic freight after steam power entered ocean shipping. The study combines launch data with historical freight prices for wheat and cotton and estimates that freight cost fell by about 15.5% for each doubling of cumulative freight volume. Space launch again shows a steeper Wright’s Law experience curve. The authors use the analogy because cheaper steam transport helped expand trade, reduce effective distance, and support new commercial relationships. They argue that falling launch cost could have an analogous effect on economic activity beyond Earth, though the mechanism and institutions differ.

The comparison is useful when kept within its limits. Transport cost can unlock markets, but lower transport cost alone does not create customers, profitable products, financing, insurance, standards, or legal certainty. New Space Economy’s feature on why launch economics shapes space markets treats reuse as an operating system that requires turnaround capability, infrastructure, labor, recovery assets, and enough demand to sustain flight rate.

Why the 2030 and 2040 Forecasts Need Care

The study converts its historical experience curve into conditional forecasts. Its central scenario assumes annual payload growth of about 13.3%, based on the 2010 through 2025 average. Under that assumption, average launch cost falls from roughly $3,800 per kilogram near the end of the observed sample to about $1,600 per kilogram in 2030 and $300 per kilogram in 2040. The model also tests a conservative scenario based on 7.5% annual payload growth, a bullish scenario based on 21% growth, and an error-trend-seasonality forecast generated from the data.

Even the conservative data-driven case produces large reductions in the study: about 45% below the 2025 level by 2030 and 75% below it by 2040. A hindcasting exercise, which estimates the curve only through 2015 and then compares forecasts with observed 2016 through 2025 outcomes, finds that the long 30-year growth trend would have produced the best predictions through 2023 within the model’s uncertainty bands. The authors also report that annual payload growth since 2020 reached 31% in their dataset, compared with 4% from 2000 through 2019, giving them reason to think the central forecast may be conservative.

The forecast is still conditional. Wright’s Law turns an assumption about cumulative payload into an implied cost path. It does not independently predict customer demand, launch licensing, launch-site capacity, access to capital, geopolitical fragmentation, satellite manufacturing output, insurance constraints, or debris-related operating limits. If cumulative payload grows more slowly than assumed, the cost curve also moves more slowly. If a technological discontinuity lowers recurring cost faster than historical experience would suggest, the model can miss that in the opposite direction.

Cost and price also need to remain separate. SpaceX’s 2026 launch-services sheet lists a Falcon 9 standard payment-plan price of $74 million through 2026 and identifies 22,000 kilograms as maximum low Earth orbit performance for a fully expendable vehicle. A customer does not automatically receive maximum vehicle capacity, and reusable missions have different performance. New Space Economy’s review of Falcon 9 and Starship cost claims reaches the same conclusion from public company and market information: Starship cost targets should not be treated as verified customer prices before full reuse and sustained cadence are demonstrated.

Launch procurement reinforces that distinction. New Space Economy’s analysis of launch-services procurement describes schedule, orbit, payload integration, risk allocation, mission assurance, and provider cadence as factors that can outweigh a simple price-per-kilogram comparison. A lower cost of transportation is economically useful only when the resulting service matches what the customer needs.

How Reusability and Competition Are Changing the Model

As of August 22, 2026, events since the study’s 2025 data cutoff have already changed some factual details around the vehicles used to illustrate its forecast. Starship remains a development program rather than a routine operational launch service. The Federal Aviation Administration closed the investigation into the preceding test and said in July that Starship Flight 13 could proceed once SpaceX met applicable safety and licensing requirements.

On July 24, 2026, Starship’s 13th flight test carried 20 Starlink V3 test satellites on a suborbital trajectory. The spacecraft completed atmospheric reentry and an Indian Ocean splashdown. The Super Heavy booster did not achieve its intended soft ocean return because five engines failed to relight during the descent, producing a harder-than-planned impact in the Gulf of Mexico. The flight nonetheless supplied additional data on the newer vehicle configuration, payload deployment, heat-shield performance, and reentry.

That status matters because the study discusses a possible Starship cost near $1,000 per kilogram at service entry and the prospect of going below $500 per kilogram as operations mature. Those are estimates tied to successful full reuse, production scale, refurbishment, launch rate, and payload utilization. They are not observed market prices. The study presents Starship as a possible accelerator of its experience curve, not as proof that the forecast has already arrived.

Blue Origin has moved beyond the prototype status attributed to New Glenn in the paper. New Glenn’s third mission launched on April 19, 2026, carrying AST SpaceMobile’s BlueBird 7 satellite. The booster returned to Blue Origin’s offshore landing platform after the company had already demonstrated a successful first-stage landing during NG-2 in November 2025.

A May 28, 2026 hotfire test then produced a significant anomaly at Launch Complex 36 and interrupted the near-term flight plan. On August 5, Blue Origin reported that its New Glenn investigation had traced the initiating problem to the main oxygen valve on one BE-4 engine. The company said recovered hardware and inspections supported that finding, that it was modifying the valve, and that updated hardware was being manufactured. Blue Origin continues to state that it plans to return New Glenn to flight before the end of 2026.

Competition is broadening outside the United States as well. On August 18, 2026, China’s LandSpace successfully recovered a Zhuque-3 first stage after an orbital launch. The landing does not establish Falcon 9-like economics or repeated reuse, but it demonstrates that controlled recovery of an orbital-class booster using landing legs is no longer confined to U.S. providers.

Policy is also shifting toward much higher cadence. The August 20, 2026 National Space Transportation Policy establishes a U.S. objective of supporting more than 1,000 launches and reentries annually by 2030. It directs federal agencies to address launch and reentry infrastructure, federal ranges, spectrum, commercial access, environmental reviews, additional sites, industrial capacity, and related constraints. That figure is a policy objective rather than a payload forecast, but it shows that federal planning is beginning to accommodate launch rates far above historical norms.

Why Market Power and Geopolitics Can Slow the Cost Curve

A steep experience curve can coexist with concentrated market power. The PNAS Nexus study explicitly separates technological cost from the price charged to customers and warns that a dominant provider may have the ability to retain part of a cost advantage rather than pass every saving through to buyers. It also estimates that SpaceX carried roughly 75% of global payload during the study period, making future average-cost outcomes unusually sensitive to one company’s flight rate, vehicle mix, and Starship execution. That percentage should be read as a study-era estimate rather than a permanent market share.

The concentration issue cuts in two directions. Scale can accelerate learning, spread fixed operating expense, support frequent flights, and make reuse more economical. Concentration can also reduce competitive pressure and give governments a strategic reason to maintain alternate launch systems even when those systems cost more. Terzi and Nicoli argue that greater geopolitical tension could recreate part of the Cold War pattern, in which states favor sovereign access and strategic autonomy over the cheapest available vehicle. Their decomposition makes this concern stronger because payload reallocation toward cheaper rockets explains so much of the historical decline.

New launch options can moderate that risk, but they do not erase it. Blue Origin, United Launch Alliance, Rocket Lab, Arianespace, European development programs, Indian providers, and Chinese launch systems differ in payload class, orbit, certification, national-security eligibility, launch site, schedule, and export constraints. The NBER research on competition and procurement also treats procurement structure and market organization as economic variables that can alter learning and launch outcomes rather than assuming technology operates separately from institutions.

Governments can rationally accept a higher launch cost if the cheapest alternative creates dependence on a foreign or commercially concentrated supplier during conflict, sanctions, export restrictions, or diplomatic disputes. That behavior can slow the global average-cost decline even when the lowest-cost rocket keeps improving. National launch programs can therefore look inefficient on a narrow accounting basis and defensible when security, industrial capacity, supply-chain continuity, and assured access are included.

What Lower Space Launch Costs Could Change in the Space Economy

Launch cost affects the economics of every payload that must leave Earth, but its influence varies by business model. Lower cost can reduce the penalty for mass, permit more shielding or propellant, increase tolerance for replacement missions, and make frequent replenishment more feasible. It can also support larger constellations, more frequent Earth-observation refresh cycles, larger communications platforms, orbital servicing equipment, scientific instruments, and early in-space manufacturing experiments.

The PNAS Nexus researchers specifically point to microgravity manufacturing and pharmaceuticals as activities that could become more financially attractive if transport cost continues to fall. The connection is straightforward: when transportation consumes less of a mission’s budget, businesses have more room to spend on hardware, operations, experimentation, redundancy, and the return or delivery of manufactured products.

The economic effect is not proportional to the launch-cost reduction. For a satellite operator, launch may be one line in a budget that also includes spacecraft production, payload development, testing, insurance, ground infrastructure, spectrum rights, operations, financing, and customer acquisition. Some missions become viable when launch price crosses a threshold. Others remain constrained by demand or by the cost of the satellite itself. Heavy infrastructure concepts, including large orbital platforms and bulk cargo, are more sensitive to dollars per kilogram than a high-value scientific spacecraft whose payload development can cost far more than launch.

Lower transport cost can also reshape design behavior. Engineers have historically paid heavily for mass reduction because every kilogram consumed scarce launch capacity. As transport becomes cheaper and more available, some programs can trade mass efficiency for lower manufacturing complexity, greater redundancy, larger power systems, or shorter development cycles. That trade is already visible in commercial constellations that manufacture satellites at higher rates and replace them more frequently than traditional government spacecraft.

Demand still has to materialize. The PNAS Nexus study acknowledges that its model focuses on launch supply and does not model the demand side needed to translate lower recurring cost into customer price or total market size. That limitation is one of the study’s most important boundaries. Cheap launch can make a business technically feasible without making it profitable.

New Space Economy’s broader treatment of launch economics reaches a similar result. High cadence requires a continuing flow of payloads, capital, ground capacity, regulatory approvals, and customers rather than reusable hardware by itself. Lower launch cost is an enabling condition. It is not proof of a self-sustaining market.

Why Orbital Sustainability Could Become an Economic Constraint

The same scale that pushes the experience curve downward can create congestion costs in orbit. Terzi and Nicoli identify orbital debris as one reason historical cost relationships may fail to extend cleanly into the future. More payload, more satellites, and more launches can raise collision-avoidance requirements, tracking needs, disposal obligations, insurance exposure, and regulatory attention. In a severe case, growth in traffic could make some low Earth orbit regions harder or more expensive to use.

Current European Space Agency statistics reinforce that concern. ESA’s Space Debris Office reported in its July 31, 2026 update that about 46,250 space objects were being regularly tracked and maintained in surveillance catalogs, about 18,840 satellites remained in orbit, and roughly 16,000 of those satellites were still functioning.

ESA’s 2026 Space Environment Report adds the growth dimension. It reports that 2025 saw more than 300 launches and about 4,000 new payloads added to the space environment. The report also evaluates the long-term collision and fragmentation risk produced by continued accumulation of orbital objects and concludes that stronger mitigation and remediation are required to move toward a sustainable orbital environment.

This creates a feedback that simple launch-cost forecasting does not capture. Lower launch cost encourages more deployment. More deployment can increase traffic-management expense and regulatory requirements. Those expenses may not appear in the rocket’s unit flyaway cost, yet they can raise the total cost of operating a space business. Environmental rules can also change vehicle design, disposal planning, upper-stage practices, licensing timelines, and satellite end-of-life requirements.

That pressure gives economic value to capabilities that once sat outside the usual launch-cost discussion. Better traffic coordination, more reliable conjunction data, faster disposal, controlled reentry, and verifiable compliance can become part of the cost of maintaining high launch cadence. A launch provider may keep lowering recurring vehicle cost even as customers face higher spending elsewhere in the mission chain. The experience curve can keep moving downward at the rocket level without making orbital access frictionless.

Summary

Terzi and Nicoli provide one of the strongest empirical cases yet that space launch costs have followed a steep experience curve. Their standardized launch dataset and analysis indicate a decline from $87,023 per kilogram in 1960 to $3,868 in 2025, with about a 21.2% average cost reduction for each doubling of cumulative payload. The modern relationship is tighter than the full historical sample, and the study finds that the post-Cold War period delivered much faster cost improvement than the Cold War era.

The more consequential finding concerns the mechanism. Lower average cost came from both learning within rocket families and a shift of payload toward cheaper systems, with reallocation contributing more over the later decades in the dataset. That makes competition, procurement, market share, and geopolitical choice part of the launch-cost story rather than side issues.

The study’s $1,600-per-kilogram estimate for 2030 and $300-per-kilogram estimate for 2040 are plausible outputs of its model under sustained payload growth, but they are not guaranteed market prices. Developments through August 22, 2026 point in both directions. Starship has completed its 13th test flight but has not demonstrated routine full reuse. New Glenn has completed three missions but is recovering from its May 28 hotfire anomaly. China has demonstrated an orbital-class booster landing with Zhuque-3. U.S. policy is preparing infrastructure and regulatory systems for much higher launch and reentry cadence.

The economic question is no longer whether launch can become cheaper. The historical record already shows that it has. The harder question is how much of the next cost reduction can be converted into lower customer prices and economically productive activity before market concentration, sovereign launch policies, infrastructure limits, and orbital congestion absorb part of the gain.