This is an idea I had kicking around for a while before SpaceX started getting serious about reusable rockets. It seems wasteful somehow, but if we had an urgent need to start lifting a lot of mass in a short time I think it could still be useful.
Consider a booster family constructed entirely of modular tanks and one or two types of engine. All components are sized to fit inside standard shipping containers. The fuel is liquefied natural gas and the oxidizer is liquid oxygen. One or more launch platforms can be built out of converted floating drydocks or oil-rig tugs; add a solar-powered cryogenic oxygen plant and docking facilities for both commercial LNG tankers and commercial containerized cargo ships.
A container ship delivers the components for several launches. The booster is assembled on the deck of the platform using cranes and quick-lock connectors. A LNG tanker delivers the fuel and the on-platform plant delivers the oxygen. Customer payload is delivered by air if it is light enough, otherwise by sea. Any nearby vessels clear the area, then the booster is launched. Wash, rinse, repeat. If there is enough onboard storage, the fuel and components for several missions can be taken onboard and then launched as fast as they can be assembled while the next shipment is in transit.
I did a first cut estimate for seven payload classes ranging from 15 to 122 metric tons into LEO (gross dV of 10km/s). I used my tank estimator and doubled the mass to account for plumbing / intertank / struts; tanks could be aluminum, titanium, Al-Li or composite overwrap and still fit the mass budget. Inconel tanks would require some consideration and stainless steel may not be profitable. I checked other fuel mixes (hydrogen, RP-1 and UDMH), but LNG / methane is adequate and much cheaper. I used performance numbers from several Russian engines, notably the RD-0139, RD-0120M-CH and RD-0141 (though most methane engines are experimental at best).
I get mass ratios of 28.0 down to 21.6 (payload fractions of 3.6-4.6%). For comparison, a Proton-M gets 32.4 (3.1%) and the Saturn V got 23.3 (4.3%). Methane engines are about 3.4:1 oxidizer to fuel. LNG is about 450 kg/m³.
A single Q-max LNG carrier might hold 220,000 m³ or 99,000 tons. A launch of the 15-ton payload configuration would need only 89.5 tons of LNG, while the 122-ton payload would require 551.6 tons of LNG. One carrier could supply fuel for six months of daily launches. In 2013, the United States alone imported nearly 3 million cubic meters of LNG by sea (quoted as 2 billion m³ of gas); launching one 122-ton payload a day would consume 15% of that amount. We exported 44.5 billion m³ of gas in 2013 (64.8 million m³ LNG). At about 450,000 m³ per year, daily launches could consume 0.7% of U.S. exports.
The liquid oxygen needs are higher, 304.3 tons for the 15-ton payload and 1,875 tons for the 122-ton payload. There are commercial plants that produce up to 7,000 tons per day of LOX, and multiple providers are willing to operate a plant of that size for you; a byproduct of those plants is access to bulk liquid argon as electric tug fuel, produced on-site.
A single Panamax cargo carrier can hold up to 5100 TEU, or 2,550 40-foot containers. Each tank is 2 TEU and three engines can fit in one 2-TEU container. A 15-ton payload requires only 16 TEU, while the 122-ton payload requires 94. That cargo ship could carry components for over seven weeks of daily launches of 122 tons. Larger ships can be used as long as they don't have to travel through the Panama canal. A larger number of smaller ships (Feeders, 2000-3000 TEU) could be used as well.
As for the components themselves, it is hard to make a good estimate. Let's assume each component on average is comparable in complexity to building a car or truck. US production of passenger and commercial vehicles in 2013 was over 11 million units. We require only 24,090 components (66 * 365) per year, a drop in the bucket in terms of industrial output. Granted a rocket engine, a fuel tank and a passenger car are not exactly the same, but today we build rocket engines by hand. In this scenario they would be manufactured by the thousands on an assembly line with minimal labor input. To compare to something the same size as the fuel tanks, somewhere between 1 and 2 million shipping containers are built every year (mostly in China); our demands would be only 2.4% of that output.
Structural mass is estimated to be about 103 tons per launch, 37,600 tons per year. North America produced 368,000 tons of aluminum just in June of this year and 4.585 million tons for 2014. The demand is only 0.8% of annual production, and that's not considering the possibility of imports.
Let's assume some kind of crash program requires daily launches of maximum payload. Existing industry is more than capable of supplying the necessary fuel, materials and transportation using existing infrastructure. Let's also assume a 90% uptime (1 out of 10 launch opportunities fail for some reason, but the platform is armored enough to avoid damage). This system could put 40,000 tons in LEO per year. If that's not enough, convert ten tugs into launch platforms and launch 400,000 tons per year; existing industry can support it with reasonable conservation.
If it's still not enough then you'll need to build everything bigger: larger tanks, larger engines, dedicated transport ships and more launch platforms. At that point we've lost most of the advantages of using existing infrastructure but we're operating at a scale that challenges the existing infrastructure anyway. With worldwide cooperation it should be possible to launch 4-12 million tons per year.
Tanks are relatively simple. I think they could be built for $50,000, but let's assume $100k. Engines are not so simple; call it $200k each. The largest vehicle uses 29 engines and 37 tanks. That's $9.5 million in hardware. The oxygen plant is a fixed cost in the neighborhood of $1 million per month or about $35k per launch. The highest price for natural gas in the last 5 years in the US was $5 per mmbtu, which is $121.35 per m³ of liquid or a bit under $150k per launch. Shipping costs might be around $5k per container or $330k per launch. That puts the marginal cost of a launch at $10.015 million. As a gut check, 103 tons of aluminum at $2 per kg is $206k; this price seems within the realm of possibility for mass-produced modular components.
The launch platform is based on a hull that runs about $200 million; let's assume the retrofits cost an additional $100 million. Maintenance will be 20% of that or $60 million per year (5% might be closer to the truth, but we're proposing a pretty stressful environment). Let's assume the design life is 10 years; that gives $900 million over 10 years. If the ship is funded by a 10-year bond at 5% that's $1.15 billion or $9.55 million per month or about $320k per launch.
The assembly lines are harder to price. Here's an article suggesting that they do not need to be as expensive as they have been historically. Here's an example factory that used an existing building and refitted for about $60 million and has over twice the necessary output. Let's double that figure and call it $120 million plus 10% per year for 10 years, $240 million funded as a 10-year bond at 5%, $318.2 million in total or $2.65 million per month or about $90k per launch.
Labor is also a bit difficult to figure. The Tesla factory I mentioned above has 3,000 employees. Due to the rigorous test requirements of aerospace components let's double that. We can also assume that launch operations will require a similar number of people. Add an extra thousand people for things I haven't considered and you are at 10 thousand people. The lowest I'd pay would be $40/hr; figure the average salary is $100k per year. Benefits and employer expenses cost another $40k per year or so. That's a $1.4 billion payroll, $3.889 million per launch.
Each individual launch now costs $14.314 million. 360 launch attempts per year put the annual cost at $5.153 billion. Assuming only 90 tons of useful payload per launch (the rest being shroud and other parts I've forgotten) and 324 successful launches, that's 29,160 tons in LEO or $177k per ton ($177 per kg). That 32 tons margin is sufficient for electric propulsion to move the cargo to GEO.
As for what we might do with that mass, who knows. Massive asteroid redirect perhaps. Build both L1 and L2 lunar elevators with a tramway connecting the poles. Permanently house a few thousand people per year on the moon or on Mars. Or we could build power satellites.
80's technology projected 60,000 tons for a 5GW solar power satellite; with three times the PV efficiency you could put 7.3 GW of capacity in orbit per year with one platform. Covering the entire 2.5 TW world demand in 20 years (roughly the lifespan of an SPS) would require launching 125GW of capacity every year, or 500,000 tons. That would take 18 platforms, or about $93 billion per year.
With a 99% uptime there are 8,672 power-generating hours per year. 125 GW of capacity yields 1.084 petawatt-hours at a launch cost of $0.0858 per kilowatt-hour in the first year. Over the 20-year life of the system, 21.68 petawatt-hours are generated at a launch cost of $0.0043 per kilowatt-hour. This ignores the cost of the SPS hardware and power receivers as well as the cost of long-range transmission lines for countries outside 30° or so of the equator.
US generating costs are as low as $0.02742 per kWhr for nuclear, the cheapest scalable power. Hydroelectric is cheaper but there is not enough potential capacity to power the world. If the station hardware is $4.01 per watt or less then it will be cheaper than nuclear power. If the hardware is $5.71 per watt or less it will be cheaper than coal. (At least one other source quotes much higher production costs.) Let's assume $5 per watt due to the long timeline and enormous production requirements ; that means $625 billion per year in SPS hardware. That's comparable to solar thermal capital costs. For systems completed in 2013, residential PV was around $4.69/W and utility scale was $3.00/W. Utility scale was projected to hit $1.80/W for installation in 2014. It seems reasonable to expect that high-efficiency PV could be produced for $5/W or less, particularly using concentrators and multijunction cells; if the overall SPS could be built for utility-scale PV prices around $2 per watt it would be half the cost of nuclear power and comparable in cost to hydroelectric.
That's a total of $718 billion per year. Add 20% margin to reach $860 billion per year and compare to gross world product of about $74 trillion; the project consumes 1.162% of the world's output and completely eliminates the use of coal for electrical power within 20 years. If the US were to do this alone it would take only 5.13% of current GDP to own the world's electricity supply while generating tens of thousands of permanent high-paying jobs. The EU could do it for 6.42% of euro-area GDP, while China could do it for 12.6% of GDP (and their government is capable of that kind of commitment). Most of the money would be spent on labor and resources within the country rather than on imports, representing a massive economic boost even before the power stations become profitable.
Maybe it would be worth doing this if only to eliminate carbon pollution during electricity production. The methane burned during launch releases CO2 (2.68 million tons of carbon per year), but compared to coal or oil it is very clean. 41% of world electrical production is from coal, which averages 278 kg carbon per MWhr. At 89.79PWhr of coal electricity that's 24.96 billion tons of carbon per year in direct emissions, not to mention the avoided methane release and transportation emissions.