From building the rocket to renting the pad to reaching orbit: who takes part in a commercial space launch?

Commercial space launch is still not an industry with much of a division of labor.

Four days ago, at a Westworld panel at the SXSW festival, a video drew several minutes of applause — and it had nothing to do with Westworld.

Jonathan Nolan, the writer and director of Westworld and the younger brother of the more famous Nolan, had recut footage of SpaceX's successful Falcon Heavy launch on February 6: a Tesla roadster drifting in space, set to David Bowie's "Starman."

When it ended, the whole room stood up and clapped, and a few people were excited enough to swear out loud.

This is Musk's own version, with the soundtrack swapped for David Bowie's Life on Mars?

"Last year, over drinks with a friend, we talked about how to inspire people, how to get people talking to each other," Nolan said after the video played. The result was that red roadster and a David Bowie song.

Then Nolan's "friend" — SpaceX CEO Elon Musk — walked on stage and quoted Konstantin Tsiolkovsky, the father of astronautics: Earth is the cradle of humanity, but humanity cannot stay in the cradle forever. The audience clapped and screamed again, the way people do when they meet an idol.

Musk spoke for two minutes. Besides saying that the work he does now is the work his heart is set on, he said SpaceX is pushing ahead with its next-generation rocket, the BFR, and that Mars should one day have factories, bars and nightclubs.

Private companies are now doing what once took the full weight of a national government — and doing it faster and cheaper.

More numbers show how hot commercial spaceflight has become. This January, the US government said it would simplify space regulation to encourage private space companies.

Commerce Secretary Wilbur Ross went further, calling space a $330 billion market that could create at least 210,000 jobs. He said there are 1,700 private space companies worldwide, 45% of them in the United States.

Rockets matter most in the space business, and anyone who wants a new one has to start from scratch

In 2001, Musk's first space plan was a plant greenhouse on Mars. He framed the project then as a way to lobby the government for a bigger space budget and to draw public interest to space exploration.

He changed his mind quickly. Without a fundamental breakthrough in rocket technology, he realized, leaving Earth would cost too much and none of the plans could be carried out.

So Musk started thinking about launching rockets himself. He flew to Moscow with the mechanic Jim Cantrell and his college friend Adeo Ressi, hoping to buy Russian Dnepr rockets and run a launch service. The Dnepr was converted from the Soviet intercontinental missile the NATO codename called the SS-18 "Satan," and the type was easy to come by after the Cold War.

He hit a wall. ISC Kosmotras, the Russian company that operated the Dnepr, dismissed the outsider out of hand. Even to buy a single rocket, Musk needed a friend who worked at In-Q-Tel, the CIA's venture arm, to help him form a consortium with Orbital Sciences; the group paid $8 million for one Dnepr.

By then Musk had lost all faith in buying rockets. The value for money was too poor, he decided, and he would build his own.

In SpaceX's first days, Musk estimated that a fully in-house "vertical integration" model, plus the modular thinking common at internet companies, could cut launch prices by a factor of 10 and still leave a 70% gross margin.

Judging by today's Falcon rockets, he mostly hit those targets. But the process was nothing like a new brand entering the phone or car business today, where partners exist for every piece of software and hardware, and money alone is enough to get a product built — whether it sells is a separate question.

Plenty of companies supply parts for government space programs, but nobody sold a kit for turning those parts into a new rocket. And as the US space program shrank after the Space Shuttle accidents, the technology itself stopped advancing.

A rocket's main parts are the outer structure, the propellant tanks, the electronics and, most critical of all, the engines.

The engine is the highest technical hurdle in the whole machine. Whatever principle it runs on, today's launch vehicle is structurally little more than a way to hold fuel and carry the payload on top.

A Falcon 9 weighs 500 tons at liftoff and carries 2.2 tons of payload — 4% of its liftoff weight. Almost all the rest is the fuel needed to break free of Earth's gravity. For most rockets the ratio is below 4%.

There was no commercial engine house to buy from, so Musk recruited Tom Mueller, a rocket engineer and former vice president at TRW. Mueller is the world's foremost authority on rocket engines; at TRW he designed two liquid-fueled engines for NASA. As a founding employee and CTO of SpaceX, he led the design of the company's core component, the Merlin engine.

SpaceX's latest engine, the Merlin 1D+, is a medium-class engine burning kerosene and liquid oxygen. It uses a gas-generator cycle, a regeneratively cooled nozzle and a pintle-type throttle valve.

In short, it can throttle thrust by 30%, restart and be reused, and it is simpler and cheaper than the engines Boeing uses on its Delta rockets.

Mueller's experience in the field also made it easier for SpaceX to source the engine's key component, the turbopump, from Barber-Nichols.

The rest of the rocket's materials are not hard to get. The 2195 aluminum-lithium alloy in the Falcon's shell is a standard material used in spacecraft, rockets and aircraft alike, and SpaceX buys it on the open market from the Canadian supplier Alcan.

For fabrication, SpaceX chose friction stir welding, a process widely used in rockets and shipbuilding. It appeared in the early 1990s and is mainly used where aluminum needs a high-strength joint. SpaceX's friction stir welding came from General Tool Company (GTC), which also supplies Boeing.

A rocket also needs electronics and onboard computers to fly it. Rockets used to carry radiation-hardened, space-grade chips: expensive, slow, and built for reliability above all.

SpaceX reasoned that a rocket only works for a few minutes, so it does not matter if commercial chips are unreliable — add enough redundant backups and they will not all fail inside the flight window.

Each Falcon 9 engine is controlled by three computers, each with two independent main chips. The operating system is written in C++ and runs on Linux, no different from an ordinary industrial computer.

The upshot: with unremarkable Merlin engines and every trick available for shaving structural mass, the Falcon 9 gives 24/25 of its launch weight to fuel. That leaves the least possible "dead weight" the rocket cannot use, which means the highest efficiency and the lowest launch cost.

In the end SpaceX makes 85% of the rocket's parts itself. Compare that to building a phone today, where a company buying off-the-shelf designs barely has to design or make anything at all.

Bringing out a new rocket is still work a private company has to do from scratch. With the engine and design bottlenecks on top of that, a latecomer cannot simply buy the technology and end up with a rocket more efficient than what is already flying. And a private company's whole value in this industry is costing less than a government program.

The private space company that came before SpaceX launched satellites on converted military missiles

SpaceX is the most technically advanced private space company today, but it was not the first.

Orbital Sciences was the first private firm to run launch missions. Founded in 1982, it launched its first Pegasus rocket successfully in 1990 — the first launch vehicle in history developed and flown by a private company. After that launch, Orbital Sciences went public.

But the Pegasus is nothing like the launch vehicles that stand a dozen stories tall. It is a small, solid-fueled, air-launched rocket roughly the size of a sedan, slung under a large aircraft, carried to altitude and released there. Its capacity tops out at small satellites of about 100 kilograms.

After the Pegasus, Orbital Sciences introduced the Minotaur, built from retired US LGM-30 "Minuteman" and LGM-118A "Peacekeeper" solid-fueled intercontinental ballistic missiles, as part of an Air Force program to convert decommissioned missiles.

The newest version, Minotaur 5, can put 630 kilograms into geosynchronous orbit. It has flown once, in 2013, sending NASA's LADEE lunar atmosphere and dust probe to the Moon.

Technically the Minotaur is an ICBM with a new destination, so Orbital Sciences' business amounts to operating it. The Antares, first flown in 2013, was the company's conventional liquid-fueled launch vehicle.

The Antares is a contemporary of SpaceX's Falcon. In 2008, to cut its budget, NASA opened public bidding for the Commercial Orbital Transportation Services program, aimed at resupplying the International Space Station and carrying crew. Winners had to develop both a rocket and a spacecraft and fly dozens of resupply missions.

After bidding against Boeing, Lockheed Martin and a dozen other companies, SpaceX and Orbital Sciences won, with contracts worth $396 million and $288 million. To date Orbital Sciences has used the Antares to launch nine Cygnus cargo craft, with three more flights planned through 2019.

Building the rocket is one thing; launching it is another job entirely

A rocket of the Falcon 9's class needs a dedicated pad. Before liftoff, its sections travel from their factories to a hangar near the pad, are assembled horizontally, and are then moved to the launch mount, which raises the several-hundred-ton machine upright, fills it with fuel and waits.

Different satellites make different demands on the site. Latitude decides how much of Earth's rotational speed a launch can borrow. Low-latitude sites near the equator suit west-to-east launches into low Earth and geosynchronous orbits; higher-latitude sites make it easier to climb into high-inclination polar or sun-synchronous orbits.

Europe, for instance, sits too far north, so the Ariane rockets fly from French Guiana in Latin America, near the equator. China once built Jiuquan in its northwest for reasons of national security, at a real cost in performance; it has since built a new site at Wenchang, on Hainan.

SpaceX operates three launch sites. One is pad SCL-40 at Cape Canaveral Air Force Station in Florida, leased from the Air Force in 2007 and used mainly for Falcon 9 flights to low Earth and geosynchronous orbit.

In 2014 SpaceX signed a 20-year lease with NASA for Launch Complex 39A at the Kennedy Space Center, also at Cape Canaveral.

That pad once launched the Apollo missions; to fit the enormous Saturn V, LC-39 was given a huge launch structure and assembly building. After Apollo ended in the 1970s, it was rebuilt for the Space Shuttle. With the shuttles retired and no other rocket able to use their equipment, NASA mothballed the complex. Once it was leased to SpaceX, the company paid to modernize the assembly building, the transport road and the launch structure, and added a landing zone, using the pad for crewed Falcon 9 missions and the Falcon Heavy. NASA, meanwhile, saves $1 million a year it used to spend keeping an idle pad in storage.

In 2011 SpaceX leased Space Launch Complex 4E at Vandenberg Air Force Base in California from the Air Force; the lease has been extended once and runs to 2019. That base handles polar and sun-synchronous missions.

Renting a pad takes more than money. It takes government review and an environmental impact assessment — and possibly a fight with a rival. When SpaceX went after the Kennedy Space Center complex, Jeff Bezos's Blue Origin bid against it, and during the bidding Blue Origin protested to the US Government Accountability Office that the process might be rigged; SpaceX said it was willing to share the pad. Blue Origin lost in the end, largely because SpaceX's plans were further along.

Having a pad does not mean you launch when you please. At Cape Canaveral, launches answer to the Air Force's 45th Space Wing, which also provides support: clearing airspace, forecasting weather, handling security. SpaceX is a private company and is not subordinate to the Air Force on a commercial mission, but it depends on the 45th Space Wing to coordinate.

At Vandenberg, launches are coordinated by the 30th Space Wing. Once SpaceX's private launch site in Texas is finished, its launches will fall under Air Force Space Command (AFSPC).

A commercial communications satellite also generally needs a green light from its home country's telecom regulator. In the United States, a domestic communications satellite launching either at home or abroad must report its functions, frequency bands, power, orbit and other details to the Federal Communications Commission for approval.

A satellite-approval mix-up surfaced just recently. The US company Swarm Technologies had four tiny relay communications satellites launched on India's PSLV rocket; only after they reached orbit did India discover that their launch license had been rejected on safety grounds. The FCC says it is investigating.

The one industrialized space business is renting space on an existing rocket — and even that is mostly government-led

Most artificial Earth satellites are communications satellites. Humanity has launched 1,337 civilian communications satellites so far, 16.07% of everything in Earth orbit and the largest share of any category. After them, the most numerous types are military spy satellites, military communications satellites, and experimental and technology satellites.

In practice a company that needs satellite service today needs one of three things: geographic information — photographing the Earth — communications, or navigation and positioning.

Navigation is a government project nearly everywhere: GPS in the United States, BeiDou in China, Galileo in Europe, GLONASS in Russia.

Photographing the Earth and communications are what commercial companies actually want satellites for.

The high-resolution overhead imagery in Google Maps and Google Earth, for example, comes from the US Geological Survey's Landsat 7 and Landsat 8. And to offer internet access everywhere on the planet, SpaceX, Facebook and others plan to rent or launch satellites and assemble them into constellations.

A company that needs a satellite of its own does not need to know how to build one. It states its requirements and budget to a satellite design-and-manufacturing firm, which builds the satellite to order; after that the job is negotiating a launch contract with a company like SpaceX, Arianespace or CASIC, and waiting for the thing to go up.

This part of the business is thoroughly industrialized, but the party at the other end of the phone is usually a government agency or a state-backed company.

The world's large satellites are built by a handful of defense contractors: Airbus Defence and Space in Europe; OHB, which builds the Galileo navigation satellites; Boeing, which builds GPS; the American firm SSL, which has built more communications satellites than anyone; and Russia's JSC, which builds the GLONASS navigation satellites and the Ekspress communications satellites.

Students and research institutes that need a small satellite with simple functions — or just want to put one up for fun — have cheaper options. One is the CubeSat standard developed at Stanford University: a 10-centimeter cube you assemble yourself, which can carry cameras, radios, experiments and other instruments, plus station-keeping hardware such as nitrogen thrusters.

Launching one of these does not require a big-rocket contract like the Falcon 9. They ride to space as "secondary payloads."

Swarm Technologies' four tiny relay satellites, the ones on India's PSLV, went up alongside the launch of a large Indian mapping satellite.

Companies exist to broker these rides. The American firm Spaceflight, for one, buys the spare payload capacity on large launches and resells it in packages to organizations with small satellites to fly.

On Spaceflight's website you can see a timetable of upcoming large-rocket launches with their orbital parameters and dates; enter your satellite's size and weight and you get a price. The company sells rides on the Falcon 9, Russia's Soyuz, India's PSLV and Orbital ATK's Minotaur. A 10-centimeter CubeSat costs $295.

Finally, launching a satellite still carries risk, which means insurance.

Facebook found this out in 2016. On September 1 that year, a Falcon 9 was to carry the Israeli-built Amos-6 communications satellite to geosynchronous orbit; Facebook had paid $95 million for five years of part of its bandwidth, planning to bring satellite internet to parts of Africa. During a static-fire test on the ground before launch, the rocket exploded and nothing was left.

Facebook, as a customer, could get its money back. The losses of Spacecom, the satellite's operator, and of SpaceX were in theory the insurers' problem. The satellite cost $200 million to build and the rocket $16 million; Lloyd's paid Spacecom $196 million. SpaceX's insurer, Aon, paid nothing: the rocket had not exploded during a launch, so the loss fell outside the policy.

There are dozens of commercial launches worldwide every year, with a cumulative success rate above 95%, and many insurers write space policies. In China, launches of the Long March family are underwritten by PICC.

The rockets all this business depends on include SpaceX's and Orbital Sciences', but more often they are leftovers of the government space race. The most successful operators today are Russia and Europe.

From the 1950s through the 1970s, under the US-Soviet space race, almost every rocket launched carried a state-sponsored spy satellite or a crewed spacecraft. After Apollo ended in the 1970s, those existing heavy-lift rockets started carrying commercial payloads such as communications satellites, and a commercial launch market took shape. The Europeans got there first.

In 1980, under the lead of the French space agency CNES, the governments and companies of 10 countries put up the money to found Arianespace, whose stated purpose was to launch satellites for customers anywhere in the world. From 1980 to now, Arianespace has fielded five rockets, Ariane 1 through 5, and flown 228 successful launches.

Ariane ranks eighth among rocket families by number of launches, and it is one of the few that mostly carry commercial payloads. Every family that has flown more often — the Soviet and Russian Soyuz and Proton, the American Delta and Atlas, China's Long March — was designed for military and government needs. Between 1979 and 2009, Ariane rockets put 300 satellites into orbit, 236 of them commercial; the rest were scientific and military satellites and interplanetary probes.

Besides Ariane, Russia's Proton is the other family used mainly for commercial launches. It has 378 successful flights to date, though most of them were government missions from before the Soviet Union broke apart.

China's Long March, Russia's Soyuz and India's PSLV and GSLV also compete for commercial launches.

SpaceX's Falcon 9 is now one of the most-flown rocket families in the world, with 18 launches in 2017. Ariane 5 flew 14 times that year, Proton-M three. For 2018, 25 Falcon 9 launches are on the books, against 15 for Ariane 5.

More companies are entering the business, most of them starting small

SpaceX's success has many causes. It caught NASA in a budget-cutting mood, took the agency's contract prepayments, and pulled market share away from the traditional military contractors. And Musk's personal pull brought in veterans like Tom Mueller, while promises of colonizing Mars brought in attention and money.

What should not be missed is that even for a company with a billionaire behind it, like Blue Origin, the engine is still the biggest bottleneck. Private firms like Orbital Sciences managed to launch satellites in the past by using cast-off ballistic missiles. Blue Origin was founded before SpaceX, and 17 years on it still has not put a rocket into orbit.

Of the 17 private companies that once bid for NASA's station resupply missions, only Orbital ATK and SpaceX are left. Even taking the Commerce Secretary at his word that 1,700 new companies have entered the industry, commercial space remains a market with an extremely high wall around it. Building rockets will probably never be as ordinary as building phones.

But demand for launches looks like it is growing. Among startups, the American company Planet flies shoebox-sized CubeSats into polar orbit to photograph the Earth in high resolution; in early 2017 it put a cluster of 88 satellites into orbit on a single Indian PSLV launch.

And OneWeb, which has proposed a 648-satellite constellation, needs to launch 600 microwave-oven-sized satellites before 2025 — more than were launched in the past 10 years.

None of these new satellite-based uses work unless rockets can serve them. Large satellites or small, rideshares or dedicated contracts, a cheaper way into space is the precondition. The defense contractors' launch services faced no competition and were priced accordingly — astonishingly high, with no room to haggle. Private launch broke that.

Go a step past satellites — Musk's trip to Mars, or the orbital tourism Virgin Galactic and others are working on — and the whole business case still rests on cost. A $10 million ticket would find takers, but a cheaper one finds more of them, and that is what grows the industry.

As long as launch costs stay at tens of thousands of dollars a kilogram, every commercial plan built on top of them is wishful thinking.

That is the work SpaceX and companies like it are doing: building cheaper rockets. And there appears to be no shortcut.

Images/SpaceX