Hyperloop Part I: A $100 billion boondoggle?
Critics claim Hyperloop will cost $60-100 billion, Elon Musk says $6. A thorough cost analysis demonstrates that it will likely cost around $11 billion.
When Elon Musk, the modern day Howard Hughes (or real life Tony Stark), unveiled his “Hyperloop” concept in August, he made a series of bold claims about how this “fifth mode of transport” would revolutionize the world. It would not only be faster and cleaner than conventional air, land and sea travel – it would be dramatically cheaper, too. It was this last benefit that would make the Hyperloop, first proposed for the San Francisco-Los Angeles’s corridor, impossible for governments, industrialists and Big Business to resist. Now that the hype and hysteria surrounding the announcement have died down, I thought it an appropriate time to investigate these claims.
Musk estimates that it would cost $6 billion to build the passenger version of Hyperloop from San Francisco to Los Angeles (with costs increasing to $7.5 billion for the vehicle version); however, others have estimated the true cost would range from $60-100 billion. Despite accusations that Musk’s costs are wildly inaccurate, as best I can tell, no one has provided detailed independent cost analysis of Hyperloop. Leaving aside the many engineering hurdles that have yet to be cleared, my best estimate is that the true cost for Hyperloop, as proposed, would range from $6.1 billion to $9.8 billion. This estimate does not include extending Hyperloop deeper into the core of Los Angeles, which would likely increase total project cost to $6.6-11.2 billion. Due to the complexity of the project, to do a true cost analysis on any on component of Hyperloop would likely be the work of a thesis project; and given the highly variable nature of costs in civil engineering projects, the new applications and combinations of technologies, the conservative assumption is to assume Hyperloop will fall towards the high end of that range.
While this estimate is close to Musk’s estimate, the potential for up to an 80% rise in cost would destroy Musk’s promise of a $20 one-way ticket, and potentially the competitiveness of the Hyperloop.
|Cost (US$ million)||Musk||MZ Low||MZ High|
|Permits & Land||1,000||117||630|
|Solar Panels & Batteries||210||290||530|
|Station & Vacuum Pumps||260||760||1,010|
For a more detailed breakdown of each line item of the passenger Hyperloop, click on the relevant header to expand:
There are two factors that will drive capsule cost:
- Number of capsules
- Cost per capsule
Musk’s initial estimates called for 40, 28-person capsules to meet an expected total demand of 14.7 million passengers per year. At $1.35 million in expected cost per capsule, this yields an overall expected cost of $54 million. However, as demand is not flat on an annual, weekly, or daily basis, an adjustment must be made to determine the capacity required to meet peak demand. Said adjustment approximates that 100 capsules will be required to meet a demand of 14.7 million passengers. Actual capsule costs are harder to predict, but if a cost-overrun safety factor of 10-20% is applied, overall expected capsule costs would increase to between $149-169 million. These two adjustments bring overall expected costs to $147-159 million.
Number of Capsules
There are three drivers of the number of capsules required:
- Transit time
- Loading/unloading time
- Peak passenger load
The slated transport time of 35 minutes for Hyperloop requires capsules experience up to 0.5g of both lateral and longitudinal acceleration which some have called similar to a roller coaster ride. Longitudinal accelerations of 0.5g should be easily handled as they fall within the range of top acceleration of medium (0.34g for RAV 4) to upper mid-market cars (0.6g for Audi S6) and are 20% of that which would be experienced on a roller-coaster.
Lateral accelerations greater than 0.2g cause nausea; therefore, Hyperloop’s 0.5g of lateral acceleration will need to be adjusted downwards. As speed of transport is the key value proposition of Hyperloop, speeds cannot be decreased materially. Thus, the only viable option to address this is altering the route of Hyperloop to increase the radius of curvature in each section by a factor of at least 2.5x (reducing lateral acceleration below the 0.2g threshold).
Assuming that the lateral acceleration issues can be addressed by adjusting the minimum radius of curvature along the route, Hyperloop will travel its stated route in 35 minutes.
Musk’s calculations for capacity of the system rely on a total of 5 minutes loading and unloading (2.5 minutes each). While 2.5 minutes may seem short (in comparison to a plane for example), the Hyperloop capsules could be designed to allow for quick loading. By having both sides of the capsule raise themselves (delorean style), passengers would simply step into their seat, stow their bag, and put on their seatbelt – all actions that would take less than 2.5 minutes.
Peak Passenger Load
Musk states in his proposal that Hyperloop will be capable of transporting up to 14.7 million people (7.35 million each way) per year. This assumes that a single capsule leaves on average every 2 minutes for all 24 hours of the day. However, as demand will not be flat throughout the day, capacity must be built for peak capacity. Musk also estimates in his alpha proposal that 70% of traffic would travel in “rush hour” and that capsules could leave up to every 30 seconds to meet this demand. Sending capsules every 30 seconds would require more than the initially proposed 40 capsules.
Some accuse 30 seconds as being too small a departure window; however, with a deceleration of ~1g (Formula 1 Drivers decelerate at a maximum of 5.4g), a non-human driver, given a fixed-location failure, it is possible; if not comfortable to stop without catastrophe.
Using comps from various municipalities, we assume that there is a +/-10% variance from mean monthly travel in throughout the year; therefore, we can assume that the maximum monthly travel for Musk’s projections would be 1.2 million passengers / month. Then assuming that 79% of travel occurs during weekdays, and that traffic can be equated with the peak hourly load during a day (which represents roughly 10% of daily weekday traffic), we find that to transport 14.7 million passenger per year, 96 capsules are required to meet the 4,072 total projected peak hourly demand, rather than the planned for 40. Traffic assessments are given below:
In order to meet Musk’s stated goal of 14.7 million passengers annually, 100 capsules should be built. This increased capsule volume will increase maximum system capacity to 36.8 million passengers annually. This suggests an average capacity utilization of Hyperloop of 40%, with 96% utilization at the peak hour. If demand is higher than expected, up to 60 additional capsules could be added before peak system capacity is reached (of one departure every 30 seconds).
Cost per Capsule
Cost per capsule is difficult to quantify as there is still much design work required. That being said, many Tesla and SpaceX engineers were involved with this project, providing comfort that costs, such as the interior, doors, capsule structure, assembly, propulsion, and batteries are closely tied to actual costs; these items represent roughly 65% of overall cost. Of the costs that may rise, items such as the compressor system (20% of cost) are based on well developed technology and are less likely to see much movement. As such, for the sake of conservatism, theoretical costs will be adjusted by applying a 25-50% riser on the remaining 35% of capsule cost. This adjustment brings per capsule cost to $1.5-1.6 million per 28-passenger capsule.
These adjustments bring the estimated capsule cost of Hyperloop from $54 million to between $147 and $159 million. Any extension of the Hyperloop would require additional capsules.
Musk and Co. estimate a tube cost of $650 million for construction and installation. Using the technical specifications provided for length (563km), diameter (2.23m) and thickness of pipe walls (20mm) and assuming that raw steel purchase is the only cost associated with this pipe, we can derive an assumed steel cost of $0.52/kg. According to the CES Materials selector this cost is at the very lowest level of the lowest grade steel that would be used. This also fails to account for the cost of shaping the steel, welding segments together, and finishing the interior.
A more realistic cost for steel would be roughly $0.60/kg for raw steel, of course the cost of buying a fully formed pipeline is more than just purchasing the base material. Pipeline costs tend to end up ranging from $1.00-1.50/kg of formed pipeline steel. The higher end of this cost is more probable, as special work would have to be done to perfect the interior of the pipe for any defects. Further, there is still the cost of welding each pipeline section together to be added to this base cost. From my conversations with those in the oil industry, these welding and installation costs for a pipe of Hyperloop’s size would likely cost ~$800,000/km and would add between $400-600 million to the project.
This revised cost estimates for pipe yields an expected cost of $2.94-4.40 million/km for steel pipe. Adding together these costs, we can assume that a more reasonable pipeline cost ranges from $1.65-2.48 billion.
Musk’s original plans called for pylons spaced at 30m intervals carrying a total cost of $2.55 billion. Others have criticize these plans, saying they wildly underestimate the cost and that based on the California High Speed Rail (CHSR) costs for viaducts would be higher by at least a factor of 10. The CHSR provides an excellent cost comparator for the reasonability of Hyperloop pylon costs as it traverses the same geological regions of California. Before directly comparing these two costs, the differences between the two structures need to be drawn out.
The CHSR reports suggest that the cost for viaducts will be $55 million/km; whereas, costs for pylons of Hyperloop are $4.53 million/km. When adjusting these costs one must account for the differences between the CHSR and Hyperloop. This is predominantly attributed to the different masses that must be supported, and that the track for Hyperloop has already been accounted for (Hyperloop pipe costs) while the track of the CHSR is a part of the viaduct cost estimate.
There are two categories of costs that makeup the Hyperloop pylon costs:
- Substructure (below ground supports)
- Superstructure (above ground supports)
As California is an earthquake fault zone, the primary design concern for the substructure (which is the subterranean support of systems) of both Hyperloop and the CHSR will be for resistance of moments. As the CHSR and Hyperloop would stand the same distance from the ground, this indicates that differences in required substructure strength would be defined by the mass of the above-ground components of the structure. Based on an assumed dimension of the CHSR track of 0.5x9x30m (per 30 m section) and a 2 metric ton/m weight for CHSR trains we can derive that the substructure for the CHSR will need to be 5.5x stronger than that of the Hyperloop (assuming the mass of Hyperloop capsules are 15,000kg, the weight of the steel tube is 1,100 kg/m). All costs of the substructure should vary linearly with mass of above-ground system.
To determine superstructure cost, the best bet is to use materials as the proxy for the overall cost, finding that the above-ground material equivalency between the two structures is 1/50.5 for CHSR/Hyperloop costs for the pylons. The reason for this vast difference is that in the CHSR, both the supports and the track are made of concrete, while in Hyperloop only the Pylons are made of concrete (the track of Hyperloop has already been accounted for by the steel pipe). While materials do not make up the entirety of the costs for Pylons, items such as labor and logistics will vary with the amount of material used. However, the comparison isn’t entirely linear, when building large civil projects such as the viaducts in the CHSR, forms must be built in place, adding labour, logistics and materials costs.
Forms are constructed along the CHSR viaducts to provide a place for concrete to be poured into. They must be built in place due to the large size of elements being used. Whereas, Hyperloop forms would be several repeated forms (to adjust for height differences between pillars) repeated for 30,000 pylons, and as such relatively small steel forms can be moved repeatedly in to place and be rapidly filled with concrete; decreasing both the cost of forms and the amount of labour required. Because of this difference in forming costs, we can assume that the cost of Hyperloop forms are essentially negligible. Thus another “forming” factor needs to be added to the cost comparison to address this significant difference. This is especially important because forming costs can be even greater than that of materials.
Cost Makeup of Systems
It is difficult to determine what percentage of the overall cost/km of the CHSR these components make up as they have not been publicly broken down (to my knowledge). Across varying civil engineering projects the ratio of these three costs for viaduct style projects can vary wildly, but we can examine what the expected cost of Hyperloop pylons would be across a range of cost makeups for the CHSR project, generating bounds on expected costs.
At the high end a reasonable mix would be sub-45%, super-45%, forming-10% and at the low end sub-35%, super-35%, forming-30% yielding a maximum cost of $5.0 million/km to a minimum of $3.8 million/km. Extrapolating out this equates to an overall cost of $2,139-2,815 million. This would suggest the Musk’s pylon cost was in the correct range for Hyperloop. Even in the most extreme (and unrealistic case), these costs yield a maximum of $5.6 billion for Pylons (100% substructure costs assumed), which while higher, is by no means a deal breaker for the overall project. These costs will decrease as tunneling increase (see Tunneling Costs breakdown to see explanation for final downwards adjustment).
Permits & Land
The cost of $1 billion that Musk assigned to land acquisition has come under scrutiny as being particularly low. The costs for land acquisitions for the CHSR provides useful guidance of what the cost of land acquisition and permitting might be. This cost is currently projected at $3.75 billion.
This cost must be taken in context however, the currently proposed Hyperloop isn’t quite comparable to the HSR for land acquisitions, there are several reasons for this:
- While CHSR is longer, it makes use of track sharing with existing rail lines in many urban areas
- Roughly 70% of the Hyperloop’s length runs along either the I-5 or I-580, where land acquisition could be negligible (given governmental buy-in)
- The CHSR travels deeper into populated areas where land costs are higher, particularly San Diego
- Right-of-way’s for the CHSR are roughly 100 ft wide; whereas, the Hyperloop right-of-way would be narrower, at 50 ft or less
Based on publicly available maps, it appears that the CHSR will need to purchase roughly 260km of rural land and 25km of urban land. Meanwhile Hyperloop will need to purchase 173km of rural land and 7km of urban land to account for deviations from highways, to keep lateral forces below 0.5g. In order to further reduce the g forces, additional deviations from the highway will be required. Assuming this doubles the required urban land purchases, and increases rural land purchases by 15% (these segments are already quite straight), total purchases will rise to 200km of rural land and 14km of urban land. Then sensitizing for a range of differences in rural and urban land pricing, the projected land acquisition costs for Hyperloop will be between $1-1.2 billion. Given the high urban land acquisition costs, if Hyperloop were faced with the possibility of large urban land purchases, tunneling through urban zones may be more cost effective (see Tunnel Construction for explanation of why final land costs will be between $117-630 million).
Musk has estimated tunneling costs of $600 million for the 24.5km that are required by the design; giving an estimated cost of $31 million/km. Exact tunneling costs are very difficult to pin down as they vary widely across different projects. The likely cost for Hyperloop (due to its small diameter) is likely in the range of $7-23 million for the excavation and reinforcement of the tunnels. Further, depending on land acquisition costs there is the opportunity to decrease overall project costs by tunneling under LA and SF urban areas (doing so would also decrease the landholder opposition that has held up CHSR).
Tunneling cost / km
This number is low relative to other rail tunneling costs (CHSR tunneling costs range from $67-90 million / km); however, expected costs for Hyperloop must be adjusted for the fact that Hyperloop will require a smaller bore area and less concrete reinforcements (due to the smaller bore size) than a traditional rail project.
To make tunnels of varying sizes comparable diameter can be used to compute a high range for cost (cost of supports will vary with diameter – change amount of material required) and area for a low range cost (boring costs will vary with amount of rock removed), giving upper and lower bounds on Hyperloop tunneling costs. Applying this adjustment yields an expected cost of between $7-23 million / km for Hyperloop. This yields an overall tunneling cost of between $171-564 million. Using data from various other tunneling projects, adjusting for station and non-tunneling costs would suggest that these tunneling costs are within a reasonable range.
CHSR provides a suitable cost comparison as it runs through similar geologic zones as Hyperloop, differences in rock characteristics (specifically hardness) are one key driver of tunneling costs. To see if a diameter/area relationship holds true across various projects of different tunneling sizes, we see what other projects imply about the expected tunneling costs for Hyperloop
Decreasing Land Purchase Costs
In order to avoid urban land purchases, and the required demolition of homes the SF approach should be built by tunneling through the entire urban segment. Tunneling costs would increase by an expected $168-552 million, but would decrease land purchases by $510-949 and pylon costs by $90-118 million. Tunneling through this area would have the added advantage of allowing for a tube without bends, allowing for increased speeds through this zone. Further, tunneling through urban areas avoids landholder opposition that would delay the Hyperloop project.
Tunneling Under SF Bay
One key cost that has not been fully addressed is the cost of traveling over the SF Bay. This cost seems to be ignored in the Alpha design yet may make up a seemingly significant portion of the cost, based on relevant comparables. Assuming that Hyperloop is tunneled under the Bay (over a 15 km span), overall costs would increase by an estimate $300-500 million. This assumes that the cost of tunneling under the bay would be above the high end of tunnels at other areas of the project.
While the construction method was sinking rather than boring, the Transbay Tube provides a good point of comparison. Completed in 1969 for $180 million, it cost roughly $1.1 billion in 2013 dollars (the cost of cement has risen with inflation since the 1970s). With a diameter of 5.2m, an equivalent cost (using a linear diameter adjustment, as per high-end estimates above) would suggest that the Hyperloop crossing would cost $478 million, a number that falls within the assumed range.
Tunneling through SF, under the SF Bay, and other tunnel portions bring overall tunneling cost to $640-1,616 million. This is above Musk’s estimates, but also decreases land acquisition and pillar costs significantly.
Propulsion & Batteries
Hyperloop is propelled by a linear induction motor that accelerates the Hyperloop capsule through the tube. This accelerator is slated to cost $140 million which will be driven by three components: materials and structural components (54%), power electronics (33%), and energy storage (13%). Not only are all elements of the linear accelerator well understood, but they are a variant of the motors used in the Tesla Model S. Musk’s team’s strong understanding of this technology means that cost overruns are unlikely. A 25% cost riser on the high end can account for this risk. Bringing the expected cost range between $140-175 million
Given current solar pricing, Musk’s estimate of $210 million to provide 21MW ($10 million/MW) of power seems an overly conservative number on a MW capacity basis, and will not provide enough power for the project given the stated capacity. Overall cost of the solar infrastructure will likely require 53MW and will carry an overall cost between $290-530 million.
Cost per MW
With a cost of $10 million/MW the Hyperloop solar system is relatively expensive. For comparison, TransCanada recently agreed to purchase 86MW of solar capacity in Ontario for $5.5 million/MW, MidAmerican Solar is building a much larger solar farm with cost/MW even lower. There are two reasons that this cost of solar might be so high:
- Batteries: The Hyperloop alpha design groups batteries with solar power, yet a more detailed look at the proposal would suggest that battery costs have been accounted for in the propulsion section
- Cost of Mounting: Hyperloop’s design calls for solar panels to be mounted on the roof of the Hyperloop, rather than making use of a solar farm, this will increase overall costs, though it’s hard to say by how much
Overall system costs could be decreased by powering Hyperloop through a traditional solar farm, which would cut cost per capacity to an estimated $5.5 million/MW.
More than 21MW of capacity will need to be installed, otherwise Hyperloop will end up capacity constrained (and be forced to purchase power from the grid, harming environmental efficiency) as it will lack suitable power to meet the adjusted system requirements. To allow for full utilization of capsules required in the system (100), the solar capacity should be increased to 53MW at a projected cost of $530 million. This will allow for full utilization of the modified installed capsule capacity (100 capsules). Further expansion of the system to peak capsule usage (160 capsules, departures every 30s) would require expansion of solar power to 84MW, which could be easily facilitated by adding an additional solar farm at some point along the Hyperloop line.
Potential Cost Savings
The initial 53MW of power could be met by only mounting panels along the top of the Hyperloop (max potential power of 57MW); however due to the available cost savings it is recommended that Hyperloop be powered by a traditional solar farm built along the line; dropping costs from $530 million to $290 million.
Station + Vacuum Pumps
The listed pricing on stations is $125 million each, for a total cost of $250 million. These costs assume that there will only be three pods in the station at any point; however, given that Musk’s original plan only called for 40% of the pods that would initially be required, and 25% of the peak design capacity station size will need to increase.
The staging zone will need to be increased to accommodate the maximum of 12 pods (required to accommodate maximum system capacity of 160 pods). Most other components of the station size will also need to be increased to accommodate peak station passenger throughput. For the sake of conservatism we can assume station costs increase roughly linearly with this capacity increase, making it safe to assume a total cost of $1 billion for the two stations. However, to reduce initial costs, the first stations built could be designed to only support staging for 8 pods, setting a low bound on station costs of $750 million.
The Hyperloop alpha report also indicated that future branch stations were a possibility, relating pod capacity to station cost, these new stations will likely cost $41.7 million per pod of capacity in the station.
With an expected cost of no more than $10 million for vacuum pumps (required to lower the pressure within the tube), even should prices increase, they remain a negligible cost, and thus I will ignore any effects they may have on overall pricing.
“Cost Margin” was a line item included in Musk’s estimate of $536 million. Perplexingly this “cost margin” is actually lower for the passenger version of Hyperloop on both an absolute and relative scale. With no explanation of the cost, it is unclear what it represents other than a plug to bring the overall cost to $6 billion (and $7.5 billion for the passenger version). While there is no rationale given for this number, as Hyperloop is still in early stages and is likely to experience cost evolution, I will assign a cost range of $300-500 million contingency value to my own cost estimate for conservatism.
A Tube to Nowhere?
The problem with Musk’s cost estimate is that the current Hyperloop only technically connects SF and LA. Critics have noted that to get from downtown LA to downtown SF, taking Hyperloop would be a longer journey than using CHSR. For a viable system to be put in place, Hypeloop’s route would need to be extended further into LA further increasing costs. This extension would increase travel times by roughly 5 minutes.
Many of the costs for Hyperloop will scale linearly; however, several design modifications would be beneficial. The 45km stretch of land purchases required would potentially yield a cost of between $1.5-3 billion dollars. Adding a significant cost to the overall project. Instead it is recommended that Hyperloop be tunneled to downtown LA, eliminating the need for land purchases and removing any additional pylon costs, for a overall tunneling cost between $315-1,035 million. Station costs would also increase by up to $100 million to allow for an underground portion for the station.
This yields the following expected cost increases to add the segment of new line from Sylmar (current proposed terminus) to Downtown Los Angeles (DT LA):
|Cost||Sylmar-DT LA Low||Sylmar-DT LA High|
|Permits & Land||0||0|
|Solar Panels & Batteries||23||42|
|Station & Vacuum Pumps||1||101|
These increases yield an overall system cost of $6.6-11.2 billion. This price would allow for a Downtown SF – downtown LA trip in 40 minutes, delivering a clear value proposition for commuters between the two cities.
Given the design capacity (assuming 160 capsules is max capacity) the maximum hourly passengers that could be transported at 100% utilization would be 6,850 passengers via Hyperloop. Given that CHSR is slated to cost $68 billion and can transport 24,000 passengers in both directions at peak capacity, Hyperloop must cost 28% or less of CHSR to be more efficient on a capital cost/peak passenger capacity basis. From this relationship it is clear that as long as Hyperloop costs less than $20 billion, it is a more efficient option on a capital cost capacity basis.
Further, by adjusting the initial design to allow for longer capsules, overall system capacity can be dramatically increased – as capacity of each pod increases by 5 passengers, overall system capacity increases by 15%. This would lead to increased capsule, station, pylon, and propulsion and power costs; however, all other costs would remain fixed. As these costs represent roughly 50% of the overall Hyperloop system, and would in most cases increase linearly (in the worst case), system costs would increase by <7.5% for every 15% increase in capacity.
Further, should an increase in capsule size not be possible, the Hyperloop line could be twinned (essentially building a new Hyperloop beside the current one) with overall system costs essentially doubling (with the potential exception of land acquisition costs).
Analysis of Hyperloop’s capital costs is only the first step in a full feasibility analysis. Clearly Musk’s cost estimates are optimistic, so the question becomes, how realistic is a $20 ticket price? I will address this question in my follow-up blog post.