By Jordan Gold and Valery Gold
The big idea: moving people without a locomotive
Every way of moving people spends energy fighting something: the friction of wheels, the weight of the engine you carry, and above all the air. NASA's drag equation says drag depends on the square of the velocity [1]. Power is force times speed, so with shape and air density fixed, the power needed to push through air grows roughly with the cube of speed (our arithmetic): double the speed and you need about eight times the power. That is why the same few ideas keep coming back. Take the air away (tubes). Take the wheel away (magnets). Take the engine off the vehicle (a pump, a flywheel, a power station somewhere else).
This is a history, not a pitch. For each idea we ask what was built, what was measured, what it cost, and where the energy came from. We bring commitments. A transport system should serve ordinary riders at a price they can afford. Energy and land are borrowed from a living planet and should be spent with care. And when someone asks the public to trust, or to fund, a new kind of train, the burden of proof is theirs. Those values shape our questions, not the answers. Where the record is thin, we say so.
How to read the labels. Measured means observed in operation. Test record means a timed or recorded test. Design target means what a design is meant to do. Model means a simulation or calculation. Company claim means a promoter's statement nobody independent has checked. Our arithmetic means we divided or converted figures printed in the source. All sources were fetched on 14 November 2026 (US Central time).
Pushing air through a pipe: the atmospheric railways (1810s-1850s)
Medhurst, Vallance, Clegg and Samuda: the idea and the piston
An 1845 London patent journal records George Medhurst's compressed-air plan of 1810 and 1812 [2] and Vallance's 1827 plan to push carriages through a large cylinder by air pressure on a disc [2].
Everyone hit the same stumbling block: a piston in a pipe must be tied to a carriage outside it, so the pipe needs a slot, and the slot must be sealed. Samuda said closing it "had been the stumbling-block, in all the preceding inventions" [2]. The answer was a valve along the whole line, a leather strip clamped between iron plates over the slit, as a German journal described the Clegg and Samuda system in 1840, translating an English account [3]. Joseph Samuda's 1841 treatise states the layout: "the moving power is communicated to the train by means of a continuous pipe or main A, laid between the rails" [4]. These are inventors describing their own designs: design targets.
Dalkey and the argument over measurement
The first commercial line was the short Kingstown to Dalkey route near Dublin. A heritage page (secondary) dates its opening to July 1844 [5]. Brunel made "several experiments upon the Dalkey railway" in 1843-44 [6].
The best primary document is Robert Stephenson's 1844 report. He built locomotives and was a rival of the system. He tested twenty trains [7] and concluded that power lost on the Euston rope incline was 45 per cent against 74 per cent at Dalkey [7]: a test record from a hostile expert. Because a quarter-second timing error shifts the calculated speed by 4 miles per hour at about 36, he averaged several intervals [7]. A pamphlet by the promoter-side Sir George Berkley says Samuda "contended that erroneous deductions had been drawn from the results" [8]. We read this as an unsettled measurement.
Brunel's South Devon line (1847-48)
Brunel concluded that the mechanical difficulties "may be overcome" and that "high speeds may be more easily attained" [6]. On the South Devon line, atmospheric trains began taking a share of passenger work on 8 September, four each way daily [6]; the year, 1847, comes from his report of 27 August in the same passage [6]. Riders noticed: the memoir calls the motion "singularly smooth and agreeable" [6].
The speeds were real. A footnote reports, from dynamometric experiments, "The highest speed recorded was 68 miles per hour" with a 28-ton train [6]. That is about 109 km/h (our arithmetic): a test record in an 1870 memoir by his son, with no year for the run. The physics worked.
Why it failed
The money and machinery did not. Brunel's report of 19 August 1848 says the engines "have not, on the whole, proved successful; none of them have as yet worked very economically" [6]. The memoir does the arithmetic: Brunel "had a right to expect" 300 horse-power during a train's passage, yet it "actually required 865 horse-power, or nearly three times the amount" [6].
The seal was the other half. "A considerable extent of longitudinal valve failed by the tearing of the leather, at the joints between the plates" [6]; the leather "seems to be destroyed—it is black, and has evidently been acted upon by the iron of the plates" [6]. A popular story blames melting grease; Brunel's report says drying and shrivelling of the leather, "and not the melting of the composition, is the principal inconvenience resulting from heat" [6]. A railway museum dates the end of atmospheric working between Exeter and Newton to 10 September 1848 [9]. The energy came from stationary steam engines of about 236 indicated horse-power at each pumping station [6]. Moving the engine off the train did not make the energy free; it made the leaks expensive.
Gyroscopes that balance: monorails, cars and one-wheel vehicles (1900s-today)
Brennan's gyro monorail
Louis Brennan's US patent 796,893 was published on 8 August 1905 [10]. It proposes to "employ two similar gyrostats rotating in opposite directions" [10]. A spinning wheel resists being tipped, and its response can push a vehicle upright, so a car can balance on one rail.
Scientific American, reporting in June 1907 on a Royal Society demonstration, noted that "The model locomotive exhibited before the Society was only six feet long" and said the system "cannot be regarded as having demonstrated its practical value, at least as applied to heavy trains" [11]. A Science Museum Group record dates Brennan's working model to about 1907 [12] and a full-size car to 1909, 40 feet long and 22 tons [12]; at the White City exhibition in 1910 "fifty passengers were driven around a circular track" [12]. It concludes that "investers were not convinced and the Mono-Rail was not adopted" [12]. The 1909 and 1910 dates rest on that museum record alone.
Scherl and Schilowsky, and why the gyro car stayed a curiosity
The Electric Railway Journal reported that the Scherl monorail car made its first test runs on 10 January 1910 at a Brooklyn skating rink [13], at "about 4 m.p.h. on straight track and 3 m.p.h. on curves" [13]. Trouble was "said to be due to a leak in the partial vacuum in which the gyroscopes revolved" [13]. A test record: tiny speeds and a seal problem.
Peter Schilowsky's US patent (Google lists a prior-art date of 1 March 1910; published 27 April 1915) [14] says "it becomes possible to obtain lateral stability of the vehicle frame with a single gyroscope" [14]: a design target. Scientific American's January 1911 article refers to London demonstrations "with the small model which he had constructed" [15]. We found no source showing a full-size Schilowsky car carrying passengers.
Why did the gyro car stay a curiosity? No source we opened says, so this is our reading: a machine that must keep a heavy rotor spinning, in a sealed chamber, to stay upright has a failure mode that a railway with two rails never had.
Monowheels, Dynasphere, Gyro-X and Lit Motors C-1
The dream moved from rails to roads. Not all of it was one wheel, and not all of it was built.
| Vehicle | What the record shows | Label |
|---|---|---|
| Dynasphere (1930s), rider inside a large wheel | A 1932 magazine says a two-and-a-half horsepower petrol motor was enough to run the thousand-pound wheel [16]. An archivist says the inventor "reportedly claimed" a high-speed future and that it "was not a commercial success" [17] | Demonstration reported; claim reported |
| Gyro-X (1967), gyroscopes, two wheels | A museum calls it a prototype built in 1967 [18], lists "Top Speed: 125 mph (claimed)" [18] and says witnesses recount instability above 70 mph [18] | Prototype; speed is a claim |
| Lit Motors C-1 (2010s-), gyroscopes, two wheels | A 2014 report says the video's test mule "had 'training wheels'" [19] and a 200-mile range is "expected" in production trim [19] | Company claim; demo with outriggers |
We give no Dynasphere speed. Lit Motors' newsletter says it is "poised to integrate the gyros, controls, and steering into one vehicle" [20] and claims "TRL 3", meaning "the core physics is proven in real hardware" [20]. Those are claims.
Flywheels and compressed air: energy you carry on board
The Swiss Gyrobus (1950s)
If a gyro can hold a car up, a flywheel can hold the energy to move it. A retrospective history by the FBW marque club says the mid-1950s Oerlikon standard flywheel could store "gut 5 kWh" (a good 5 kWh) of usable energy [21], spinning to 3,000 revolutions per minute [21], with charging taking about two minutes [21]. These are design figures. Service at Yverdon opened on 30 September 1953 [21]. The engineers designed it for a maximum of 50 km/h on the flat [21]: a design speed, not a recorded one.
In Léopoldville, a quoted Oerlikon chief designer recalls a top speed of 60 km/h and a mean consumption of 1.6 kWh per vehicle-km [21]: a manufacturer's undated recollection. For scale, a battery-bus fleet at Foothill Transit used 2.15 kWh per mile, about 1.34 kWh per km (our arithmetic) [22], per the National Laboratory of the Rockies (formerly NREL) [23]: different vehicle, era and method.
The workshop memory: "We regularly had to tow the two buses to Oerlikon on a truck" [21]. The Yverdon pair was replaced by three diesel vehicles in 1960 after about 720,000 km in total [21].
Mekarski compressed-air trams and the AirPod
Compressed air cools as it expands, which can freeze an engine. Louis Mekarski's US patent, published on 23 May 1876 [24], answers by "passing the compressed air through a body of heated water" [24]; an 1881 Scientific American Supplement describes a boiler at about 120° Cent. before the cylinder [25]. These are design descriptions. We give no efficiency, because we found no traceable source for the popular figures.
The idea came back as a car. Tata Motors signed a 2007 licence with France's Motor Development International to make compressed-air cars in India [26]; that is a company plan, and we found no evidence of series production. A peer-reviewed 2009 model found that, "even under highly optimistic assumptions", the compressed-air car "is significantly less efficient than a battery electric vehicle" [27]. That is a model, and it is the only analysis we found.
Tubes, vacuum and magnetism: the vactrain lineage
Goddard to Salter
Robert Goddard's US patent 2,511,979, "Vacuum Tube Transportation System", was filed on 21 May 1945 [28] and granted on 20 June 1950 [28]. It describes preventing friction "by non-magnetic means, such as the application of fluid pressure between relatively movable parts" [28] and refers to an earlier application of 18 November 1944, now abandoned [28]. We could not retrieve that application, so we report only the 1950 patent.
In August 1972 R. M. Salter's RAND paper proposed "tubecraft" that ride on electromagnetic waves "much as a surfboard rides the ocean's waves" [29], at speeds "on the order of thousands of miles per hour on the long-haul links" [29]: design targets in a think-tank paper. Its cost line is not an estimate. It says "we can afford a $90 billion cost for the central corridor part of the system", which "works out to be about $30 million/mile of double tube system" [29], on an assumed fare of about $50 each way [29]. That is an affordability ceiling. The paper states no price year, so we call the figures nominal and do not adjust them.
Hyperloop Alpha (2013)
The 2013 Hyperloop Alpha white paper was released by Elon Musk on 12 August 2013 [30][31]. That credits the release, not authorship; the PDF prints no byline. Its Los Angeles to San Francisco figures are design targets, not results. It proposes capsules at "760 mph (1,220 kph, Mach 0.99 at 68 ºF or 20 ºC)" [32], a trip time of "35 minutes from county line to county line" [32], a cost "under $6 billion USD for two one-way tubes and 40 capsules" [32] and a ticket at "$20 USD plus operating costs" [32]. Its energy claim appears only in a chart with no printed values [32], so we leave it out of our comparison. It lists "sub-scale testing" as a next step [32].
What was actually built and run
The nearest thing to a built system was a 500-metre test track. The BBC reported that two passengers, both company staff, "travelled the length of a 500m test track in 15 seconds, reaching 107mph (172km/h)" [33]. Reuters, via the Guardian, put it more carefully: "the company said" so [34]. A test record on a short track, a fraction of the white paper's design speed.
In a release of 18 December 2017 the company said it had set "a historic test speed record of nearly 387 kilometers per hour" [35]. That is a company claim. We found no independent timing, so we do not call it a record or plot it.
Setbacks and the 2023 shutdown
On 21 December 2023 TechCrunch reported that Hyperloop One "will cease operations on December 31 according to Bloomberg News" [36], having "pivoted its focus to cargo in early 2022" [36]. Fortune, citing PitchBook, said it "raised more than $450 million since its founding in 2014" [37]. Engadget, relaying Bloomberg, said it had "notably never secured a contract to build a working hyperloop system" [38]. That is nearly a decade of private capital, one short test track and no contract. Anyone asked to fund such a venture with public money should demand that answer first.
Magnetic levitation in service
EMS or EDS: two ways to float
Maglev trains hover by one of two methods. A 2018 comparison describes Germany's Transrapid as an "electromagnetic system - supporting and guiding systems based on attraction" [39] and Japan's L0 as using "superconducting magnets in the vehicle" [39]. Linimo uses a "Normal conducting magnetic levitation type / linear induction motor propulsion system" [40]. That paper's authors work for maglev organisations and its figures are "based mainly on simulation results" [39].
Shanghai, Linimo, Changsha, Incheon
Shanghai. Engineering News-Record reported 30 km of guideway [41] as an "estimated $1.2-billion project" [41]. The operator gives a maximum speed of 430 km/h [42]; Guinness says 431 km/h on each run [43]. A government portal's 2018 timetable shows 430 km/h only in two daily windows and 300 km/h otherwise [44], and the operator's chronology records the off-peak limit of 300 km/h from 1 December 2005 [45]. So we write a maximum service speed of 430 km/h at set hours. A Japanese trade site reports that all service was cut to 300 km/h from 1 May 2020 [46]; that is secondary, and we make no claim about any later change.
Linimo. The Aichi line is 8.9 km long [40] with a maximum speed of "Approx. 100 km per hour" [40]. We found no construction cost, so we give none.
Changsha. Hunan's transport authority said the line opened for trial operation on 6 May 2016, runs 18.55 km at up to 100 km/h (our translation) [47]. It gives total investment as 4.29 billion yuan [47]; a project page from a company involved gives 4.604 billion [48] and 248 million yuan per kilometre [48]. The sources disagree, so we show both.
Incheon. The airport line is the cautionary case, and the public paid. A 2018 report says it opened in February 2016 and its utilization rate was 16% [49]; the first year brought 880,000 riders against 4.52 million expected [49]. It ran free [49]. The 450 billion won construction cost was split among national government (350 billion won), the airport corporation (79 billion) and the city (19 billion) [49]. The city says service stopped in July 2022 and resumed on 17 October 2025, three years and three months later [50]. That is no argument against magnets, only for asking about ridership before building.
Chuo Shinkansen: costs up, opening date open
Japan's superconducting maglev from Shinagawa to Nagoya is the biggest project under way. Central Japan Railway (JR Central) says the total was "initially estimated at 5.52 trillion yen" [51], rose to 7.04 trillion yen when announced in April 2021 [51], and is now forecast at 11.0 trillion yen on 29 October 2025 [51]. That is roughly double the first estimate (our arithmetic). The forecast "includes the cost for rolling stock and excludes the expenses already spent on the Yamanashi Maglev Line" [51]. All three are estimates, not final spending. We show no per-kilometre figure, because JR Central's notice gives no line length and the only length we found is in the 2018 comparison paper.
JR Central says it will fund the work "mainly through operating cash flow" and financing [51]. Its 2026 fact sheet says "we cannot predict a new opening date" [52]. We state no opening year, because the operator does not.
Speed over time
Eleven dated speeds passed our checks.
The oldest has a contemporary source. Dingler's Polytechnisches Journal says the AEG "A" car reached 210.2 km/h on 28 October 1903 [53]; a railway-statistics chronology agrees [54]. Dingler says the Siemens car reached 201 km/h on 2 October [53]; Siemens' own history gives 6 October [55]. The sources disagree on that date, and we do not resolve it.
The rest: the LNER A4 Mallard, 126 mph (about 203 km/h, our conversion) in July 1938 [56]; the Tokaido Shinkansen at 210 km/h in revenue service from 1964 [57]; a TGV test at 515.3 km/h in 1990 [58]; Transrapid 07 at 450 km/h in 1993 [59]; Shanghai at a 430 km/h maximum service speed, plotted at 2004 (full-day public trial operation began 29 January, operator chronology) [42][45]; Linimo at 100 km/h [40], plotted at 2005 (the operator lists 6 March) [40]; the TGV V150 at 574.8 km/h in 2007 [58]; the L0 at 603 km/h in 2015 [60]; Changsha at 100 km/h in 2016 [47]; and the 172 km/h pod test in 2020 [33]. Six are test records, four are operating speeds as the operators state them and one is a short-track test. Shanghai and Changsha are dated to the start of trial operation, which is how the timeline labels them.
What does not appear: the Gyrobus's 50 km/h (a design speed), Hyperloop Alpha's 1,220 km/h and Salter's thousands of miles an hour (design targets), Virgin's 387 km/h (a company claim), and Brunel's 109 km/h (no year given). A straight line on a log axis through the eleven would rise about 0.2% a year, with a 95% interval from -1.1% to 1.6% and an R-squared of 0.014 (our calculation). That is no trend: it mixes records with service speeds, and a 100 km/h urban shuttle with a 603 km/h test.
What it costs per kilometre
The best audited numbers come from the European Court of Auditors (ECA), which reviewed ten high-speed rail lines. It reports that "on average, the lines we audited cost 25 million euro per km (not taking into account the more expensive tunnelling projects)" [61] and that the latest two lines are "above 40 million euro per km" [61]. Dividing Table 4's total cost by its total length gives 23.6 (our arithmetic; Table 4 counts a 133 km overlap twice [61]). For Madrid-Galicia we use ECA's own 13.7 million euro per km, which excludes the overlap [61]. ECA says figures "are expressed in nominal terms" [61] and prints no price base year, so we assume none and adjust nothing.
Maglev cannot go on the same axis. Shanghai's $1.2 billion over 30 km is about 40 million US dollars per km (our arithmetic; an estimate at opening). Changsha is about 248 million yuan per km on the project page's figure, or about 231 million on the official total (our arithmetic). Incheon is about 73.8 billion won per km (our arithmetic from 450 billion won over 6.1 km [49]). Those are three currencies in three years; we convert none.
The energy question
Energy per passenger-kilometre
The only broad, tabulated comparison we could verify is the US Transportation Energy Data Book, Edition 41, for 2023, which counts "only end-use energy ... for electricity", with no generation losses [62]. Converted from Btu per passenger-mile (our arithmetic), in megajoules per passenger-km: car 1.83 [62], transit bus 3.68 [62], domestic air 1.47 (domestic service only) [62], Amtrak 1.02 [62], rail transit 0.74 [62] and commuter rail 1.68 [62]. Its authors warn that "it is not possible to obtain truly comparable national energy intensities among modes" [62]. An e-bike literature review gives 1.5 plus or minus 0.5 kWh per 100 km at tank-to-wheel [63]. That is per vehicle-km and the review does not count riders [63], so our 0.05 MJ per passenger-km assumes one rider.
We looked for five or more independent sources per mode and found one per mode, so the chart draws dots, not boxes. Maglev, high-speed rail and hyperloop are excluded because the numbers that exist are per seat. In this dataset rail transit uses about a fifth as much energy per passenger-km as a transit bus (0.74 against 3.68), partly because it carries 16.4 people per vehicle against 6.0 [62]. Energy per passenger mostly reflects how full the vehicle is.
Maglev vs wheel-on-rail, and what a vacuum might save
A 2018 simulation-based comparison puts Transrapid at 45 Wh per seat-km at 330 km/h on Hamburg-Berlin against 59 for the ICE 3, and says Transrapid at 430 km/h uses just 7% more than the ICE at 330, at 63 [39]. Those are model values, per seat, from authors with maglev affiliations. The one measurement we found comes from the Shanghai team's 2004 paper: 76 Wh per seat-km at 430 km/h at propulsion level (70 at converter output, 80.5 at the 20 kV side) [64], for a five-section, 462-seat train [64], on runs in November 2003. It is per seat, not per passenger, measured by the developers, and read through a Wayback Machine copy because the original server returns an error. We do not pool it with the per-passenger data. Seat figures flatter an empty train.
For vacuum tubes, a Delft-based paper models "0.1 for 200 km, 0.036 for 600 km, and 0.025 for 1000 km distance between stops", in kWh per seat-km [65]. That is a model for a system that does not operate.
From energy source to wheel
Between the energy you buy and the wheel there are losses. The US Department of Energy says "about 30% for a conventional gasoline vehicle, depending on the drive cycle" [66], and that for an electric vehicle energy to the wheels "is 65% - 69%", before regenerative braking, on the EPA combined cycle [66]. That is tank-to-wheel and battery-to-wheel only. It ignores how the gasoline was refined or the electricity generated, and the two figures are worded differently in DOE's text.
We have no consistent, verified source for well-to-wheel or hydrogen pathways, so we show none.
Hydrogen trains in practice (Coradia iLint)
Lower Saxony's Alstom Coradia iLint fleet is the hydrogen-train case we could document. A joint Alstom and LNVG release of 24 August 2022 says the 14 fuel-cell vehicles belong to LNVG, the state's transport authority [67]; the operator evb said they would replace 15 diesel trains [68]. The operator evb headlined the launch a "Weltpremiere" and called it "das weltweit erste Netz mit Wasserstoffzügen im Passagierbetrieb" (the world's first network of hydrogen trains in passenger service) [68]; that is the operator's own claim. How many were running at the start is where the sources part.
The release says "Derzeit rollen fünf der neuen Züge" (five of the new trains are currently rolling) [67]. LNVG's slides of 6 November 2023 say "nur sechs Fahrzeuge ausgeliefert" (only six vehicles delivered) [69] and show "2 Fahrzeuge Betrieb" (two vehicles in operation) in August 2022 [69], growing to twelve by December 2023 [69]. Five and two may count different things, and we do not know how either document defines them, so we resolve neither.
On fuel, the same slides give "rechnerisch ca. 22 kg pro 100 km" (calculated, about 22 kg per 100 km) [69] against a "nachgewiesener Verbrauch bei bis zu 28 kg pro 100 km" (proven consumption of up to 28 kg per 100 km) [69]: a design target and an upper figure from operation. On 6 October 2026 evb said that, of 14 hydrogen vehicles originally planned [70], on average only half the fleet is available for daily service [70], and that LNVG's diesel replacement vehicles "cannot fully make up" the shortfall [70]. In January 2025 another operator, RMV, said line RB12 "continues to be served with hydrogen trains" [71]. This is no verdict on hydrogen trains in general, only one fleet running below plan in October 2026.
Where the power comes from
Every idea here needs energy from somewhere; promoters skip where. Electric traction itself is old. Siemens ran a small electric train on a roughly 300-metre loop at a Berlin exhibition, opened 31 May 1879 [72], at 150 volts direct current [73], with three carriages [73]; we give no speed, because no primary source we found states one, and the passenger count is Siemens' own. Frank Sprague's Richmond street railway began operating on 2 February 1888 [74]; Electrical World that month counted about a dozen cars running and forty more under way, on about twelve and a half miles of track [75].
Efficiency claims need their boundary. The US Department of Energy puts 60%-65% of wall energy on the road for an electric car, before regenerative braking [76] (a different boundary from the 65–69% battery-to-wheel figure in the chart above). It also says a typical EV is "87%-91% efficient (after taking regenerative braking into account)", counting from the energy stored on board [66]. All three are different boundaries. For electric modes, "grid electricity" is the whole question: the EIA says about 58% of US utility-scale generation in 2025 came from fossil fuels, about 18% from nuclear and about 24% from renewable sources [77].
Solar is the place to be most skeptical. Australia's Byron Bay train carries 6.5 kW of roof panels by the operator's account [78] and 6.6 kW by the Clean Energy Regulator's [79]. The operator says 23% of its shed-roof array's output feeds the train, and the rest goes to the grid [78], so "solar-powered" is a net-energy claim. Riding Sunbeams' Aldershot installation is about 100 panels, "just under 40kW" per the project [80] and 30 kWp per its consultant, and it powers signalling and lights, not trains [81]. Indian Railways reported a 1.7 MW plant at Bina "under extensive testing" in July 2020 [82] and says it has "tied up" 4,260 MW of solar [83]; those are government or contracted figures, not delivered traction energy. Nuna 3, a racing car, averaged 102.75 km/h over 3,021 km in 2005 [84]. Aptera's "up to 40 miles per day" of solar driving is a company claim [85].
What the history tells us
This section is opinion.
The ideas that need new infrastructure and near-perfect seals, rotors or tunnels have a poor record in what we could document. The atmospheric railway fell to a leather valve and pumping costs. The Scherl car's first test runs were dogged by a reported leak in its vacuum. The Gyrobus regularly had to be towed back to the factory on a truck. The Hyperloop company ended with one short track and no contract. Incheon, built with public money, drew 880,000 riders in its first year against 4.52 million expected. The ideas that fit existing rights of way (rail transit, battery buses, e-bikes) are the ones this evidence shows in routine, measured use.
That is a pattern in this evidence, not a law. JR Central is funding its line mainly from operating cash flow and financing, its estimate has roughly doubled, and maglev may yet prove itself. But the habit this history teaches suits riders and the public purse: ask for measured results on an operating line before trust or money is given, ask who carries the risk, and ask who gets to ride. The numbers nobody has are telling: a verified cost per kilometre for any maglev line, and a measured per-passenger energy figure for any of them.
Methods, limits and corrections
Cutoffs. Every cited source was fetched on 14 November 2026 (US Central), and each quote was found in the fetched copy. US energy intensities are 2023; the ECA data are from a 2018 audit; Shanghai's energy measurement is from 2003-04; the iLint status runs to October 2026.
Labels and boundaries. We keep tested or measured results apart from design targets and company claims, and do not compare per-seat with per-passenger figures, tank-to-wheel with well-to-wheel, or design with measured values without saying so.
Charts. The energy chart has one source per mode, so it shows dots. The speed chart is descriptive only. Cost is nominal euros for audited European lines, with maglev not pooled. The tank-to-wheel chart is two pathways from one source.
Known gaps. We could not open Brennan's own demonstration records or any record of a full-size Schilowsky car. We have no Linimo cost, no confirmed Chuo line length, no verified hydrogen or well-to-wheel efficiency, and no exchange rates for the maglev costs. Several sources are secondary or retrospective and are labelled. Wayback Machine copies were used for the Shanghai paper, the operator's page and the Hyperloop Alpha PDF because the original servers refused or failed.
Corrections. If you see an error, please use the corrections form on our site.
Sources
Numbers are in order of first citation. Every link was fetched on 14 November 2026 (US Central), and each quoted passage was checked against the fetched copy.
- NASA Glenn Research Center, "Drag Equation" (Beginner's Guide to Aeronautics). link.
- "Scientific Notices," The London Journal of Arts, Sciences and Manufactures, and Repertory of Patent Inventions (1845); Internet Archive scan. link. Archive.org OCR scan; physical original not compared.
- "Beschreibung der atmosphaerischen Eisenbahn der HHrn. Clegg und Samuda," Dingler's Polytechnisches Journal 77 (1840), translated from the Civil Engineer and Architect's Journal, Aug. 1840. link.
- Joseph d'Aguilar Samuda, A Treatise on the Adaptation of Atmospheric Pressure to the Purposes of Locomotion on Railways (1841); Internet Archive scan. link. Archive.org OCR scan.
- Ingenious Ireland, "The last journey on Dublin's atmospheric railway" (2013). link. Secondary (heritage and tourism site).
- Isambard Brunel, The Life of Isambard Kingdom Brunel, Civil Engineer (1870); Project Gutenberg. link. Biography by his son, reproducing the engineer's own testimony and reports.
- Robert Stephenson, Report on the Atmospheric Railway System (J. Weale, 1844); Internet Archive scan. link. Written by a rival locomotive engineer; Archive.org OCR scan.
- Sir George Berkley, The Peculiar Features of the Atmospheric Railway System (1845); Internet Archive scan. link. Pamphlet by a promoter-side author; Archive.org OCR scan.
- Didcot Railway Centre, "Brunel's Atmospheric Pipe." link.
- L. Brennan, "Means for Imparting Stability to Unstable Bodies," US Patent 796,893 (published 8 Aug. 1905), Google Patents. link.
- "The Brennan Gyroscope Monorail," Scientific American 96(22), 1 June 1907; Internet Archive scan. link. Archive.org OCR scan.
- Science Museum Group Collection, "Brennan's gyroscopic mono-rail car" (object co27065). link. Museum catalogue record; the 1909 and 1910 claims rest on this record alone.
- "The Scherl Gyroscope Car," Electric Railway Journal 35 (15 Jan. 1910), p. 116; Internet Archive scan. link. Archive.org OCR scan.
- P. Schilowsky, US Patent 1,137,234 (filed 1 Mar. 1910, published 27 Apr. 1915), Google Patents. link.
- "The Schilowsky Gyroscopic Car," Scientific American 104(4), 28 Jan. 1911; Internet Archive scan. link. Sentence confirmed against the page image.
- "Rolling Hoop Carries Passengers," Popular Science, May 1932; Internet Archive scan. link. Archive.org OCR scan.
- IET Archives blog, "The Dynasphere" (15 Nov. 2022). link. Secondary (archivist's blog).
- Lane Motor Museum, "Gyro-X (1967)." link. Museum catalogue entry.
- Green Car Reports, "Lit Motors C-1 video shows self-balancing motorcycle on the road" (7 Nov. 2014). link.
- Lit Motors, "July Newsletter" (company blog post). link. Company statement.
- FBW Club, "Geschichte - Der Gyrobus." link. Retrospective club history; German, our translation.
- Foothill Transit Battery Electric Bus Evaluation: Final Report, National Laboratory of the Rockies (formerly NREL), 2021. link.
- U.S. Department of Energy, "Energy Department Renames NREL as National Laboratory of the Rockies" (1 Dec. 2025). link.
- L. Mekarski, US Patent 177,736 (published 23 May 1876), Google Patents. link.
- "Compressed Air Locomotives," Scientific American Supplement no. 275 (9 Apr. 1881); Project Gutenberg. link.
- Autocar Professional, "Tata Motors air-powered car project" (14 Feb. 2017). link. Trade press relaying company statements.
- Creutzig, Papson, Schipper, Kammen (2009), Environmental Research Letters 4, 044011 (abstract). link. Peer-reviewed model.
- R. H. Goddard, "Vacuum Tube Transportation System," US Patent 2,511,979 (filed 21 May 1945; granted 20 June 1950). link. Copy hosted by the Los Angeles Metro library.
- R. M. Salter, The Very High Speed Transit System, RAND Paper P-4874 (Aug. 1972). link. Scanned PDF; figures confirmed on the page image. A plain request returns 403; a browser user-agent receives the PDF.
- Casey Johnston, "Hyperloop—a theoretical 760 mph transit system made of sun, air, and magnets," Ars Technica (Aug. 2013). link. Contemporary trade press; dated by URL (Aug. 2013), day confirmed by the WSJ item.
- Brian R. Fitzgerald, "Here's Elon Musk's Hyperloop Alpha (With Images)," Wall Street Journal, Digits blog (12 Aug. 2013). link. Contemporary news blog.
- Hyperloop Alpha white paper (August 2013); no printed byline. Archived PDF copy (Wayback Machine capture of 9 March 2014). link. Archive copy; the paper prints no byline and its figures are design targets.
- BBC News, "Virgin Hyperloop: First passengers ride levitating pod" (9 Nov. 2020). link.
- "First passengers travel in Virgin's levitating Hyperloop pod," The Guardian (9 Nov. 2020, Reuters). link.
- Virgin Hyperloop One, "Virgin Hyperloop One names Richard Branson chairman, closes $50 million financing and sets new speed record" (18 Dec. 2017). link. Company release; no independent timing record found.
- TechCrunch, "Hyperloop One is shutting down" (21 Dec. 2023). link.
- Fortune, "Hyperloop One shut down" (21 Dec. 2023). link.
- Engadget, "Hyperloop One is shutting down" (Dec. 2023). link.
- Fritz et al., "Energy consumption of track-based high-speed trains: maglev vs wheel-rail," Transportation Systems and Technology 4(3 Suppl. 1), 2018. link. Authors affiliated with maglev organisations; values mostly from simulation.
- Aichi Rapid Transit, Linimo company outline. link.
- "Shanghai's Express Train," Engineering News-Record (16 Feb. 2004). link.
- Shanghai Maglev Transportation Development Co., "Brief Account of Maglev Train" (operator English page; Wayback capture). link. Operator marketing page, undated; archive copy.
- Guinness World Records, "Fastest maglev train in regular public service." link. Record keeper relaying operator data.
- Pudong government English portal, "By maglev" (China Daily item, updated 1 Sep. 2018). link. Government host republishing a newspaper item.
- Shanghai Maglev Transportation Development Co., chronology of major events (operator Chinese page; Wayback capture). link. Archive copy; Chinese, our translation.
- TRAICY, report of 31 May 2020 on the Shanghai Maglev speed reduction. link. Secondary report of an operator statement; Japanese, our translation.
- Changsha Municipal Transport Bureau via Hunan Transport Department, "Changsha Maglev Express officially opens for passenger trial operation" (9 May 2016). link. Chinese; our translation.
- Jumbo (project case page), Hunan Changsha maglev PPP project. link. Page of a company involved in the project; Chinese, our translation.
- Kyunghyang Shinmun, "4500억 인천공항 자기부상열차 무료에도 이용률은 16%" (8 Oct. 2018). link. Korean; our translation.
- Incheon Metropolitan City press release, 17 Oct. 2025 (airport maglev reopening). link. Korean; our translation.
- Central Japan Railway Company, notice on total construction costs of the Chuo Shinkansen, Shinagawa-Nagoya section (29 Oct. 2025). link.
- Central Japan Railway Company, Chuo Shinkansen fact sheet (July 2026). link.
- Dingler's Polytechnisches Journal 319 (1904), p. 483, on the Marienfelde-Zossen high-speed trials. link. Contemporary journal.
- bahnstatistik.de, chronology of the Berlin military railway (KED Berlin). link. Secondary compilation.
- Siemens Historical Institute, Lifelines 7: Walter Reichel. link. Company history.
- Science Museum Group blog, "Mallard 75." link.
- Japan Railway & Transport Review 48, "Increasing Shinkansen speeds." link.
- Alstom press release, "574,8 km/h: new world rail speed record" (3 Apr. 2007). link.
- University of Washington, Transrapid chronology. link. Secondary compilation.
- Guinness World Records, "Fastest maglev train." link. Secondary compiler.
- European Court of Auditors, Special Report 19/2018, "A European high-speed rail network: not a reality but an ineffective patchwork." link.
- Oak Ridge National Laboratory, Transportation Energy Data Book: Edition 41 (2025), Table 2.13. link.
- Weiss et al. (2015), review of electric two-wheelers, Transportation Research Part D 41, Appendix Table A1 (Europe PMC full text). link. Mirror of the publisher text.
- Shi, Yang, Hu, Li, "The Research on Power Test of Shanghai High-Speed Maglev Transportation," Maglev 2004 proceedings, Table 2 (Internet Archive copy; original server returns an error). link. Written by the system's own team; archive copy.
- van Goeverden, Janic, Milakis (2018), conference abstract on Hyperloop energy use, TU Delft repository. link. Model, not measurement.
- U.S. Department of Energy, Fact of the Week #1360 (16 Sep. 2024). link.
- Alstom and LNVG joint release, 24 Aug. 2022 ("Weltpremiere: erstes Netz mit 14 Wasserstoffzuegen"). link. German, our translation.
- evb Elbe-Weser, "Weltpremiere: Erste Wasserstoffzuege fahren in Bremervoerde" (24 Aug. 2022). link. German, our translation.
- Nawrocki (LNVG), presentation on hydrogen trains in Lower Saxony, 6 Nov. 2023. link. German, our translation.
- evb Elbe-Weser, notice on the RB33 line and missing railcars (6 Oct. 2026). link. German, our translation.
- Rhein-Main-Verkehrsverbund press release, 9 Jan. 2025. link. German, our translation.
- berliner-bahnen.de, "Siemens-Ausstellungsbahn" timeline. link. Enthusiast reference site; German, our translation.
- Siemens Historical Institute, "In Focus": the 1879 electric railway (Wayback capture, 2016). link. Company history; archive copy.
- Engineering and Technology History Wiki, "Milestones: Richmond Union Passenger Railway, 1888" (IEEE Milestone). link.
- "The Sprague Electric Railway at Richmond, Va.," The Electrical World 11(7), 18 Feb. 1888; Internet Archive scan. link. Archive.org OCR scan.
- U.S. Department of Energy, Fact of the Week #1045 (3 Sep. 2018). link.
- U.S. Energy Information Administration, "Electricity in the U.S." (generation, capacity and sales). link.
- Byron Bay Railroad Company, "Sustainability." link. Operator statement; rated, not metered, output.
- Clean Energy Regulator (Australia), case study "Byron Bay solar powered train." link.
- Riding Sunbeams, "First Light: Aldershot." link. Project originator; rated output.
- Ricardo, press release on the Aldershot solar-connected railway (28 Aug. 2019). link. Project consultant press release.
- Press Information Bureau, Government of India, Ministry of Railways release (13 Jul. 2020). link. Government statement; installation status as of July 2020.
- Press Information Bureau, Government of India, Ministry of Railways release (24 Feb. 2025). link. Government statement of contracted capacity.
- Guinness World Records, "Fastest average speed in the World Solar Challenge" (Nuna 3, 2005). link. Race result by a university team.
- Aptera Motors, vehicle page. link. Company marketing claim.
Timeline-only sources (events shown on the timeline that the text does not discuss): Clegg valve patent, 3 Jan. 1839 (grant list only; the specification is not verified): The London Journal of Arts and Sciences, vol. 13 (1839), list of patents sealed in England; Internet Archive scan. link Samuda patent, 30 Apr. 1844: "Verbesserungen in der Construction und Anordnung ... atmosphaerischer Eisenbahnen," Dingler's Polytechnisches Journal 95 (1845), on the Samuda patent of 30 April 1844. link Gyrobus trials at Flüelen–Altdorf, 1950: Schweizerische Bauzeitung 68(38), 23 Sep. 1950, p. 533, "Probebetrieb mit Gyrobus auf der Strecke Flüelen-Altdorf" (ETH Library e-periodica). link