On the first of October 2026 a SpaceX rideshare called Transporter-18 carried, among its other cargo, two satellites belonging to companies that sell electricity. One was Protostar, built by Star Catcher Industries of Jacksonville, which is meant to gather sunlight, turn it into laser light, and aim the beam at the solar panels of a cubesat released from its own hold (Gent, 2026; Woodburn, 2026). The other was Reason-1, the first spacecraft of Cowboy Space Corporation, built to send a kilowatt-class laser down to a receiver on the ground (Cowboy Space Corporation, 2026). A third passenger, by WIRED’s account, was Google’s first attempt at a data center in space (Kahn, 2026). Six weeks on, neither power company’s newsroom carries a figure for power delivered. Star Catcher’s chief executive had prepared his audience for the interval. “Testing this on orbit is success in and of itself,” Andrew Rush told Payload before the launch (Woodburn, 2026). The trade has sold ahead of the watt all the same. Star Catcher reports ten power purchase agreements, modeled on terrestrial utility contracts, under which a customer puts down a deposit to hold a rate for a set period, and forty letters of intent behind them. It has raised $88 million in venture capital, beside a Space Force award that WIRED puts at $30 million and Payload at $60 million (Kahn, 2026; Woodburn, 2026).

Twenty-five days later, on the evening of October 26, Cowboy Space filled a room in Palo Alto for an event it called Big Sky Rising and showed a rocket. Big Iron will be 86 meters of stainless steel, and its upper stage does not come home. It stays in orbit, unfolds solar wings, and operates as a data-center satellite of about one megawatt. The constellation, called Stampede, is to number about 20,000 satellites, one to a rocket, so the plan requires 20,000 launches to build and several thousand more every year to hold the count. First flight is set for late 2028, and the founder, Baiju Bhatt, said of the schedule that “it’s not a guarantee” (Wall, 2026). The company was founded in 2024 as Aetherflux, and its purpose then was the one Reason-1 still carries, to beam solar power from orbit to the Earth. In May 2026 it raised $275 million, took its present name, and filed with the Federal Communications Commission for a constellation it described as in-orbit data centers built for the computational demands of AI, placed in space to get around a strained terrestrial grid (Jewett, 2026). Its own release now describes the business as “building vertically integrated orbital infrastructure for the AI era” (Cowboy Space Corporation, 2026). The beam to the ground, which was the product, has become the pathfinder for a different product.

The same Monday had begun with a prospectus. Anthropic’s registration statement became public that morning, and by its second amendment, filed on November 10, it priced 825 million shares at $70 to $78 and disclosed aggregate committed spending of approximately $900.9 billion, “substantially all of which is related to compute and infrastructure” (Anthropic, PBC, 2026; Foxley, 2026). Between the filing and the amendment came November 5, when the S&P 500 fell 4.36 percent and the Nasdaq 5.52 percent in a sell-off led by chip stocks and attributed in part to worries about AI regulation after the midterm elections (Izquierdo, 2026). A rocket was unveiled on the evening of a day that opened with a ledger of obligations, and the two documents describe one economy from its opposite ends.

The argument of this essay is that the orbital power trade recapitulates, at one further remove, the townsite speculation that grew up beside Jay Cooke’s Northern Pacific Railroad between 1869 and 1873. The Northern Pacific was a road built ahead of its traffic and financed on bonds secured by a land grant it had not yet earned. Beside it stood a second kind of company, the Western Land Association and the Lake Superior and Puget Sound Company, whose business was to plat and sell lots in towns that would exist if the road arrived and if the road, having arrived, found freight (Sakolski, 1932, Chapter 14). The mechanism of recurrence is derived demand read as present demand. A first enterprise borrows against revenue it expects. A second tier of firms treats its forward commitments as a market, and a third treats the second tier’s letters and deposits the same way. Each remove is capitalized now, and each stands further from anyone who has paid for a finished thing. The historical difference has to be put early, because the rest of the essay leans on it. Cooke’s road ran toward a country with almost no freight in it. The laboratory at the head of the present chain has customers, and they are multiplying; the prospectus that lists $900.9 billion in commitments also reports $17.0 billion of revenue in a single quarter (Anthropic, PBC, 2026). The analogy therefore bears on the derivative tiers, the orbital data center and the orbital grid that would feed it, and on the paper that ties them back to the first. Artificial intelligence enters here only as the banks treat it, as a weight in an index and a column of forward obligations. The essay ends on one property of orbit that no railroad shared, which is how briefly anything lasts there.

The strongest objection is that this is how infrastructure has always been built, and it is made with most force by Carlota Perez. In her account every technological revolution passes through an installation period, in which financial capital, drawn by early returns, pours into the new networks faster than use can absorb it. The frenzy ends in a crash. What the crash leaves behind is an installed base, bought with other people’s losses, on which a longer and broader period of deployment is then built (Perez, 2002). The bubble on this reading is the funding mechanism, and perhaps the only one available, for assets whose returns are collective, late and uncertain. Duluth is her case in miniature. On January 27, 1871, Representative J. Proctor Knott of Kentucky rose in the House against a land grant for a railroad to the head of Lake Superior and delivered a long comic oration at the expense of the promoters’ town. He claimed to have hunted it through every map in the Library before concluding that it must be “a place of untold delights, a terrestrial paradise” (Knott, 1871/1971). The House laughed and the bill died. For several years afterward the Northern Pacific handed the speech to passengers in its dining cars as a souvenir, and by the 1880s, when Duluth had some 30,000 inhabitants, its chamber of commerce reprinted the satire as a prospectus that had come true (Knott, 1871/1971). Cooke, ruined in 1873, rebuilt his fortune on the Duluth land he had kept (Sakolski, 1932, Chapter 14). The paper town became a port. On Perez’s reading the mockery of 1871 is the error to avoid, and an essay that laughs at a megawatt satellite before it has flown is Knott with footnotes. The objection has a second prong, and it is empirical. The demand at the head of this chain is real, as it was not for Cooke: a quarter’s revenue of $17.0 billion is freight.

Perez’s sequence is sound history, and the answer to it lies in what the sequence leaves unspecified: which assets survive the turning point, and who was holding the paper on the others. Richard White’s study of the transcontinentals supplies the first half. The western roads, he told an interviewer, were “built way ahead of demand,” by three or four decades in many cases, and would not have been built at all without public land and public credit (Inskeep, 2011; White, 2011). Sakolski’s older account of Cooke supplies the second half, because it follows the money into the companies that stood beside the track. Duluth in 1869 was six or seven frame houses, a school and a land office. Cooke and a partner held some 40,000 acres around it. The Western Land Association was formed to work that holding, along with other town sites on the line to St. Paul, and when the Northern Pacific agreed to fix its eastern terminus at Duluth the Association sold 2,000 of its shares to the Lake Superior and Puget Sound Company for $120,000. The buyer was the railroad’s own real-estate subsidiary, capitalized at $2.4 million, owned half by the Northern Pacific and half by Cooke’s pool. About 13,416 town lots, most of them in Duluth, were then divided among the insiders. The railroad’s bonds paid 7.30 percent and were secured by a mortgage on a land grant that would be earned only as track was laid. They were sold the way Cooke had sold war loans, with heavy advertising, touring parties over the route, and pamphlets whose map showed the grant as a broad curved band across the continent. From that band the road took its nickname, Jay Cooke’s Banana Belt (Sakolski, 1932, Chapter 14).

The townsite company’s customer was a settler who had not come. Its counterparty, the buyer of its shares and the guarantor of their value, was a subsidiary of its own supplier. Its asset was a location, worth something only if a second enterprise, itself unfinanced beyond the next bond sale, completed a road and then found traffic for it. The bonds sold badly at home and abroad. Jay Cooke & Company closed on September 18, 1873, the Northern Pacific went into bankruptcy, and before the reorganization of 1875 the railroad bought in the share capital of its land company for $721,630 in defaulted bonds. The company was wound up in 1891, and its remaining Puget Sound lands fetched $72,000 (Sakolski, 1932, Chapter 14). Duluth did get built. It was built for the holders who came after the default, on lots that had sat through it, and the bondholders of 1873 were not in the dining car when the speech was handed round.

Against that structure the orbital power sector sorts into tiers by distance from a paying user. Beneath the tiers lies an older and smaller thing, the state’s long interest in beamed power, which owes nothing to the present boom. NASA’s own study, presented in 2024, put space-based solar designs delivering power to the grid at 12 to 80 times the cost per unit of electricity of terrestrial renewables (Rodgers et al., 2024). The patrons who stayed were military. Cowboy Space credits multi-year support from the Operational Energy Capability Improvement Fund for the laser and optics aboard Reason-1; Star Catcher holds its Space Force award; and the Pentagon’s Defense Innovation Unit issued a call in July for power-beaming prototypes, with operational capability sought by fiscal 2030 (Cowboy Space Corporation, 2026; Woodburn, 2026). This is a real customer with a real use, a forward base or a satellite that cannot afford to go dark, and it is not large.

The first tier of this structure is the laboratory and the hyperscaler, the firms that owe for compute. The second is the orbital data center, which proposes to supply them. SpaceX has filed with the FCC for a system that the American Astronomical Society describes as one million data centers in space (American Astronomical Society, n.d.). Google’s researchers published a design study in November 2025 that begins from space-based solar power, credits the idea to a 1941 story of Asimov’s, and then turns it over: leave the power in orbit and send the computation up to meet it (Agüera y Arcas et al., 2025). Its first hardware was the third passenger on Transporter-18 (Kahn, 2026). Cowboy’s Stampede belongs to this tier, and so does Starcloud. The economics of the tier are upside down for now, by the trade’s own reporting. One engineer’s public model puts a one-gigawatt orbital data center at roughly $42.4 billion, almost three times its equivalent on the ground (Fernholz, 2026a). Launch on a reusable Falcon 9 runs about $3,600 a kilogram, against the $200 that Google’s paper says parity requires and expects in the 2030s. The working life of an AI satellite is put at around five years (Fernholz, 2026a). What this tier has by way of demand is a sentence. Announcing its lease of SpaceX’s Colossus 1 data center in May, Anthropic added that it had expressed interest in working with SpaceX on “multiple gigawatts of orbital AI compute capacity” (Anthropic, 2026). An expression of interest is the freight on which the second tier is being capitalized.

The third tier sells power to the second. Star Catcher’s named purchasers include Starcloud, which is to say that the orbital grid’s customer is an orbital data center still waiting for a customer of its own (Woodburn, 2026). Overview Energy, of Ashburn, Virginia, holds a capacity reservation from Meta for up to one gigawatt from a fleet of as many as a thousand spacecraft in geosynchronous orbit, which would brighten existing solar farms after dark with near-infrared light. Its first demonstration satellite is planned for January 2028 and its deliveries for 2030, and TechCrunch could not establish that any money had changed hands (Fernholz, 2026b). Cowboy Space shows where the logic leads when the counterparties fail to appear. It began in the third tier, selling a beam. It moved into the second, selling compute. Now it proposes to build the rocket as well, because a firm whose suppliers and customers do not yet exist can only integrate until it is both. The Western Land Association solved the same problem in the same way, by selling itself to its supplier’s subsidiary.

The other end of the chain is the paper. By October a JPMorgan basket of forty-one companies tied to artificial intelligence stood at 49 percent of the S&P 500’s market value (Capul, 2026). A November report drawing on S&P Dow Jones Indices data put AI-related stocks at about 45 percent of the index. It credited them with 22 of the 33 percentage points of the index’s price return since January 2024, and relayed Michael Hartnett’s estimate at Bank of America that the share of the American market could reach 48 percent once the large private laboratories list (Esebame, 2026). The same report supplies the multiplier for everyone else: the United States is about a quarter of the world’s economy and about 63 percent of the world’s stock-market value. Set that 63 percent against the index share, and something near three-tenths of the planet’s listed equity rests on a single expectation. The arithmetic is mine and it is rough, since the S&P 500 is not the whole American market, but the order of it is right. The International Monetary Fund’s October stability report states the conditional in its own flat idiom, warning that weaker or less durable investment and profitability in AI could bring “a correction in AI-related asset prices” (International Monetary Fund, 2026). And when Bank of America’s strategists went looking for a precedent for the concentration they found one: the railroads, at about 63 percent of American market capitalization in the 1880s (Esebame, 2026). The banks had chosen this essay’s lens before the essay did.

Under the equity lies the schedule of what has been promised for compute, and it has a short, a middle and a long term. The short term runs to September 30, 2027, and inside it the filing shows approximately $70.7 billion of Anthropic’s commitments coming due (Anthropic, PBC, 2026). That is more than the offering would raise for the company at the top of its range, about $62.5 billion on the 801 million shares it is selling for its own account, and it is of the same order as four quarters of revenue at the third-quarter rate. Both comparisons are mine, not the filing’s. Against them stand $95.6 billion in cash and $36.3 billion in short-term investments. In the same window the builders keep a calendar of their own. BloombergNEF counts $13.4 billion of fixed-rate, dollar-denominated data-center debt maturing in 2027, which at October’s Treasury yields would refinance 2.85 percentage points dearer, inside at least $1 trillion of such borrowings due within five years (Cheng, 2026). The middle term runs to the end of 2030, by which date the filing expects approximately $451.6 billion of the total to have been incurred. It is also where the chip-leasing firms meet their wall. Their debt maturing between 2030 and 2032 comes to $49 billion, about 65 percent of what they owed in September, and two companies owe four-fifths of it (Cheng, 2026). The long term is nearly all of the remainder, about $449 billion, which the filing places between 2031 and 2045 (Anthropic, PBC, 2026). One company’s obligations thus reach nineteen years forward. The hardware they pay for, on the trade’s own estimate, is obsolete or dead in about five.

The schedule is also uneven in how firmly it binds, and the unevenness runs in one direction. The largest single arrangement, up to $120.5 billion with SpaceX’s AI subsidiary, runs at about $3.1 billion a month once capacity is delivered. Its non-cancelable periods end in January and March 2027, after which either party may leave on ninety days’ written notice (Anthropic, PBC, 2026; Foxley, 2026). At the foot of the chain a power purchase agreement holds a rate with a deposit, and a reservation holds a place in a queue. In between stand long bonds and five-year loans. The promise is firmest where it lies furthest from the user, and loosest where the user stands. Cooke’s land-grant mortgage worked the same way: the bond was a fixed claim, and the settler was a hope.

A different objection comes from the history of technology, and it charges the essay with the fault the essay describes. Edgerton would observe that the lens has been chosen for its glamour. “The technological boosterism of the past has too often been turned into the history of our material world” (Edgerton, 2006/2008, p. 4), and a railroad panic with a banana-shaped map is boosterism’s reverse image, equally loud. A use-centered account would start from what is in use. Power for these machines is being supplied by gas turbines, grid interconnections, terrestrial solar and batteries; Meta’s data centers drew more than 18,000 gigawatt-hours in 2024, every one of them from the ground (Fernholz, 2026b). Against that the orbital power sector is a few hundred million dollars of venture money, a handful of defense awards and two satellites launched on someone else’s rocket. To call it an industry built on borrowed paper is to inflate it. The debt belongs to the first tier. The space firms are financed with equity, which is the capital whose owners expect to lose it, and their steadiest customer, the military, predates the boom and will outlast it. Edgerton’s test of significance applies as well: the “hidden counterfactual assumption” behind any claim of importance “is that there was no alternative” (p. 6), and here the alternative is winning. Most inventions are properly refused. “Resistance is required” (p. 9). The market’s refusal of orbital power, if it comes, will be ordinary selection and will need no help from Jay Cooke. Google’s engineers add the technical form of the same calm. Their paper calls the project a moonshot, reports that its chips survived a radiation dose equivalent to a five-year mission, and expects launch near $200 a kilogram by the mid-2030s on a learning curve (Agüera y Arcas et al., 2025). Nobody in that document is selling town lots.

Most of this has to be conceded. The sector is small. Its money is equity; nobody has lent against a laser, and the borrowing in this story sits one tier up. The defense customer is real and was there first. And the chain this essay has traced, from a laboratory’s expression of interest to an orbital grid’s deposit, is made of intentions. What the concessions leave standing is the question of what these firms say they are for, and on that the record is their own. Cowboy’s FCC filing gives the computational demand of AI as the reason for its constellation, and the company took its new name in the same month it took the $275 million (Jewett, 2026). Star Catcher lists an orbital data center among its purchasers (Woodburn, 2026). Overview’s one announced reservation is from Meta, for data centers (Fernholz, 2026b). Small firms can be priced on a large expectation, and these have said which one. Edgerton’s test then turns on his own objection. If the ground is winning, as he would rightly insist, the orbital tiers have no demand of their own. They are waiting on the overflow of someone else’s, and that someone’s largest contract can be ended on ninety days’ notice from next spring.

Cooke’s structure had two layers, a road and its towns. The present one has three, and the third has come almost wholly loose from the first: a laser company in Jacksonville holds a deposit from an orbital data-center company that hopes for a contract from a laboratory that has, so far, expressed interest in the general idea. Nothing in the nineteenth-century case stands this far from a customer.

The holder is new as well. A buyer of Northern Pacific bonds in 1871 had read a pamphlet and made a choice. The holder of an index fund in 2026 has made no choices about low-orbit investment, and holds the theme at nearly half the benchmark by default, which is why the November report on concentration was written for people with retirement accounts (Esebame, 2026). The paper has been distributed more widely than any promoter of Cooke’s generation could have managed, and with far less persuasion.

Orbit supplies the answer to Perez. Her sequence needs the installed asset to outlast the crash. Roadbed did, and a Duluth lot could wait fifteen years for the city to arrive, because waiting is what land does. Low Earth orbit does not permit it. The satellites of this trade are given about five years, their chips are expected to be outdated inside that term, and Stampede at full strength would need several thousand launches a year merely to stand still (Fernholz, 2026a; Wall, 2026). At about 15 metric tons a satellite the constellation is some 300,000 tons of hardware placed in orbit and then replaced on a cycle shorter than a car loan; the multiplication is mine, on the company’s figures as Space.com reports them. Smil calls it a category mistake to suppose that systems of mass and energy move at the speed of the electronics they serve (Smil, 2022, Chapter 7), and an orbital data center is a chip strapped to exactly such a system. Vinsel and Russell warn that those who build without providing for future care end with “a stress-inducing mountain of deferred maintenance and infrastructural debt” (Vinsel & Russell, 2020, Chapter 4). In orbit there is no deferral to be had. Maintenance there means replacement, and replacement means launch. A constellation that is not maintained does not decline into a ruin that a later owner can buy cheaply and put to work. It reenters. A bust in these tiers therefore leaves no roadbed for the deployment period to inherit, and the liabilities written at the head of the chain run to 2045 against assets at its foot that will not see 2032.

Perez is right that a great deal of what the frenzy buys on the ground will still be standing afterward, the substations and the transmission and perhaps the buildings, and right that ridicule is a poor forecaster; Knott’s speech is very funny and it was wrong about Duluth. Edgerton is right that the sector is small, that refusal is the ordinary fate of inventions, and that the durable use is probably the dull one: a laser that keeps a military satellite alive may be the real product of all this, as a county’s cloud-seeding contract was the real trade behind the rainmaker. Google’s engineers are right that no law of physics forbids a computer in orbit. Each of these judgments, however, is passed on a technology or on an era. The thesis here is a claim about an instrument: when it was capitalized, against what, by whom, and whether the thing it bought can afford to wait. It is stronger than the objections because it accounts for what each of them holds constant. Perez’s legacy accrues to durable assets and to later owners, and neither condition is met a few hundred kilometers up. Edgerton’s smallness measures the firms and leaves out what they are written on, which is $900.9 billion of one company’s forward commitments and nearly half of an index. The engineers’ learning curve arrives in the mid-2030s, by which time two generations of the hardware now being financed will have burned up. And the contractual facts all point one way, with the firm claims held far from the user and the cancellable ones beside him.

The story Google’s paper names as the origin of the whole idea is Asimov’s “Reason,” published in 1941. Cowboy’s first satellite shares its title. In it a robot placed in charge of a solar power station concludes that the Earth is a fable its human minders tell, and goes on holding the beam steady, through a storm and better than they could, for a planet in which it does not believe. The present trade keeps its faith the other way round. It believes ardently in the customer and has yet to show the beam. The Northern Pacific could afford to pass Knott’s speech around its dining cars because the lots were still there when the joke turned. Nothing in low Earth orbit lasts long enough to be vindicated.

References

Agüera y Arcas, B., Beals, T., Biggs, M., Bloom, J. V., Fischbacher, T., Gromov, K., Köster, U., Pravahan, R., & Manyika, J. (2025). Towards a future space-based, highly scalable AI infrastructure system design. arXiv. https://doi.org/10.48550/arXiv.2511.19468

American Astronomical Society. (n.d.). Action alert: Provide input to the FCC on proposed satellite systems. Retrieved November 12, 2026, from https://aas.org/action-alert-provide-input-fcc-proposed-satellite-systems

Anthropic. (2026, May 6). Higher usage limits for Claude and a compute deal with SpaceX. https://www.anthropic.com/news/higher-limits-spacex

Anthropic, PBC. (2026, November 10). Amendment No. 2 to Form S-1 registration statement (Registration No. 333-299633). U.S. Securities and Exchange Commission. https://www.sec.gov/Archives/edgar/data/0002130357/000162828026072484/anthropicpbc-sx1ax2.htm

Capul, J. (2026, October 8). JPMorgan’s AI basket represents 49% of the S&P 500. Seeking Alpha. https://seekingalpha.com/news/4651484-jpmorgans-ai-basket-represents-49-of-the-sp-500

Cheng, M. (2026, November 9). AI developers face 2027 rate shock after $1 trillion debt binge. Bloomberg. https://www.bloomberg.com/news/articles/2026-11-09/ai-developers-face-2027-rate-shock-after-1-trillion-debt-binge

Cowboy Space Corporation. (2026, October 1). Cowboy Space launches Reason-1 to demonstrate first laser power beaming from orbit to Earth [Press release]. Business Wire. https://www.financialcontent.com/article/bizwire-2026-10-1-cowboy-space-launches-reason-1-to-demonstrate-first-laser-power-beaming-from-orbit-to-earth

Edgerton, D. (2008). The shock of the old: Technology and global history since 1900. Profile Books. https://openlibrary.org/books/OL25573148M (Original work published 2006)

Esebame, D. (2026, November 9). 401k balances depend on one move most investors never make. TheStreet. https://www.thestreet.com/investing/sp-500-ai-concentration-401k-risk

Fernholz, T. (2026a, February 11). Why the economics of orbital AI are so brutal. TechCrunch. https://techcrunch.com/2026/02/11/why-the-economics-of-orbital-ai-are-so-brutal/

Fernholz, T. (2026b, April 27). Meta inks deal for solar power at night, beamed from space. TechCrunch. https://techcrunch.com/2026/04/27/meta-inks-deal-for-solar-power-at-night-beamed-from-space

Foxley, W. (2026, October 26). Anthropic discloses up to $120.5 billion in SpaceX compute spending through 2029. Yahoo Finance. https://finance.yahoo.com/technology/ai/articles/anthropic-discloses-120-5-billion-140613686.html

Gent, E. (2026, October 1). Star Catcher is about to beam power between two satellites for the first time. Singularity Hub. https://singularityhub.com/2026/10/01/star-catcher-is-about-to-beam-power-between-two-satellites-for-the-first-time/

Inskeep, S. (Host). (2011, July 11). How trains “railroaded” the American economy [Radio broadcast transcript]. NPR. https://www.npr.org/transcripts/137497772

International Monetary Fund. (2026). Global financial stability report: Resilience, risk, and transformation. https://www.imf.org/en/Publications/GFSR/Issues/2026/10/13/global-financial-stability-report-october-2026

Izquierdo, R. (2026, November 5). Stock market today, Nov. 5: Intel falls on weak fourth-quarter sales outlook amid chip selloff. The Motley Fool. https://www.fool.com/coverage/stock-market-today/2026/11/05/stock-market-today-nov-5-intel-falls-on-weak-fourth-quarter-sales-outlook-amid-chip-selloff/

Jewett, R. (2026, May 15). Cowboy Space files for 20,000-satellite constellation for data center services. Via Satellite. https://satellitetoday.com/technology/2026/05/15/cowboy-space-files-for-20000-satellite-constellation-for-data-center-services/

Kahn, B. (2026, September 28). Space lasers are about to get their first real test generating energy. WIRED. https://www.wired.com/story/space-lasers-are-about-to-get-their-first-real-test-generating-energy/

Knott, J. P. (1971). The untold delights of Duluth (D. G. McCullough, Ed.). American Heritage, 22(4). https://www.americanheritage.com/untold-delights-duluth (Original speech delivered 1871)

Perez, C. (2002). Technological revolutions and financial capital: The dynamics of bubbles and golden ages. Edward Elgar. https://www.e-elgar.com/shop/usd/technological-revolutions-and-financial-capital-9781843763314.html

Rodgers, E., Sotudeh, J., Mullins, C., Gertsen, E., Hernandez, A., Le, H., Smith, P., & Joseph, N. (2024, July 30-August 1). Space based solar power study [Conference presentation]. 2024 ASCEND, Las Vegas, NV, United States. https://ntrs.nasa.gov/citations/20240008752

Sakolski, A. M. (1932). The great American land bubble: The amazing story of land-grabbing, speculations, and booms from colonial days to the present time. Harper & Brothers. https://mises.org/online-book/great-american-land-bubble/chapter-xiv-jay-cookes-banana-belt

Smil, V. (2022). How the world really works: A scientist’s guide to our past, present and future. Viking. https://www.penguin.co.uk/books/319141/how-the-world-really-works-by-smil-vaclav/9780241989685

Vinsel, L., & Russell, A. L. (2020). The innovation delusion: How our obsession with the new has disrupted the work that matters most. Currency. https://www.penguinrandomhouse.com/books/576816/the-innovation-delusion-by-lee-vinsel-and-andrew-l-russell/

Wall, M. (2026, October 27). Cowboy Space unveils “Big Iron” rocket to launch 20,000 AI data centers. Space.com. https://www.space.com/space-exploration/launches-spacecraft/cowboy-space-unveils-big-iron-rocket-to-launch-20-000-ai-data-centers

White, R. (2011). Railroaded: The transcontinentals and the making of modern America. W. W. Norton. https://wwnorton.com/books/Railroaded/

Woodburn, V. (2026, September 28). Star Catcher’s orbital power-beaming demo is ready for launch. Payload. https://payloadspace.com/star-catchers-orbital-power-beaming-demo-is-ready-for-launch/