Summary

Former SpaceX executive 洪力德, who oversaw 7 departments and more than 3,000 rocket components and is now GP at Aries Fund, sees a SpaceX IPO as “the space industry’s ChatGPT moment.” But that would mark an inflection point in how capital understands the sector, not in the underlying technology. The real turning point was December 21, 2015, when SpaceX first reached orbit and recovered a rocket: “Once you can recover the vehicle, that is the starting point for everything.”
The first-principles logic behind the xAI acquisition is a space-based data center. U.S. data-center construction is constrained by permitting and power shortages: most facilities across the 3 major grids are more than 30 years old, the newest nuclear plants were built in the 1990s, and the power gap had already reached 25-40% before AI. Space needs no permits, solar conversion efficiency is more than 10 percentage points higher, and signals travel twice as fast through a vacuum as over the most advanced terrestrial fiber. “Whoever gets there first gets the position.”
Musk bought the x.com domain in 2014, and 洪力德 speculates that his companies will eventually converge into an Alphabet-like X holding company. Tesla would handle physical AI, Neuralink the most direct human-machine interface, and the tunneling company may be tied to radiation-shielded underground housing: “I don’t think these companies are accidental.” Grok’s foothold may be Physical AI and world models; he says 2026 Grok will be worth watching.
Rockets are a means, not the end. Falcon 1 was retired immediately after its fourth launch succeeded in 2008; the cost of putting a kilogram into orbit fell from $10,000-$20,000 before SpaceX to $3,000 with a reusable Falcon 9, while Starship targets less than $100. Once the cost falls to one-two-hundredth, “projects that used to sound like science fiction in space start to pencil out on the books.”
Firsthand testimony captures Musk’s management style. He holds only 3 kinds of meetings: when something is late, cannot be built, or needs more money. A 30% efficiency improvement earns “get out—you’re not thinking,” because the bar is an order-of-magnitude or even 100x leap. After the team developed a high-pressure vessel at 90% lower cost, Musk’s only response was, “Have you ever looked at how soda cans are made?” When a veteran engineer said something was impossible, Musk calmly replied, “Okay, then I accept your resignation.”
Space startups split into a “SpaceX chain” (Apple chain) and an “Android chain,” with the core founders forming a “new Elon mafia.” His investment targets include space manufacturing—perfect spheres for artificial corneas in zero gravity and next-generation wafers made in a vacuum—and controllable-gravity space stations, a field with 17 companies competing globally.
He has barely sold any SpaceX stock, calling it “at least a trillion-dollar company”—and stressing the “at least” repeatedly. He compares it with the East India Company during the Age of Discovery and says it is only getting started. On competition, Musk’s exact words were: “Our competitor is the speed of our own progress, but we are watching China’s space program closely.”

Deep dive

1. The IPO Is Capital’s ChatGPT Moment; the Real Inflection Point Came on December 21, 2015

洪力德 uses a dual framework. From the market’s perspective, this year “does indeed feel like a ChatGPT moment for space”: ever since reports that SpaceX may pursue an IPO, inbound calls have poured in asking about the stock and space projects. From the industry’s perspective, however, “I don’t think there has been a particularly large inflection point this year.” It is simply the result of sustained execution.
The real turning point was December 21, 2015, “a hugely important day for every SpaceX employee”: the first successful orbital launch and rocket recovery. That rocket is still displayed at SpaceX’s Los Angeles headquarters. His logic is straightforward: if any transport vehicle is discarded after getting from A to B, “you can’t scale anything or do anything.” Recovery is the starting point for everything—“even bigger than the GPT-3 moment.”
The industry broadly considered the achievement irrational and economically incoherent. A startup did it anyway. More important was SpaceX’s foresight: in 2012, when its rocket could barely launch once a year, the Mars team had already been formed and Starship had already been conceived. “The things everyone now thinks Musk is randomly posting on X—we had already been talking about them at SpaceX more than 10 years ago.”
SpaceX had a master plan similar to Tesla’s, but “it wasn’t the I Ching”; it did not spell out specific years. The company’s core mission has never changed: “The company exists to open up the universe for all of humanity to explore.” Musk never defined it as a rocket company.

2. The First-Principles Case for the xAI Merger: Space as a Data Highway and an Escape Valve for Data Centers

Starlink has 2 purposes. First, no one will fund Mars directly: “If you tell any investor or government that you want to go to Mars, I don’t think anyone will pay for it.” SpaceX therefore has to generate the capital itself. Second, physics gives space a natural advantage: information travels through a vacuum “another 1x faster” than over any of the most advanced optical fiber on Earth—in other words, twice as fast. Starlink began in 2015, and SpaceX moved to the next step immediately after mastering recovery.
U.S. AI infrastructure is constrained by 2 things: approvals and power. Securing land, connecting electricity, obtaining permits and winning community consent “takes a tremendous amount of time.” The U.S. effectively has only 3 major power grids—the East Coast, West Coast and Texas grids. Most of their facilities are more than 30 years old, the 2 newest nuclear plants were built in the 1990s, and the power shortfall had already reached 25-40% in the electrification era. “Add AI, and it gets even harder.”
Space offers the opposite setup: no permits are required, and “whoever gets there first gets the position.” Experiments have shown solar conversion efficiency more than 10 percentage points higher than on Earth. There is effectively unlimited energy and space, with no one restricting development. “That is the underlying logic behind combining xAI with SpaceX to build space-based data centers.”
When Sam Altman mocked space-based compute as “very laughable,” 洪力德 responded, “I think he’s jealous,” and declined to continue the trash talk. His engineering distinction was more rigorous: it may be possible, but “even for SpaceX it would be a difficult job.” The benchmark is not merely building it; the system must compete with or outperform a terrestrial data center on efficiency. “If you don’t achieve that, it has no meaning.” Launch cost is “a relatively simple problem” by comparison.

3. The X.com Clue: Musk’s Companies May Eventually Converge into a Single X

The 2014 lunch gossip went like this: a colleague said, “The boss bought the x.com domain today.” 洪力德’s guess is that all of Musk’s companies will one day sit under a holding company like Alphabet. Tesla would focus on physical AI—robots and autonomous driving. Neuralink would deliver the most direct interface between humans and AI through brain-computer interfaces rather than a new iPhone or smart glasses. The tunneling company may be linked to radiation-shielded underground housing. “I don’t think these companies are accidental.”
The host pointed out that xAI’s Day One vision was to “understand the nature of the universe,” with a black hole as its logo—suggesting an innate link to SpaceX. 洪力德 conceded, “You’ve read more deeply than I have,” but offered his competitive map: Gemini dominates text through Search; Claude has focused on B2B coding tools from the beginning; OpenAI targets consumers; and Grok’s unique foothold is the data inside Musk’s ecosystem—Physical AI and world models.
He was candid about whether xAI is falling behind: “That’s not my expertise.” But his forecast for 2026 is that investments such as last year’s 100,000-GPU Super Data Center “will emerge this year.” “This year’s Grok is something everyone can watch.”
Does Starship need an AI brain? “I think it’s still pretty early.” But he accepted the host’s reference to Moss in The Wandering Earth: in deep space, how many people can you send, and who is your most reliable companion? “AI and robots.” The 2 organizations also share a similar culture—Musk culture, summed up as “move fast, break things.”

4. Rockets Are a Means, Not the End: From $20,000/kg to $100/kg

Falcon 1’s retirement was a counterintuitive decision. After its fourth launch succeeded in 2008—following 3 failures, near-bankruptcy and an all-in effort to produce a fourth rocket—SpaceX retired it immediately. Any rocket company would have used the success to take orders and generate revenue. “SpaceX never treated rockets as the company’s purpose. Rockets are a means, not the end.” Rocket Lab’s Electron carries less than one-third of Falcon 1’s payload but now launches more than 70 times a year. SpaceX chose not to monetize Falcon 1 further and even rejected U.S. government contracts to shift into defense.
The product ladder is straightforward. Falcon 9 carries 25x Falcon 1’s payload and has only 2 missions: generate revenue and “teach us how to build a big rocket from a small rocket.” Starship is then 5x Falcon 9. The industry’s core KPI is cost per kilogram to orbit: $10,000-$20,000 before SpaceX, $3,000 after Falcon 9 reuse, and a Starship target below $100. A 200x reduction turns “the fantasies once discussed in space into projects that can be calculated on the books.”
Falcon 9 is “the Model T of aerospace”—the product with which Ford first mass-produced automobiles. “It’s not the Tesla era yet; we’re still 100 years away.” Musk wants Starship to jump straight to the Model 3. 洪力德’s reservation is unchanged: “I think that’s a little ambitious. Reaching the Model S would already be good.” Falcon 9 took 10 years from development to the recovery milestone; Starship has covered half the distance in half the time. “Perhaps it can reach that level within the next 2-3 years.”
Falcon 9 will not disappear. “It’s basically the industry’s 737.” Boeing has the 747 and 777, but the 737 remains its workhorse model.

5. Hiring Philosophy: A Food Engineer’s Letter, Generalists First, Veterans After 6 Months

Every SpaceX candidate traditionally faced a final interview with Musk. 洪力德 was one of the few exceptions because Musk was traveling in Australia; instead, he wrote a letter explaining why someone from food could build rockets. Food is widely misunderstood: it is “the hardest product,” requiring huge volumes to be produced in a very short time, with every item subjected to the world’s highest quality inspection—your mouth. HR later told him Musk decided within 1 minute.
SpaceX did not treat aerospace specialists as the only qualified candidates. “The industry had not changed much in the past few decades, and these were the people responsible.” It valued how quickly someone could absorb information and grow over decades of prior experience, recruiting from many industries to bring new blood into aerospace. 洪力德 himself had no rocket background when he took over Dragon’s interior and later became Falcon 9’s lead manufacturing engineer.
When he joined, SpaceX had roughly 1,000 employees; Boeing, Lockheed and ULA used 10,000 people to do comparable work—a 1:10 labor-efficiency ratio. The organization was extremely flat, with 1 engineer accountable for each component. “If that person were hit by a train tomorrow, there might genuinely be a hole.” If you cannot deliver, the entire company waits. The internal joke was: “Stay at SpaceX for more than 6 months and congratulations—you’re a veteran.” Turnover was high because many people quickly realized, “This is absolutely not an environment I can handle.”
The dark joke about rewards was: “Do a good job, and the only reward is that the company gives you more work.” The hours were the U.S. version of 996: engineers and managers were salaried and received no overtime pay. “SpaceX had a group of completely free engineers.” Twelve-hour days and 6-day weeks were the baseline.

6. Meetings with Musk: 3 Types of Meetings, Soda Cans, and “I Accept Your Resignation”

Meetings with Musk covered only 3 things: “First, your thing is late. Second, your thing can’t be built. Third, your thing needs more money.” “There was never a meeting that was simple.” Musk’s claim to be chief designer was “not entirely wrong,” because he really did make the final call on designs large and small across the company.
At other companies, saving 10-30% can earn a promotion and a bonus. Tell Elon you increased a project’s efficiency by 30%, and “he’ll tell you to get out—you’re not thinking.” The demand was for 10x or 100x improvements: an order-of-magnitude leap.
The full soda-can story involved Falcon 9’s high-pressure vessel. A car tire holds 40 PSI; the vessel operated at the level of tens of thousands of PSI. Suppliers charged more than $100,000 per unit and could deliver only a few a year. The team built its own version at 90% lower cost and at a production rate the program could support, convinced it was “the best in the world.” Musk did not even say “good.” His first question was: “Have you ever looked at how soda cans are made?” A soda can is also a pressure product from first principles—more than 1,000 produced per minute at a few cents each. “Do you still have room to improve?” From first principles, there was no rebuttal.
Musk did not yell, but what he did was harsher. When a veteran engineer reported, “Elon, the thing you’re asking for is impossible,” Musk calmly replied, “Okay, then I accept your resignation.” “For the next 30 seconds, the air froze.” 洪力德 rejects the outside label of Musk as a tyrant: “He is very even-tempered and has no sense of hierarchy.” When happy, he can be “like a child.” He never praises people directly; “If I’m not talking to you, you’re actually doing a good job—just keep going.” Early Musk was “extremely introverted”; during annual speeches, “you felt more uncomfortable sitting in the audience than he did on stage.” The great storyteller of today was trained over more than 10 years, but “he has always been a very genuine person, consistent inside and out.”

7. The Manufacturing Revolution: Mini Cooper Teams, the World’s Most Advanced IKEA, and 40 Launches a Year

When SpaceX expanded production, its first hires were not from Boeing or Mercedes-Benz and BMW, but from a group of Mini Cooper teams. Mini had the most SKUs of any vehicle line and had genuinely achieved “high-volume production with high-variation control.” That matched Falcon 9’s nature: in the early days, “no 2 Falcon 9s were exactly the same.” Every launch generated data for the next rocket, followed by immediate iteration and another launch.
Dragon’s interior was self-deprecatingly defined as “the world’s most advanced IKEA.” It had to be as light as possible, withstand several times the gravity that any piece of furniture would encounter, and integrate perfectly with every other system. “Some cabinets may have cost more than a house in the American Midwest.”
The signature analogy for rocket precision was: “If you drive from here to Los Angeles today, and just as you pull into your front door, the last drop of fuel runs out—that is a rocket. If you have a lot of fuel left, you have a bad rocket.” Add a one-way journey through a temperature range of several hundred degrees and a complete vacuum. That is the gap between a car and a rocket. Musk nonetheless insisted that “even at such a high specification, you have to simplify it until it is like a car.”
In 2012, the entire world—including the U.S., China and Russia—conducted 77 launches. SpaceX launched once a year, while Musk demanded that “everything be planned around launching at least 40 times a year.” That meant 40x the capacity. “No one knew what Elon wanted. Elon said so, and everyone believed Elon.” His general observation: SpaceX should have a T-shirt that reads, “I told you so,” signed by Elon.

8. The Darkest Hours: Consecutive Failures in 2015-16 Were Harder Than 2008

Outsiders assume the hardest period was Falcon 1 in 2008. 洪力德’s firsthand account says it was 2015-16, after SpaceX had achieved multiple successes and become the industry benchmark. 2 launches failed in succession, and “both were related to my departments.” The problem was that “you were already ahead of the entire industry, and then you failed—who could you ask?” The system was considered the most perfect version, and no one knew the answer. NASA rules required a root cause before another flight; no flight meant no money.
The June 28, 2015 launch took place on Musk’s 44th birthday. He was unusually late getting into the office and asked colleagues how the launch looked. The answer was, “It doesn’t look good.” “A very short time later, something went wrong with the rocket.”
The investigation was brutally honest. For months, with no known cause, the team tested every possibility by intuition. “Luck was part of it—we really did guess correctly.” The failure came from an extreme-edge design, an unknown unknown. “Knock on wood—so far, we’ve had more than 600 launches without much trouble.”
The company’s character showed in the crisis: “Elon’s expression looked exactly the same as usual.” The team simply continued working through the problem one item at a time, “solving it our way.”

9. The Fork in the Road: No Regrets About Missing Google, a Lifetime of Regret If He Had Missed SpaceX

洪力德 was born in Taiwan and grew up in Germany, Hong Kong and Guangzhou. He entered Stanford in 1999, when China sent only 2 students. After graduating in 2003, he turned down an early Google employee offer for an Epson ink-cartridge plant in Portland, where automation pitted 2 U.S. workers against 55 in China. He became Epson’s youngest global executive the following year. The day he graduated from Oxford’s MBA program, Lehman collapsed and his investment-banking offer was rescinded. He later worked as a supply-chain consultant in Hong Kong and then started a food business, coming from a family with 3 generations in food processing.
His definition of regret is worth preserving: “You regret something when you know you should have done it, but fear or the easier path stopped you—and later your intuition proved right. Google was not my intuition at the time, so I have nothing to regret.”
In 2012, a Stanford friend and one of SpaceX’s first 100 employees called him and said, “We are a rocket ship.” At the same time, his former boss was urging him to join the launch of Amazon Japan. Everyone around him gave him the same advice: “Do you really have to ask? Of course you should join Amazon. Why go work on space?” His decision was based on a different calculation: “If SpaceX really succeeded, it would create a world we had never seen before. If I didn’t join, I would regret it for the rest of my life.”
His eventual departure was triggered in quintessentially Musk fashion. On December 31, 2018, Musk emailed all employees about ordering a Tesla that day, taking delivery that day and receiving full self-driving. 洪力德 bought his first new car, a Model 3, and concluded from an engineering perspective that the EV platform was not a gimmick. “If there is an opportunity bigger than SpaceX, it must be in electric vehicles.” He joined a battery company in 2019, caught the EV wave and exited in 2020. Does he still have a rocket fixation? “This thing can never be fully completed. Starship is only the first step.”

10. The Industry and Investment Map: SpaceX Chain vs. Android Chain, with Space Manufacturing Next

He divides the space value chain into 3 layers. Upstream is the hard work—building and launching rockets, satellites and space stations. The middle layer covers constellation management and data downlink. The downstream layer is applications. Starlink is only the beginning: “In low Earth orbit, are satellites the only thing there? This space is several times larger than Earth.” Run the same math and “10 Starlinks outside Earth might still not be enough.” He calls it “the second wave of AWS.”
His fund’s first-principles framework divides space startups into 2 camps. The “Apple chain” or “SpaceX chain” follows SpaceX’s vertically integrated system and requires insiders who deeply believe—and know—that it can be built. The “Android chain” serves the rest of the world beyond SpaceX. The core founders are a “new Elon mafia”: people from SpaceX and Tesla who know how to build sophisticated hardware from zero to one in the U.S., typically in their 20s or 30s, “with both brains and resources.” Their companies are not limited to space; they include supersonic aircraft, small-scale nuclear power and new chips.
He cites 2 concrete use cases for space manufacturing. In zero gravity, anything becomes “a perfectly formed sphere”; the first application could be artificial corneas. The bigger opportunity is wafers. Building a fab costs billions of dollars and amounts to “using a huge amount of resources on Earth to simulate space.” “So why not do it in space?” The fund is tracking a company using SpaceX’s platform to develop next-generation wafers.
The fund has just invested in a space-station project. Of the 17 companies building space stations globally, it chose what it sees as the newest and most deeply infused with SpaceX DNA: the first controllable-gravity space station, which would give humanity control over a third dimension beyond temperature and pressure. Its most accessible external pitch is “the world’s first space-flushing toilet.” 洪力德 stresses that this is an engineering problem, not a science problem: physics has already established that it is possible. “Why did no one do it before? Why did no one recover rockets before?”

11. The Big Picture: NASA’s Self-Revolution, a U.S.-China Duopoly, and a Stock He Refuses to Sell

The old guard’s arc is striking. Early on, Boeing and ULA had 10x the resources and “poured cold water on SpaceX at every turn.” Even Musk’s idol Buzz Aldrin declared that SpaceX would “eventually screw it up,” nearly bringing Musk to tears in an interview. “The strangest thing is that the same people who poured cold water on it back then now say SpaceX is the most powerful company in the world.” The real institutional shift happened at NASA. Cost-plus contracts produced SLS, launched in 2011 with roughly 5x SpaceX’s funding—$25B and continuing—yet yielded a rocket inferior to Starship on almost every dimension. Commercial contracts, beginning with SpaceX’s station-resupply contract, shifted NASA from government-led to government-enabled. After the ISS, which cost nearly $100B and is the most expensive structure ever built, retires in 2030, NASA has decided not to build its own replacement and will commission private companies instead. Why the change? “NASA had already been marginalized. After lying down for 30 years, one day it had to answer for itself.”
On the U.S.-China question, Musk’s answer was measured: “Our competitor is the speed of our own progress, but we are watching China’s space program closely.” 洪力德 expects the 2 biggest participants in space to be the U.S. and China. China “can absolutely outcompete the world in production and iteration.” The irony is that there is no separate Chinese space or American space: everything orbits the same globe, and space is the one domain where the world must work, cooperate and develop together. There are already 77 space agencies globally. “If you own a space station, which country does it belong to?” That remains an entirely undeveloped question.
His valuation view is direct: SpaceX is “the East India Company of the Age of Discovery, and it is only getting started.” Low Earth orbit is merely the first stop; Mars is also merely the first stop. It is “at least a trillion-dollar company—at least, and I’ve said at least many times.” He has barely sold the stock—“not that I’ve sold absolutely none; I sold a little.” In a rapid-fire exchange, he put it more starkly: “If I don’t need to sell, I will never sell.”
His closing view on AI and aerospace runs in both directions. SpaceX “may have been one of the earliest AI companies”: decades of NASA data and extensive simulation let a small company accomplish what large companies had spent decades doing. Musk’s time-allocation principle was 洪力德’s deepest takeaway from 7 years of working with him: “His time belongs only on the company’s hardest problem.” He would not join celebrations after 300 successful launches, but if a mechanical failure stopped the entire line, Musk would be there solving it with the team until the problem was fixed. The recommended book is Silicon Valley Iron Man, rather than the other autobiography, which had 10x the resources but missed the essence of what mattered most inside Musk’s companies.

Episode 186 - June 12, 2026 ‘Hard Fork’ Live, Part 1: Satya Nadella and Cindy Cohn