Not everyone is on board with Elon Musk’s concept of space-based data centers. 

Masayoshi Son, founder and CEO of Softbank, expressed at a recent shareholder gathering that constructing data centers in orbit won’t significantly lower expenses and will require too much time, since “in the race for AI, the upcoming years will be far more crucial than events that may emerge a decade from now.”

In the latest episode of TechCrunch’s Equity podcast, Kirsten Korosec, Sean O’Kane, and I talked about Son’s comments within a wider conversation that covered OpenAI’s intentions for custom chips, Groq’s new $650 million investment round, and more.

Kirsten pointed out the “great irony” in Son being the skeptic, considering SoftBank’s “long history of daring investments.”

Sean added that when Musk discusses “creating a constellation of satellites — which need to be replaced every few years as well — to form an ‘orbital data center,’” he’s simply “assuring that much greater business” for SpaceX.

Continue reading for a sneak peek of our dialogue, refined for brevity and clarity.

Sean O’Kane: Listen, neo-clouds are the new oil, and everyone eager to profit is making the shift to a neo-cloud. I’m excited to announce that TechCrunch is now a neo-cloud; give us all your money.

I mean, this is the strategy. It appears there are numerous players constrained by compute limitations, so whoever can lease out that compute is capitalizing on it, whether that’s Groq, a company somewhat diminished by Nvidia, or Allbirds, which went into bankruptcy and emerged rebranded as a new neo-cloud vendor instead of just selling shoes — Tim Fernholz conducted an interview with the new CEO of that initiative that I highly recommend.

Or consider SpaceX, which proposed: I’m going to establish an AI platform that targets a market as broad as U.S. GDP, but first, we’ll just lease out our compute. We’ve seen this trend with SpaceX; although their deals with Google or Anthropic are more significant, they just inked another agreement, [their] inaugural post-IPO deal, to lease compute to a smaller entity. They are persistently moving in that direction.

I can foresee this becoming a near-term business for Groq. The real question with all these options is their long-term viability.

Anthony Ha: If we’re discussing SpaceX and their AI and data center ventures, we must also touch on Masayoshi Son’s recent statements, where he essentially questioned the purpose of space-based data centers, which is a query we’ve raised on this program. 

It reflects, again, the industry’s acute compute constraints — they need to establish as many data centers as possible, [and] numerous factors make this challenging on Earth, so perhaps space could be the solution. However, I believe Son raises some legitimate concerns: even if everything we’re discussing works — and the expenses are going to be extremely significant to implement it — this isn’t materializing for years and years, so it’s not a fix for any pressing issues regarding current data center requirements.

Kirsten Korosec: I’d like to emphasize that SoftBank has a longstanding tradition of making bold bets. It’s noteworthy when Son asks the question many have pondered. 

There are plenty of VCs and founders who have been captivated by the idea of orbital data centers, and it feels like suddenly everyone is on board. Just a couple of years prior, mentioning this would have been dismissed somewhat. Thus, I think it’s significant that a high-profile individual is raising this concern. Yet, it seems quite ironic that he is the one questioning it, considering his pitch deck features substantial investments in several ambitious concepts.

Sean: WeWork! Mark my words, we’re going to raise this point repeatedly over the next few years. The concept of placing these systems in space poses an intriguing engineering and economic challenge. 

Anthony, what you pointed out is certainly somewhat accurate. Elon Musk is someone who despises bureaucracy, and as there are no NIMBYs in space, he’s bound to pursue that path. 

To me, it ultimately boils down to: The business as it currently stands for SpaceX, particularly its launch segment, is overwhelmingly dependent on Starlink. The reason they command 80 or 90% of the global launch market isn’t solely due to their superior performance compared to nearly every other launch provider, but also because of Starlink, which significantly boosts that number. Without Starlink, they’d likely be around — I don’t know, maybe 20% or 30% of the launch market, or 40%, but certainly not 90%. 

When you discuss creating a constellation of satellites — which also need to be upgraded every few years — to constitute an “orbital data center,” you’re simply promising much more work for your launch business. I can’t help but emphasize that point.

Kirsten: I’d like to quickly mention that [SpaceX’s] other significant business is renting out their compute, by the way. So, circling back to the chip discussion, we’ve come full circle.

Anthony: One of the recurring themes in this episode could be the idea of advocating your own interests. This isn’t a novel occurrence. Executives at tech firms, or any other organizations, project their visions for the future, ultimately favoring their business interests. 

However, this is something important to keep in mind as we engage in discussions about major AI companies, especially during this moment of profound uncertainty. We’re all contemplating: What will the job market look like in the future? What impacts will this have on the environment? What skills will I need to acquire? 

All these AI CEOs or investors have perspectives on those issues. It’s not that they’re incorrect or intentionally misleading; however, there’s always an asterisk beside these forecasts. In Musk’s situation, he’s discussing something that would greatly benefit SpaceX. In SoftBank’s case, they hold extensive investments in terrestrial data center initiatives. Sam Altman is another notable figure who has raised his eyebrows at the notion of orbital data centers — and, again, he and Elon Musk undoubtedly share a long and complicated history.

All this points to the fact that there are no unbiased, objective observers in this discussion. Everyone involved has their own baggage and significant financial stakes.

Conno Christou leaves nothing to randomness. He monitors his sleep utilizing a Whoop band, cross-references this with an Oura ring, and annually has nearly 100 biomarkers evaluated. For four years straight, he underwent annual blood tests, adhering to the guidelines set forth by longevity experts like Peter Attia and Rhonda Patrick. He was fine-tuning his supplements, circadian cycle, and protein consumption.

At 35, while establishing his second venture, he was as well-informed about the latest health research as anyone in his circle. His most recent health assessment in 2025 displayed positive results across all metrics. “It was the best I’d had in years,” he remarks.

Then, post-exercise, his arm began to swell.

Initially, he didn’t think much of it. It took a week before he consulted a doctor, who discovered two blood clots in his veins and arranged for surgery. However, the pre-operative tests altered everything. A doctor re-entered the room, informing him that the surgery would not be proceeding.

“We’ve identified an 11-by-11-by-8 centimeter mass behind your sternum,” the physician revealed.

A biopsy validated a reality Christou had never considered. He had an aggressive, rapidly developing type of non-Hodgkin’s lymphoma — a rare diagnosis occurring in roughly one in 420,000 individuals, resulting from a random genetic mutation unrelated to lifestyle, diet, or stress.

The tumor had only been present for approximately three months. In another three weeks, it would have progressed to stage four.

“Fortunate in my misfortune,” Christou shared with this editor this week from his residence in Athens, where he spends part of his time. “It was discovered solely because I sought treatment for something completely different.”

What ensued was an education in the constraints of the healthcare system, and the proactive measures a determined patient can undertake with the tools presently accessible.

His first oncologist, a reputable expert, proposed the less intensive of two chemotherapy options. Christou scheduled his first infusion for three days later. However, on the eve of the appointment, he sought another opinion.

The second physician was decisive. He advocated for the more aggressive regimen — continuous in-hospital infusion, cycling every three weeks over six months — considering Christou’s unique pathology. The lighter option had an approximate 60% success rate for his case. The more intense one raised that figure to around 85%. Two top-tier doctors with fundamentally opposing suggestions.

“As entrepreneurs, we steer the ship,” Christou articulates concerning the tendency of many individuals to accept what they’re told — and why more shouldn’t. “You encounter numerous opinions. You’re not obligated to adhere to the first recommendation.”

He didn’t simply choose to accept the second doctor’s guidance, either. In the subsequent two days, he procured 12 opinions overall — leveraging his professional connections, reaching out to hematologists and oncologists both domestically and internationally, calling in every favor possible. Eleven to one favored the more intense approach. He pursued it. The choice, he claims, felt more logical than courageous. As a data-driven individual, the stakes had become existential for him.

Throughout six months of treatment, Christou approached chemotherapy similarly to how he approached launching a business, as a series of marathon sprints — each with a predetermined cycle, filled with weekly data points. He had completed a compulsory 25-month military service in Cyprus at 18 and drew on that experience as well. He told himself he would be a diligent soldier. Trust the process. Six cycles. Get through it.

He wore his Whoop throughout the process and found it astonishingly accurate in predicting the days his immune system would be at its lowest, occasionally signaling them before symptoms appeared. He maintained a symptom diary using voice transcription, logging every change, side effect, medication, and counter-medication. He concentrated on three key factors: sleep, nutrition, and, most importantly, psychology. (“It influences the outcome more than anything,” Christou stated. “I never asked ‘why me’ — not once. That question yields no useful answer.”)

He input all of this — blood results, scan data, wearable readings, journal entries — into Claude. He’s not alone in utilizing chatbots for medical advice. A public opinion survey released in March indicated that a third of American adults now rely on them for health insights and guidance. Online accounts suggest that for some patients, AI is providing solutions that the traditional system failed to deliver.

Experts recommend caution; Danielle Bitterman, clinical lead for data science and AI at Mass General Brigham, has recently informed the New York Times that general-purpose chatbots often provide inaccurate information and “have not undergone thorough evaluation” for personalized diagnoses.

Christou agrees. “It didn’t substitute for the doctors,” he asserts, but it “assisted me in posing the right questions.”

For a condition as uncommon as his — one that an oncologist might encounter once annually — access to a model that had absorbed the entire scope of medical literature was, he insists, simply not equivalent to a Google search.

The model became essential toward the conclusion of treatment. His final PET scan — the imaging method used to identify active disease — returned inconclusive results. His oncologist began discussing a second line of treatment, potentially involving radiotherapy near his heart and lungs. It was a distressing turn of events.

Christou once again conducted thorough research. He learned that for this specific lymphoma, the false-positive rate on end-of-treatment PET scans hovers around 60% — a statistic that continues to astound him. “It’s 2026,” he remarks. “Sixty percent.”

He input all three of his PET scans and his MRI into Claude, which identified a known but often overlooked phenomenon: in patients under 40 recovering from this type of lymphoma, the thymus gland can reactivate post-chemotherapy, appearing on imaging as if active disease. Given his age and the characteristics of his scans, the model estimated the likelihood of that explanation at roughly 90%.

He sought three additional opinions. The fourth physician verified it: thymus rebound. There was no active disease. Radiotherapy was unnecessary. He was in the clear.

Christou is still working through what the past year has signified for his health, his work, and his perspective on time. He founded Keragon, his current venture, before any of this unfolded; it’s an AI-driven platform designed to assist medical practices in automating their administrative tasks.

However, experiencing the system as a patient has imparted him with a new understanding. He observed healthcare professionals overwhelmed with tasks unrelated to patient care. He received the same chemotherapy regimen as an 80-year-old woman, with side effects managed through a complicated sequence of additional medications, each introducing their own issues. He believes that we will look back at this phase of treatment with dismay.

Now, he generally takes Sundays off. He strives to be present — during lunch with friends, at home with his dog, in conversations that may have previously felt like interruptions to work. A VC friend told him years before something that he said he kept replaying during treatment: Be happy now. He admits it’s one of the hardest things to accomplish but he finally recognizes its significance.

He expresses willingness to converse with anyone going through similar experiences, to exchange insights, and compare stories. He seems earnest in this offer.

“It’s not something happening in ten years,” he remarks regarding what AI can already accomplish for patients ready to embrace it. “It’s happening now.”