ELON MUSK SAYS MOST NEOCLOUDS WILL DIE BUT THESE FOUR COULD THRIVE Elon Musk's latest warning about the neocloud industry is one of the most important developments investors should be paying attention to because while his prediction of widespread failures and consolidation across more than 200 operators sounds incredibly bearish, I actually think it could create an enormous competitive advantage for the handful of companies that have already figured out how to finance their AI infrastructure expansion. Musk argues that debt markets have almost entirely shut off financing to companies without meaningful revenue outside AI data centers, but I think the more important distinction is that financing isn't disappearing as much as it's becoming considerably more selective since lenders are increasingly distinguishing between operators borrowing against speculative future demand and those financing infrastructure backed by contracted revenue, customer prepayments and valuable assets that can serve as collateral. That's an important difference because a company with billions of dollars in commitments from Microsoft or Meta presents a very different credit profile than a smaller operator building GPU capacity without knowing who will ultimately rent it. In fact, I think this environment could ultimately favor four companies in particular, $NBIS, $IREN, $CRWV and $CIFR, although each has a very different business model and level of financial risk. What makes Nebius and IREN particularly interesting is that both have developed financing structures designed to reduce their dependence on traditional unsecured borrowing while continuing to expand their infrastructure at an extraordinary pace. Nebius finished Q2 with approximately $8B in cash, expects more than $9B in customer prepayments during 2026 and has already demonstrated its ability to secure asset-backed financing through a $775M facility supported by deployed GPUs and contracted cash flows from an investment-grade customer. With more than $40B in customer commitments, I think Nebius has a potentially repeatable financing model where customers effectively help fund the infrastructure they're purchasing, reducing the amount of external capital required for each deployment. IREN is approaching the same problem from a slightly different direction because its recent contracts include customer prepayments covering approximately 45% to 55% of GPU capital expenditures, while the remaining investment can be financed through loans secured by those GPUs. Its Microsoft agreement supported a $3.65B investment-grade financing facility at approximately 6%, while another $2.4B facility was arranged at a fixed 9% rate for deployments involving customers without investment-grade credit ratings. I think those transactions are particularly important because they demonstrate that even in a tightening credit environment, institutional lenders remain willing to finance AI infrastructure when the underlying economics, customer commitments and collateral are sufficiently attractive. Meanwhile, CoreWeave and Cipher demonstrate two additional ways companies can navigate this environment, although I think their risk profiles are materially different from those of Nebius and IREN. CoreWeave is arguably the best example of why investors need to distinguish between unsecured corporate borrowing and asset-backed project financing because while its unsecured bonds have traded at yields approaching 15%, the company has continued accessing financing through structures supported by GPU collateral and contracted cash flows from customers like Microsoft and Meta. Goldman Sachs recently highlighted how those financing arrangements can produce borrowing costs considerably below what distressed unsecured bond yields would suggest, while Nvidia's commitment to purchase unused capacity provides another layer of demand support. However, I still consider CoreWeave the most financially vulnerable of these four because its substantial leverage, enormous capital requirements and dependence on maintaining attractive GPU rental economics create meaningful refinancing and execution risks. Cipher, on the other hand, isn't really a traditional GPU neocloud because its business increasingly revolves around developing and leasing the physical infrastructure that AI operators need rather than directly purchasing GPUs and renting compute. Its Barber Lake facility now carries more than $9B in contracted revenue over 20 years, while its broader portfolio has approximately $16.6B in contracted revenue, creating a very different financing proposition where lenders can evaluate long duration leases and underlying real estate infrastructure rather than relying primarily on the economic life of rapidly depreciating GPUs. That doesn't eliminate construction overruns, financing risk or tenant concentration, but I think it gives Cipher a more infrastructure-oriented business model that could become increasingly attractive as capital markets prioritize predictable contracted cash flows. Ultimately, I think Musk is probably right that we're heading toward a meaningful consolidation across the neocloud industry, but I disagree with the idea that this automatically translates into a collapse in AI infrastructure demand or that every company without a large non-AI business is destined to struggle. If anything, I think the more likely outcome is that weaker operators without investment-grade customers, meaningful prepayments, secured financing arrangements or sufficient liquidity find themselves increasingly unable to fund expansion, while better-capitalized competitors gain opportunities to acquire valuable power capacity, data center infrastructure and customer relationships at more attractive prices. It's also worth remembering that Musk has an enormous strategic interest in how this industry develops because SpaceX is pursuing its own compute ambitions and would naturally benefit from competitors facing higher financing costs or struggling to expand. That doesn't invalidate his warning, but I think it's important to recognize that the company delivering the warning is also positioning itself to become one of the industry's biggest competitors. My broader takeaway is that investors should stop treating every neocloud as though it has the same balance sheet, customer quality and access to capital because $NBIS, $IREN, $CRWV and $CIFR represent four very different approaches to financing the same enormous infrastructure opportunity. None is immune to a prolonged credit squeeze, but I think the companies that can consistently demonstrate strong customer commitments, attractive project economics and access to financing will be in a much better position to survive consolidation and potentially emerge with considerably greater market share. In other words, I don't think the biggest question is whether the neocloud industry survives, but rather which companies emerge as the dominant operators once the weaker participants can no longer afford to compete.