Fellow Degens bears and priced in Doomers, Ive been researching investing and gambling on the ai infrastructure sector for the past year. In fact I’ve spent so much time on the sector that my poor university grades have slipped as I spend more and more hours focused on dilution credit risk contracted computing and options pricing rather than financial planning database coding and fictitious internal machinery financing. All of this is to say that the sector is wildly irrational seemingly misplacing multiple areas with the risk being beared by debt riddled but operational companies “Coughweave”. Speaking on neoclouds specifically and regarding debt weaves bond pricing you can see multiple bears jump the gun and declare black Monday prematurely. For large scale defaults these companies still need to take on a hell of a lot more debt. Personally I don’t own any of the Neoclouds at the moment as after I moved from micron to nebius I cashed out at around 208$ my pick for infrastructure is significantly de risked comparatively and already delivering meaningful revenue but more on that later. Now if we look at the pure financial picture core weave is running a debt to equity of 10.27x but before you get the pitchforks 16 billion of the 50 billion in debt is long term leases on infrastructure not high yield debt. So while Coreweave is in heaps of debt only 10b is due in the next 15 months 4.4b in 2026 and 6b in 2027 (found in the 10Q) Notably to pay this debt they’re still receiving vast funding with another 14 billion in long term financing being made available this year which adds to the current 3.5b in ever increasing quarterly cash flow. Essentially what I’m saying here is as a bear your chances of a coreweave default in coming 2 year period is like betting on the housing market to collapse but in 2003 and as they famously say the market can stay irrational longer than you can remain solvent, from what were seeing in the media and the market since June, things can get far far more irrational than you can currently claim they are. Now for those familiar with the infrastructure space you will know of a type of company type of datacenter landlord. If not here’s the rundown, at the top of the ai money cycle is the Hyperscalers and the labs who need the compute, then you have the neoclouds and Hyperscalers filling the compute with the GPUS and finally the company’s supplying the data centre leases. They are the physical infrastructure power site and cooling that house the GPUS for the neoclouds and hyperscalers. Now these company’s APLD HUT8 WULF CIFR and CORZ move in direct relation to the neoclouds despite a glaring difference. They operate at incredibly high margins once sites and MWs are operational. The major glaring downside to the neoclouds is that they have to deal with GPU depreciation meaning they not only have to deal with their massive financing but they also have to get enough cash together to pay for more GPUs once the old ones are obsolete. My broader argument is that out of the AI infrastructure sector, the market is failing to differentiate between companies taking on significant technological and financing risks and those developing long lasting physical infrastructure backed by contracted revenue. This is where I think the bearish infrastructure argument begins to fall apart, particularly when looking at the timelines involved. Many of these companies are still in the construction phase of their vast multiple GW development pipelines, meaning their current financial statements reflect substantial capital expenditure without the full benefit of their future contracted revenue leading to them taking on large scale debt which is allowing the default risk idea to be sprung Take Applied Digital, for example. The company has secured long-term leases with hyperscale customers, yet its financial performance is still heavily influenced by the costs associated with constructing and financing its facilities. Applied Is down to $18 from its $50 high based purely on hype dying down and overblown macro factors. It has a market cap under 6 billion with nearly 12 billion in assets and 6 billion in debt. It has 600mw coming online and already fully financed which will produce over a billion dollars in yearly revenue at an 89% noi margin from q2 2027 so why is it priced like this ? Apld is much larger than its competitors being hut8 wulf cifr ect with more financed mw more delivered mw and a larger contract base 1.4 gw at 36 billion in revenue and a smallercloud business Chronoscale that serves to reap the benefits of cloud mw pricing with a contract from Microsoft. So why is APLD so devalued, my guess is the overblown risk associated with the 2 original customers being coreweave and oracle, but they have larger contracts with Meta which are incredibly low risk. This risk combines with the fact APLD will have to take on more debt to fund the meta sites, unfortunately for the bears this debt like almost all of APLDs debt isn’t due until after 2030 while they will reap rent revenue of over 2 and a half billion plus additional expansions from 2028 onwards. The market is systematically undervaluing APLD especially when Hut8 boasts a 2 billion dollar higher market cap despite having equal debt a smaller contract backlog and a canary in the coal mine for a tenant (Anthropic). The question Bears ask shouldn't simply be how much debt these companies have accumulated. It should be whether the cash flows generated by their contracted infrastructure will be sufficient to service that debt once the facilities become operational.This is why I think 2028 is such an important year for the sector. By then, a much larger proportion of contracted infrastructure should be operational, providing investors with a clearer picture of sustainable margins, debt coverage and the actual economics of these projects.That doesn't mean defaults are impossible before 2028. Companies with inadequate liquidity, uncontracted capacity or significant refinancing requirements could experience financial distress much sooner.But I think the market is making a mistake by extrapolating the risks associated with highly leveraged GPU operators across the entire AI infrastructure industry. I'm not arguing that AI infrastructure is cheap, nor that every company will survive. I'm arguing that the market is pricing certain infrastructure companies as though they are being run by inept fools and not experts who have projected these financials years in advance and know what they’re doing. TLDR Coreweave won’t go bankrupt tomorrow, Datacenter landlords are undervalued and imo the unicorns of the market with an actual road to profit for shareholders mirroring memory last year. APLD is beaten down despite great numbers and forecasts with 22% short float 4 days to cover and a Ceo and co founder who are both actively buying millions worth. positions are a couple thousand APLD shares and calls -Stay7089
CryptoAlerta — análise de criptomoedas e mercado em tempo real