Brad Thomas
Founder & Lead Analyst of Wide Moat Research
Fermi America asks investors to value a blueprint rather than a building, and expectations have outrun reality; shares are down more than 70% from the IPO. Mission-critical infrastructure with existing cash flows is preferable to paying today for possibilities that may not materialize until 2040. x.comÂ
Fermi has no revenue and is only promising a 17 MW Texas project by 2032. Its steep share-price decline looks like capital markets disciplining a speculative business. x.comÂ
Fermi carries a true chance of going to zero, so it can't be owned outright, even though it may well get a deal done in time. In a world of high energy costs and constant backlash against turning backyards into data centers, Fermi holds a strong hand. @rahmbotrades sees a good chance Amazon partners with Fermi for its Amarillo build-out, citing an updated Amazon job listing for a data center cluster operations leader. x.comÂ
Fermi's valuation has now fallen to where starting a position makes sense, because the market has finally discounted risks that were apparent more than a month earlier. Those risks were clear from conversations with other management teams and Texas regulators; @chinoalemano notes the market concluded only about a third of the potential 1,600MW for 2027 would actually happen. x.com x.comÂ
Higher financing and build costs squeeze short-term returns for neoclouds. Iran-linked fuel inflation raises Fed hike risk; even better projects may borrow in the low 7s, while weaker ones risk dilution or delays. CoreWeave, IREN, Nebius, New Era Energy & Digital and WhiteFiber need higher contract dollars per MW to offset costs. Funded projects with strong customers and liquidity fare better, but almost none can fund the next phase without easier markets. No need to panic. x.com x.comÂ
AI infrastructure providers' heavy capital needs and aging equipment may leave little cash for shareholders despite revenue growth. Depreciation is among their most underappreciated costs: shorter equipment lives raise it, while replacement and financing erode returns. Leverage and dilution magnify economic risks. Some may prosper, but until pricing power, depreciation, financing and long-term returns on capital are clearer, their investment appeal remains hard to share. x.com x.comÂ
Fermi is an overlooked way to play AI because almost no one is watching who powers the chips. Rick Perry's company has built an 11 GW private, behind-the-meter energy grid for AI hyperscalers, with a 6 GW Clean Air Permit, a nuclear pilot, $785M in new equipment financing and a $500M MUFG loan. At $9.5, the stock sits well below analyst targets of $20 or more, and two buy signals are in play. x.comÂ
Neoclouds can be an incredible business only if they turn delivered GPUs into revenue on time. Power delays are harder to remedy than commissioning issues; idle GPUs earn nothing while financing costs mount, especially for smaller operators and those reliant on colocation. Energized sites and vertical control help, but GPUs or secured power alone are not enough. IREN has financing secured and Nebius looks less exposed to Musk's debt warning; smaller players may struggle to fund expansion. x.com … Show more
Nebius has pricing power because its hard-to-build AI infrastructure and software serve enterprise clients who cannot find capacity elsewhere. It owns some data centers and land but has mostly operating leases; the share of active compute at owned sites is unspecified. It looks like the best infrastructure play despite slight skepticism about its path to profitability and valuation, and would work even if capex slowed 20% or the cycle paused. x.com x.com x.comÂ
Fermi's pullback on its $375 million convertible note looks like a chance to buy the fear, because the capital funds its gigawatt-scale Project Matador AI power campus in Texas. Capped calls limit dilution for shareholders up to $14.64 a share. Demand for dense, redundant power for hyperscalers keeps accelerating, with reports of possible major tenant interest such as OpenAI. Analyst targets sit well above the price, and with the proxy fight over, management can focus on execution. x.comÂ
IREN looks better valued than failed IPO candidate Firmus on operating data-center capacity: it has twice Firmus's operational MW at half the valuation, and its operating capacity is heading to 300MW, about five times the current level, by year end. x.comÂ
