The Data Center Mirage: Why Speculation Fails and Energy Wins
Alex Davis of @disruptivetech put words to a concern that many people in infrastructure and AI have been quietly sharing, even while the hype machine keeps accelerating. His warning about speculative data center construction is not bearishness about AI. It is a sober assessment of how capital is being deployed, and where that deployment starts to break down when it is disconnected from fundamentals.
Davis frames the current moment well. AI is real. The opportunity is enormous. At the same time, too many business models are being justified by momentum rather than margins, by capex intensity rather than customer traction, and by financial engineering rather than operational reality. Nowhere is this more visible than in the speculative data center market.
The core of his argument is simple and uncomfortable. The “build it and they will come” strategy is a trap. Hyperscalers do not outsource their strategic compute long term. They own it. They integrate it deeply into their networks, power strategy, and software stack. When speculative landlords assume hyperscalers will eventually fill empty shells, they misunderstand how these buyers behave. The result is a growing pile of assets that look impressive on paper but lack durable demand.
Davis’s concern about a financing crisis in the 2027 to 2028 window follows naturally from that mismatch. Rising interest rates, delayed grid interconnections (or communities in outright revolt as we are seeing increasingly lately), and utilization that fails to meet underwriting assumptions create stress very quickly. When power is not secured, when tenants are not locked in, and when margins depend on perfect future conditions, the system becomes fragile. That fragility does not show up in pitch decks. It shows up when refinancing cycles hit.
An Axios piece recently highlighted these concerns in the context of Groq and the broader AI infrastructure boom, but the more important signal is not how Axios framed it. The signal is that experienced capital allocators like Davis are publicly drawing a line between durable infrastructure and speculative excess.
What comes next is the real question.
The future of AI infrastructure is not endless rows of generic data center shells waiting for someone else to make them useful. It is compute that is built with a clear purpose, tied directly to real workloads, and paired with energy strategies that are controlled rather than hoped for. Energy independence is not a philosophical preference. It is a structural requirement once AI inference, latency sensitivity, and power density are taken seriously.
This is where the conversation shifts from warning to direction.
Distributed, decentralized infrastructure brings compute and networking closer to where demand actually exists. AI inference does not need to live exclusively in massive centralized campuses. In many cases, it performs better when it is closer to users, networks, and data sources. Distributed architecture reduces single points of failure and avoids forcing every workload through the same grid constrained choke points.
@Vailinor addresses the other half of the equation. Compute without energy certainty is speculative by definition. Vailinor’s model centers on AI ready, renewable biomass data center infrastructure designed around long term ownership and integrated power strategy. These are not landlord assets waiting for tenants. They are owner aligned facilities built to serve defined workloads with energy secured from day one.
Taken together, this approach directly answers the concern Davis raised. Owner users over speculative landlords. Integrated energy over grid dependency. Real workloads over hopeful utilization curves.
AI is not the bubble. Poorly aligned infrastructure is.
Davis’s letter is a reminder that discipline still matters, even in a once in a generation technology shift. The next phase of AI infrastructure will reward those who understand that compute, power, and demand must be designed together. Energy independent compute capacity is not the future because it sounds good. It is the future because it is the only model that survives when the hype fades and the balance sheets are tested.
Read the article from @axios here: https://www.axios.com/2025/12/29/groq-alex-davis-data-center-concerns
Originally published on X by Jason Brink, Jan 1, 2026.