The consensus is comfortable. It is also, in specific places, wrong.
HampTex builds its analysis on two proprietary methodologies, the Six Forces Framework and the Factory Model, and applies them with a single discipline: the firm diverges from industry consensus only where the divergent position is highly defensible against the most sophisticated counterparty in the room. The positions below are the ones the firm will defend, by name, against the institutions that hold the other view: Goldman Sachs, Bain & Company, McKinsey, Deloitte, JLL, and CBRE.
Announced capacity is not energized capacity
Roughly 150 GW of U.S. AI data center capacity has been announced through 2030. HampTex's analysis puts what actually energizes this cycle at 32 to 55 GW. The headline pipeline is read as a delivery schedule; it is a filter, and most of what enters it does not survive. Diverges from: the pipeline-as-schedule reading implicit across sell-side capacity forecasts.
Revenue utilization is 75%, not ~100%
Building Stack economics are commonly underwritten at the near-100% utilization implicit in headline IRR decks. HampTex underwrites a 75% time-weighted revenue utilization base case (60% downside, 90% upside), the fraction of contracted capacity actually generating rent across a 15-year hold. Pre-leasing at 100% is a snapshot of commitment, not a measure of rent collected over time. Diverges from: headline pre-investment underwriting; sits deliberately above Citi's 50 to 70% capacity-utilization range.
Compute revenue realizes at 65%, not in full
The market assumes planned Compute Stack capacity converts near-fully into AI services revenue by 2030. HampTex models 65% realization at base (50% bear, 80% bull), unifying five drag factors, energization delay, deployment ramp, workload utilization, hardware failure, and refresh, that consensus treats as separate tail risks. Diverges from: Bain, McKinsey, and Goldman growth forecasts that imply near-full conversion.
The SLA-to-realization gap
Hyperscalers project forward revenue against a 99.99 to 99.999% SLA availability standard. That standard is correct for uptime and wrong as a proxy for revenue realization: it is satisfied while delivery delays, ramp, utilization gaps, and failures all erode the revenue a given tranche of capex actually produces. Diverges from: the industry practice of treating availability and realization as the same number.
The IRR is 9 to 12%, not 12 to 18%
Once revenue is underwritten at a defensible utilization rather than the headline near-100%, the equity return compresses. The marketed 12 to 18% becomes 9 to 12%. The projects are still investable; they are simply not as mispriced upward as the pitch deck implies. Diverges from: the headline IRR ranges quoted in pre-investment materials.
The colocation feasibility benchmark
Financial feasibility turns on the realized lease rate per kW. The current benchmark, per CBRE's North America Data Center Trends (H2 2025), is a record $196.25 per kW per month for a 250 to 500 kW requirement in primary markets, up 6.6% year over year. This is an asking rate for colocation capacity, not a construction cost, and reading it as either revenue certainty or build cost is a common and material error. Source: CBRE, H2 2025. HampTex's divergence is on how the figure is applied in feasibility models, not on the figure itself.
Compute capital is 10 to 20x building capital
Across the useful life of one Building Stack facility, the Compute Stack capital cycled through it exceeds the building capital by roughly 10 to 20x, because GPUs refresh every 3 to 6 years while the shell depreciates over 20 to 30. The risk is concentrated where the depreciation is fastest. Consistent with: Man Group's March 2026 duration-mismatch analysis and the Chanos and Burry depreciation thesis; quantified here as a lifecycle ratio.
Energy is the binding constraint, not chips
HampTex separates the capital funding AI infrastructure into three pools, Compute, Construction, and Energy, and identifies Energy as the slowest at 36 to 84 months for new builds. It is the rate-limiter on the entire cycle, the one pool that cannot be accelerated by writing a larger check or running a tighter operation. Extends: McKinsey, JLL, Grid Strategies, and LBNL, who name power as a constraint; HampTex names it the binding one.
Behind-the-meter is not the escape valve
Roughly 48 GW of behind-the-meter capacity was announced by late 2025, a 24-fold jump in a year, and the market reads it as the way around the interconnection queue. It is not. BTM does not remove the Six Forces constraints; it relocates them. HampTex's realistic read of BTM realization is 13 to 22 GW against that 48 GW announced. Diverges from: S&P Global and Wood Mackenzie coverage treating BTM as a primary mitigation pathway.
The Factory dividend is worth ~$300B
Factory-model architecture, smaller, modular, energy-ready sites, recovers roughly 15 points of the compute realization gap versus Monument-class bespoke builds, a steady-state dividend of about $300B against Bain's $2T 2030 target, plus $2 to 3B per gigawatt in earlier revenue from staggered delivery. Extends: McKinsey's modular-construction findings from a cost-and-timeline lens to a revenue-realization one.
Smaller hyperscaler sites, in specific contexts
Sub-50 MW hyperscaler sites are emerging, but not as a wholesale shift away from gigawatt campuses. They appear in three specific contexts: inference edge, neocloud partnership clusters, and latency-bound placements. Precision here matters; the consensus either dismisses the trend or overstates it. Diverges from: commentary treating smaller sites as either marginal or as a wholesale market turn.
Community opposition is capital-material
Standard underwriting treats community and social opposition (Force 3) as a marginal residual risk handled by public affairs. HampTex documents it as a structurally binding force with capital-material exposure, and shows that firm-level practices, opacity, salami-sliced announcements, unmet commitments, erode credibility at the sector level, not just for the offending operator. Diverges from: project-finance and infrastructure-equity frameworks that score opposition as a checklist item.
HampTex maintains a documented defense of each position to a primary-source standard, structured for adversarial review by senior institutional partners. Capital allocators and operators evaluating these divergences against their own models are invited to engage directly. Request a briefing →